{"authors":[{"name":"Martin Stühmer","url":"https://daily-devops.net/authors/martin/"},{"name":"Jendrik Brack","url":"https://daily-devops.net/authors/jendrik/"}],"description":"Recent content in GitHub Copilot AI Code Assistant on Daily DevOps \u0026 .NET","favicon":"https://daily-devops.net/images/logo_hu_6465d873dfa490cf.png","feed_url":"https://daily-devops.net/tags/github-copilot/feed.json","home_page_url":"https://daily-devops.net/tags/github-copilot/","icon":"https://daily-devops.net/images/logo_hu_5926de77762241ba.png","items":[{"authors":[{"name":"Martin Stühmer","url":"https://daily-devops.net/authors/martin/"}],"content_html":"\u003cp\u003eEvery time you drop into a Claude Code session, it starts indexing your project. In a .NET solution, that means \u003ccode\u003ebin/\u003c/code\u003e and \u003ccode\u003eobj/\u003c/code\u003e directories first: tens of thousands of compiled IL files, PDB symbols, generated Razor views, and NuGet extraction caches that serve exactly zero purpose in an AI context window. Then come the test coverage reports, the generated code, the environment files. All of it lands in context unless you tell Claude otherwise.\u003c/p\u003e\n\u003cp\u003eSo you ask Claude how to control which files it can see.\u003c/p\u003e\n\u003cp\u003eAnd Claude tells you about \u003ccode\u003e.claudeignore\u003c/code\u003e.\u003c/p\u003e\n\u003cp\u003eOf course it does. The concept is obvious: \u003ccode\u003e.gitignore\u003c/code\u003e keeps files out of version control, \u003ccode\u003e.dockerignore\u003c/code\u003e keeps files out of Docker build context, so naturally \u003ccode\u003e.claudeignore\u003c/code\u003e keeps files out of Claude\u0026rsquo;s context. Clean, consistent, intuitive. You add it to the project root, list your secrets and production config files, commit it alongside your \u003ccode\u003e.gitignore\u003c/code\u003e, and get on with your day.\u003c/p\u003e\n\u003cp\u003eThe AI assistant is properly restricted. Your teammates are protected. The security audit will go smoothly.\u003c/p\u003e\n\u003cp\u003eOne small problem.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"the-problem-claude-hallucinated-its-own-feature\"\u003e\u003ca href=\"/posts/claudeignore-dotnet/#the-problem-claude-hallucinated-its-own-feature\" title=\"The Problem: Claude Hallucinated Its Own Feature\"\u003eThe Problem: Claude Hallucinated Its Own Feature\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eHere\u0026rsquo;s what actually happened. Someone asked Claude how to prevent it from touching third-party licensed code in their repository. Claude, being helpful, explained that Claude Code supports a \u003ccode\u003e.claudeignore\u003c/code\u003e file with the same syntax as \u003ccode\u003e.gitignore\u003c/code\u003e: just drop it in the project root, and Claude will leave those files alone.\u003c/p\u003e\n\u003cp\u003eExcept that feature doesn\u0026rsquo;t exist. Claude invented it.\u003c/p\u003e\n\u003cp\u003eThe hallucination spread remarkably well. It turned up in blog posts, Stack Overflow answers, Reddit threads, and engineering wikis. Claude then read those discussions, concluded that \u003ccode\u003e.claudeignore\u003c/code\u003e must be real because everyone was talking about it, and started recommending it again. Confidently. Every time. An AI coding assistant invented a capability it doesn\u0026rsquo;t have, the made-up documentation spread across the internet, and now the AI keeps training on that documentation to reinvent the same capability. You couldn\u0026rsquo;t design a better hallucination feedback loop if you tried.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"how-a-false-sense-of-security-spread\"\u003e\u003ca href=\"/posts/claudeignore-dotnet/#how-a-false-sense-of-security-spread\" title=\"How A False Sense Of Security Spread\"\u003eHow A False Sense Of Security Spread\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eMeanwhile, developers were sleeping soundly. They had a \u003ccode\u003e.claudeignore\u003c/code\u003e in their repository. They had listed \u003ccode\u003eappsettings.Production.json\u003c/code\u003e. They had excluded \u003ccode\u003e.env.*\u003c/code\u003e. They had blocked \u003ccode\u003esecrets/\u003c/code\u003e. They had done the responsible thing, ticked the box, and assured their security teams the AI assistant was properly restricted.\u003c/p\u003e\n\u003cp\u003eProduction database connection strings: protected. API keys: safe. OAuth client secrets: completely off-limits.\u003c/p\u003e\n\u003cp\u003eNone of that was true. The file did nothing. It still does nothing. It is a text file with glob patterns that no tool reads.\u003c/p\u003e\n\u003cp\u003eTo be precise about the scope of the problem: \u003ccode\u003e.claudeignore\u003c/code\u003e is fiction, and it isn\u0026rsquo;t unique to Claude. Neither Claude Code, nor GitHub Copilot, nor OpenAI Codex offers a simple \u003ccode\u003e.claudeignore\u003c/code\u003e-style file at the project root. The mechanisms that actually exist are more fragmented, more verbose, tool-specific, and in several important cases considerably less capable than a plain ignore file would be.\u003c/p\u003e\n\u003cp\u003eHere\u0026rsquo;s what you\u0026rsquo;re actually working with.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"claude-code-permissionsdeny-in-settings\"\u003e\u003ca href=\"/posts/claudeignore-dotnet/#claude-code-permissionsdeny-in-settings\" title=\"Claude Code: permissions.deny in Settings\"\u003eClaude Code: \u003ccode\u003epermissions.deny\u003c/code\u003e in Settings\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eClaude Code\u0026rsquo;s real mechanism for restricting file access is the \u003ccode\u003epermissions\u003c/code\u003e section in \u003ccode\u003e.claude/settings.json\u003c/code\u003e. Instead of a clean, readable standalone file, you get JSON with tool permission syntax:\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-json\" data-lang=\"json\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e  \u003cspan class=\"nt\"\u003e\u0026#34;permissions\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e:\u003c/span\u003e \u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"nt\"\u003e\u0026#34;deny\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e:\u003c/span\u003e \u003cspan class=\"p\"\u003e[\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e      \u003cspan class=\"s2\"\u003e\u0026#34;Read(./.env)\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e      \u003cspan class=\"s2\"\u003e\u0026#34;Read(./.env.*)\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e      \u003cspan class=\"s2\"\u003e\u0026#34;Read(./secrets/**)\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e      \u003cspan class=\"s2\"\u003e\u0026#34;Read(./appsettings.Production.json)\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e      \u003cspan class=\"s2\"\u003e\u0026#34;Read(./appsettings.Staging.json)\u0026#34;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e]\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e  \u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eEach entry in the \u003ccode\u003edeny\u003c/code\u003e array is a tool call pattern. \u003ccode\u003eRead(path)\u003c/code\u003e denies Claude Code\u0026rsquo;s Read tool access to the matched path. The path supports glob patterns, so \u003ccode\u003eRead(./secrets/**)\u003c/code\u003e blocks the entire secrets directory tree.\u003c/p\u003e\n\u003cp\u003eThe settings file comes in two variants. \u003ccode\u003e.claude/settings.json\u003c/code\u003e is committed to version control and applies to the whole team. \u003ccode\u003e.claude/settings.local.json\u003c/code\u003e is personal, automatically gitignored by Claude Code when created. For security-relevant exclusions, use \u003ccode\u003esettings.json\u003c/code\u003e so restrictions are consistent across the team.\u003c/p\u003e\n\u003cp\u003eA practical baseline for .NET projects:\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-json\" data-lang=\"json\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e  \u003cspan class=\"nt\"\u003e\u0026#34;permissions\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e:\u003c/span\u003e \u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"nt\"\u003e\u0026#34;deny\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e:\u003c/span\u003e \u003cspan class=\"p\"\u003e[\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e      \u003cspan class=\"s2\"\u003e\u0026#34;Read(./.env)\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e      \u003cspan class=\"s2\"\u003e\u0026#34;Read(./.env.*)\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e      \u003cspan class=\"s2\"\u003e\u0026#34;Read(./secrets.json)\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e      \u003cspan class=\"s2\"\u003e\u0026#34;Read(./appsettings.Production.json)\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e      \u003cspan class=\"s2\"\u003e\u0026#34;Read(./appsettings.Staging.json)\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e      \u003cspan class=\"s2\"\u003e\u0026#34;Read(./**/bin/**)\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e      \u003cspan class=\"s2\"\u003e\u0026#34;Read(./**/obj/**)\u0026#34;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e]\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e  \u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\n\n\n\n\u003ch3 id=\"why-deny-rules-are-not-filesystem-blackouts\"\u003e\u003ca href=\"/posts/claudeignore-dotnet/#why-deny-rules-are-not-filesystem-blackouts\" title=\"Why Deny Rules Are Not Filesystem Blackouts\"\u003eWhy Deny Rules Are Not Filesystem Blackouts\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eNow here\u0026rsquo;s the part most writeups about this topic quietly skip: \u003cstrong\u003e\u003ccode\u003epermissions.deny\u003c/code\u003e rules are tool-level controls, not filesystem-level blackouts.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003ccode\u003eRead(./bin/**)\u003c/code\u003e blocks Claude Code\u0026rsquo;s Read tool from opening files in \u003ccode\u003ebin/\u003c/code\u003e. But Glob and Grep are separate tools, and \u003ccode\u003eRead\u003c/code\u003e deny rules don\u0026rsquo;t cover them. Claude can still discover those paths and search their contents. It just can\u0026rsquo;t read a file directly.\u003c/p\u003e\n\u003cp\u003eI asked Claude Code directly whether it had access to \u003ccode\u003ebin/\u003c/code\u003e folders after a \u003ccode\u003eRead\u003c/code\u003e deny rule was active. The answer: yes, Glob found them just fine.\u003c/p\u003e\n\u003cp\u003eSo \u003ccode\u003epermissions.deny\u003c/code\u003e alone does not make a directory invisible. For build artifacts like \u003ccode\u003ebin/\u003c/code\u003e and \u003ccode\u003eobj/\u003c/code\u003e, the more reliable mechanism is the \u003ccode\u003e.gitignore\u003c/code\u003e file you already have. Claude Code respects \u003ccode\u003e.gitignore\u003c/code\u003e by default (\u003ccode\u003erespectGitignore: true\u003c/code\u003e), and that exclusion applies across all discovery mechanisms, not just the Read tool. In practice, directories already in \u003ccode\u003e.gitignore\u003c/code\u003e are not surfaced by Glob or Grep in normal operation. Adding \u003ccode\u003eRead\u003c/code\u003e deny rules on top gives you a second layer, but \u003ccode\u003e.gitignore\u003c/code\u003e is doing the heavier lifting.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"what-this-means-for-secrets-in-the-repo\"\u003e\u003ca href=\"/posts/claudeignore-dotnet/#what-this-means-for-secrets-in-the-repo\" title=\"What This Means For Secrets In The Repo\"\u003eWhat This Means For Secrets In The Repo\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eFor \u003cstrong\u003esecrets and credentials\u003c/strong\u003e specifically, this distinction matters in a way that should make you uncomfortable. A production config file sitting in the repository root is not protected by a \u003ccode\u003eRead\u003c/code\u003e deny rule the way you might assume. Claude can still see that the file exists, discover its path with Glob, and match content patterns against it with Grep. The deny rule prevents reading the full file contents, but the file remains visible. The real answer, as always, is not having that file in the repository at all.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"github-copilot-exists-but-read-the-fine-print\"\u003e\u003ca href=\"/posts/claudeignore-dotnet/#github-copilot-exists-but-read-the-fine-print\" title=\"GitHub Copilot: Exists, But Read the Fine Print\"\u003eGitHub Copilot: Exists, But Read the Fine Print\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eGitHub Copilot does have a content exclusion feature. Before you get optimistic: it requires \u003cstrong\u003eCopilot Business or Copilot Enterprise\u003c/strong\u003e. Individual, Free, and Pro tiers don\u0026rsquo;t get it.\u003c/p\u003e\n\u003cp\u003eIt also isn\u0026rsquo;t a file you commit to your repository. There\u0026rsquo;s no \u003ccode\u003e.copilotignore\u003c/code\u003e. Exclusions are configured through GitHub\u0026rsquo;s web UI at the repository, organization, or enterprise level, and the configuration lives on GitHub\u0026rsquo;s end, not yours. If you want to know what\u0026rsquo;s excluded in a given repository, you navigate through GitHub\u0026rsquo;s admin interface to find out.\u003c/p\u003e\n\u003cp\u003eThe configuration format at least uses familiar YAML path patterns:\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-yaml\" data-lang=\"yaml\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"s2\"\u003e\u0026#34;*\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e:\u003c/span\u003e\u003cspan class=\"w\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"w\"\u003e  \u003c/span\u003e- \u003cspan class=\"s2\"\u003e\u0026#34;**/.env\u0026#34;\u003c/span\u003e\u003cspan class=\"w\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"w\"\u003e  \u003c/span\u003e- \u003cspan class=\"s2\"\u003e\u0026#34;**/appsettings.Production.json\u0026#34;\u003c/span\u003e\u003cspan class=\"w\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"w\"\u003e  \u003c/span\u003e- \u003cspan class=\"s2\"\u003e\u0026#34;**/secrets/**\u0026#34;\u003c/span\u003e\u003cspan class=\"w\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\n\n\n\n\u003ch3 id=\"why-copilot-exclusions-skip-agent-mode\"\u003e\u003ca href=\"/posts/claudeignore-dotnet/#why-copilot-exclusions-skip-agent-mode\" title=\"Why Copilot Exclusions Skip Agent Mode\"\u003eWhy Copilot Exclusions Skip Agent Mode\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eHere\u0026rsquo;s the part that really matters: these restrictions apply when Copilot is accessed through the IDE extension for code completion and inline chat. \u003cstrong\u003eThey do not apply to Agent mode, Copilot CLI, or the cloud agent.\u003c/strong\u003e GitHub states this explicitly in their documentation.\u003c/p\u003e\n\u003cp\u003eThe agentic workflows are exactly the ones where unrestricted file access is most concerning. They\u0026rsquo;re also the ones the exclusion feature doesn\u0026rsquo;t cover.\u003c/p\u003e\n\u003cp\u003eGitHub\u0026rsquo;s content exclusion is a governance feature for enterprise administrators, not a developer-facing tool for managing what an agent can see. The use case it serves is \u0026ldquo;prevent this entire codebase from reaching the AI at the org level.\u0026rdquo; The use case it doesn\u0026rsquo;t serve is \u0026ldquo;control what the agent accesses when I\u0026rsquo;m actively using it.\u0026rdquo;\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"openai-codex-nothing\"\u003e\u003ca href=\"/posts/claudeignore-dotnet/#openai-codex-nothing\" title=\"OpenAI Codex: Nothing\"\u003eOpenAI Codex: Nothing\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eAs of May 2026, OpenAI Codex has no built-in mechanism for file exclusion. Codex is an API. What goes into the context window is entirely the responsibility of the client making the call. If you\u0026rsquo;re using Codex through a third-party tool or IDE integration, that tool may or may not have its own ignore mechanism. The model itself is indifferent to what you send it.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"what-this-means-for-net-projects\"\u003e\u003ca href=\"/posts/claudeignore-dotnet/#what-this-means-for-net-projects\" title=\"What This Means for .NET Projects\"\u003eWhat This Means for .NET Projects\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eGiven the actual state of tooling, here\u0026rsquo;s where that leaves you:\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOn Claude Code\u003c/strong\u003e: use \u003ccode\u003epermissions.deny\u003c/code\u003e in \u003ccode\u003e.claude/settings.json\u003c/code\u003e, but understand you\u0026rsquo;re adding a tool-level restriction, not a filesystem blackout. Prioritize secrets and environment-specific config files. Build artifacts are better handled by \u003ccode\u003e.gitignore\u003c/code\u003e, which you almost certainly already have and which provides broader coverage across all discovery mechanisms.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOn GitHub Copilot Business or Enterprise\u003c/strong\u003e: configure content exclusions at the organization level for sensitive repositories if you need the governance coverage. Accept that none of this applies to agent-based workflows. For individual context management in the IDE, the exclusion feature is your only option. For agents, there is no equivalent.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOn Codex or API-based tools\u003c/strong\u003e: the exclusion problem is entirely at the integration layer. Read your tool\u0026rsquo;s documentation, not Codex\u0026rsquo;s.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"the-feature-that-should-exist-but-doesnt\"\u003e\u003ca href=\"/posts/claudeignore-dotnet/#the-feature-that-should-exist-but-doesnt\" title=\"The Feature That Should Exist But Doesn\u0026rsquo;t\"\u003eThe Feature That Should Exist But Doesn\u0026rsquo;t\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eThe \u003ccode\u003e.claudeignore\u003c/code\u003e hallucination spread so effectively because it named something developers genuinely need: a simple, version-controlled file that tells an AI coding assistant which files to leave alone. Same syntax as \u003ccode\u003e.gitignore\u003c/code\u003e. Lives in the project root. Gets committed like any other configuration. Readable by anyone on the team.\u003c/p\u003e\n\u003cp\u003eThat doesn\u0026rsquo;t exist. What exists instead is a JSON config with Read tool denial patterns that doesn\u0026rsquo;t cover Glob and Grep, a web-based admin feature that costs extra and doesn\u0026rsquo;t work for agents, and an API that does whatever the client tells it to.\u003c/p\u003e\n\u003cp\u003eThe gap between what developers want and what the tooling provides is real. Claude filling that gap by inventing a feature called \u003ccode\u003e.claudeignore\u003c/code\u003e is understandable: the concept is obvious, the need is legitimate, the name is perfect. It\u0026rsquo;s also deeply unhelpful, because developers have implemented \u003ccode\u003e.claudeignore\u003c/code\u003e files in their repositories, committed them, documented them in their onboarding guides, and assumed they were doing something. They weren\u0026rsquo;t.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"keep-credentials-out-of-the-repo\"\u003e\u003ca href=\"/posts/claudeignore-dotnet/#keep-credentials-out-of-the-repo\" title=\"Keep Credentials Out Of The Repo\"\u003eKeep Credentials Out Of The Repo\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003eThe actual security takeaway\u003c/strong\u003e: don\u0026rsquo;t put credentials in your repository. Azure Key Vault, user secrets during development, environment variable injection in CI/CD: these are not just best practices, they\u0026rsquo;re the only mechanisms that actually guarantee your production credentials don\u0026rsquo;t end up in an AI context window. \u003ccode\u003epermissions.deny\u003c/code\u003e and GitHub\u0026rsquo;s content exclusion are useful layers on top. They are not a substitute for that foundation.\u003c/p\u003e\n\u003cp\u003eThe \u003ccode\u003e.claudeignore\u003c/code\u003e situation is a concrete illustration of a broader rule worth internalizing: AI tools recommend nonexistent features with the same confident tone they use for everything else. A hallucinated configuration file and a correct one read exactly the same way. Verify against official documentation before you trust it, especially for anything security-adjacent.\u003c/p\u003e\n\u003cp\u003eThe official documentation for Claude Code\u0026rsquo;s actual file exclusion mechanism is at \u003ca href=\"https://code.claude.com/docs/en/settings#excluding-sensitive-files\" target=\"_blank\" rel=\"noopener external noreferrer\"\u003ecode.claude.com/docs/en/settings#excluding-sensitive-files\u003c/a\u003e. That\u0026rsquo;s where a search for \u003ccode\u003e.claudeignore\u003c/code\u003e should have taken you from the beginning.\u003c/p\u003e\n","date_modified":"2026-05-26T10:22:03+02:00","date_published":"2026-05-05T17:00:00+02:00","id":"https://daily-devops.net/posts/claudeignore-dotnet/","language":"en","summary":".claudeignore is a hallucination. Claude invented it, the internet spread it, and now Claude keeps recommending it. Here is what actually works in .NET.\n","tags":["ai-code-assistant","dotnet","github-copilot","devops","security"],"title":".claudeignore Doesn't Exist. Here's What Does.","url":"https://daily-devops.net/posts/claudeignore-dotnet/"},{"authors":[{"name":"Martin Stühmer","url":"https://daily-devops.net/authors/martin/"}],"content_html":"\u003cp\u003eIn \u003ca href=\"../code-sharpens-thinking/\"\u003ePart 1\u003c/a\u003e, we established that \u0026ldquo;vibe coding\u0026rdquo;—describing what you want and shipping what AI generates—creates productivity illusions that collapse spectacularly under production load. \u003ca href=\"../feedback-loop-ai-cant-replace/\"\u003ePart 2\u003c/a\u003e explored the feedback loop that AI can\u0026rsquo;t replicate.\u003c/p\u003e\n\u003cp\u003eNow we confront the practical question: \u003cstrong\u003eWhat skills define real professionals when typing code becomes trivial?\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAI code assistants accelerate the mechanical part extraordinarily well. GitHub Copilot autocompletes functions. ChatGPT generates entire APIs from prompts. The typing is handled.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eYet you remain indispensable.\u003c/strong\u003e Not in spite of AI\u0026rsquo;s code generation capabilities, but \u003cstrong\u003ebecause of them\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWhy?\u003c/strong\u003e When code generation becomes commoditized, the differentiator isn\u0026rsquo;t typing speed. It\u0026rsquo;s accumulated experience. Watching systems fail in production. Understanding \u003cstrong\u003ewhy\u003c/strong\u003e they failed. Applying that hard-won knowledge to prevent the next failure.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHere\u0026rsquo;s the uncomfortable truth:\u003c/strong\u003e Organizations that confuse \u0026ldquo;lines of code generated\u0026rdquo; with \u0026ldquo;productivity\u0026rdquo; discover the difference when production incidents spike—and the bill arrives.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"why-prompt-engineering-isnt-architecture\"\u003e\u003ca href=\"/posts/real-professional-software-engineering-ai-era/#why-prompt-engineering-isnt-architecture\" title=\"Why Prompt Engineering Isn\u0026rsquo;t Architecture\"\u003eWhy Prompt Engineering Isn\u0026rsquo;t Architecture\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003e\u003cstrong\u003eAI code generation creates a seductive trap.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eYou think:\u003c/strong\u003e If I can describe what I want in natural language and get working code, isn\u0026rsquo;t that sufficient? Why spend time understanding implementation details when AI handles them?\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHere\u0026rsquo;s why that\u0026rsquo;s wrong:\u003c/strong\u003e Prompts describe intent. Not constraints.\u003c/p\u003e\n\u003cp\u003eAnd software engineering? It\u0026rsquo;s fundamentally about managing \u003cstrong\u003econstraints\u003c/strong\u003e. Performance budgets. Memory limits. Concurrency safety. Error handling. Maintainability. Operational cost. Security boundaries.\u003c/p\u003e\n\u003cp\u003eConsider asking an AI to \u0026ldquo;implement caching for customer data.\u0026rdquo; You\u0026rsquo;ll get code that caches. But you \u003cstrong\u003ewon\u0026rsquo;t\u003c/strong\u003e get answers to:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eWhat\u0026rsquo;s the \u003cstrong\u003ememory budget\u003c/strong\u003e? When does caching become more expensive than repeated database calls?\u003c/li\u003e\n\u003cli\u003eHow do you handle \u003cstrong\u003ecache invalidation\u003c/strong\u003e across multiple application instances?\u003c/li\u003e\n\u003cli\u003eWhat\u0026rsquo;s the \u003cstrong\u003econsistency model\u003c/strong\u003e? Can stale data cause correctness issues downstream?\u003c/li\u003e\n\u003cli\u003eHow do you \u003cstrong\u003emonitor\u003c/strong\u003e cache hit rates to verify it\u0026rsquo;s actually improving performance?\u003c/li\u003e\n\u003cli\u003eWhat happens during \u003cstrong\u003ecache warming\u003c/strong\u003e? Do users experience degraded performance on cold starts?\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003eAI generates code that addresses the prompt. Professionals understand these questions emerge from production experience\u003c/strong\u003e—from watching systems fail, from debugging race conditions at 3 AM, from analyzing cost reports that show caching is more expensive than the problem it solved, from responding to incidents where stale cache data caused customer-visible bugs.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePrompt engineering optimizes for generating code quickly. Software architecture optimizes for systems that survive production reality. These are orthogonal skills.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eI\u0026rsquo;ve seen teams adopt AI-heavy workflows where \u003cstrong\u003ejunior developers generate features rapidly using prompts\u003c/strong\u003e, and \u003cstrong\u003esenior developers spend weeks later refactoring the accumulated technical debt\u003c/strong\u003e. The AI-generated code worked in isolation. It failed as a system because no one understood how the pieces interacted, what assumptions each component made, or where performance would degrade under load.\u003c/p\u003e\n\u003cp\u003eThe skill that AI can\u0026rsquo;t replace: \u003cstrong\u003erecognizing which questions to ask before writing code\u003c/strong\u003e, not generating syntax after questions are answered. That recognition comes from the feedback loop—you write code, watch it fail, understand \u003cstrong\u003ewhy\u003c/strong\u003e it failed, and internalize the lesson.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePrompt-driven development skips this loop entirely, outsourcing both the implementation and the learning.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eReal professionals don\u0026rsquo;t reject AI tools. They use them to accelerate the mechanical parts while maintaining ownership of the architectural decisions, performance analysis, and failure mode understanding that prompts can\u0026rsquo;t capture.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"technical-debt-where-abstract-design-becomes-concrete-burden\"\u003e\u003ca href=\"/posts/real-professional-software-engineering-ai-era/#technical-debt-where-abstract-design-becomes-concrete-burden\" title=\"Technical Debt: Where Abstract Design Becomes Concrete Burden\"\u003eTechnical Debt: Where Abstract Design Becomes Concrete Burden\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eTechnical debt is abstract thinking\u0026rsquo;s deferred consequences manifesting as maintenance burden. Design decisions that felt reasonable in isolation accumulate into complexity that resists change, harbors bugs, and drains productivity.\u003c/p\u003e\n\u003cp\u003eEvery architecture discussion includes statements like \u0026ldquo;we\u0026rsquo;ll refactor later\u0026rdquo; or \u0026ldquo;this is temporary\u0026rdquo; or \u0026ldquo;once we prove the concept, we\u0026rsquo;ll clean it up.\u0026rdquo; These are thought patterns that treat code as temporary scaffolding rather than operational reality. Code doesn\u0026rsquo;t stay temporary—it becomes production reality that teams maintain for years.\u003c/p\u003e\n\u003cp\u003eI\u0026rsquo;ve inherited codebases where \u0026ldquo;temporary\u0026rdquo; solutions from 2015 still run in production, calcified by dependencies and surrounded by defensive code that works around their limitations. The abstract thinking that justified shortcuts—\u0026ldquo;we\u0026rsquo;re moving fast,\u0026rdquo; \u0026ldquo;we\u0026rsquo;ll fix it in v2\u0026rdquo;—never accounted for the operational reality: v2 got deprioritized, teams changed, knowledge evaporated, and the technical debt persisted.\u003c/p\u003e\n\u003cp\u003eMicrosoft\u0026rsquo;s own guidance on technical debt management emphasizes measurement and prioritization based on impact—not on abstract severity, but on actual operational burden:\u003c/p\u003e\n\u003cblockquote\u003e\n\u003cp\u003e\u0026ldquo;Prioritize technical debt items based on their effects on workload functionality. Focus on addressing the issues that have the most significant effect on the performance, maintainability, and scalability of the workload.\u0026rdquo;\u003c/p\u003e\n\u003c/blockquote\u003e\n\u003cp\u003eThis requires executable code that can be measured, profiled, and analyzed. Abstract architectural concerns translate into concrete technical debt only when code exists to evaluate. You can\u0026rsquo;t measure maintainability, performance impact, or operational cost without code that runs in production-like conditions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAI-accelerated development amplifies this pattern.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWhen junior developers generate features using prompts, the code works immediately but accumulates technical debt invisibly. The AI optimized for \u0026ldquo;works now,\u0026rdquo; not \u0026ldquo;maintainable long-term.\u0026rdquo;\u003c/p\u003e\n\u003cp\u003eSix months later, when requirements change? \u003cstrong\u003eThe bill comes due.\u003c/strong\u003e What took 2 days to generate takes 2 weeks to refactor. Why? Because no one understands the generated foundations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eReal cost:\u003c/strong\u003e Senior developers spending 40+ hours untangling AI-generated code instead of building new features. That\u0026rsquo;s €4,000-8,000 in lost productivity—per feature.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"real-professionals-in-the-ai-era-mastering-the-feedback-loop\"\u003e\u003ca href=\"/posts/real-professional-software-engineering-ai-era/#real-professionals-in-the-ai-era-mastering-the-feedback-loop\" title=\"Real Professionals in the AI Era: Mastering the Feedback Loop\"\u003eReal Professionals in the AI Era: Mastering the Feedback Loop\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003e\u003cstrong\u003eDavid\u0026rsquo;s comment about real professionals not being replaced wasn\u0026rsquo;t wishful thinking or gatekeeping.\u003c/strong\u003e It was recognition that professional software engineering has \u003cstrong\u003enever\u003c/strong\u003e been about typing code—and in an era where typing is automated, that distinction becomes \u003cstrong\u003ebrutally\u003c/strong\u003e clear.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThe skills that define professionals in 2026 and beyond:\u003c/strong\u003e\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"understanding-execution-characteristics\"\u003e\u003ca href=\"/posts/real-professional-software-engineering-ai-era/#understanding-execution-characteristics\" title=\"Understanding Execution Characteristics\"\u003eUnderstanding Execution Characteristics\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eWhen AI generates code, professionals can read it and immediately recognize:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eAllocation patterns that will cause garbage collection pressure\u003c/li\u003e\n\u003cli\u003eDatabase access patterns that create N+1 problems\u003c/li\u003e\n\u003cli\u003eSynchronization primitives that risk deadlocks\u003c/li\u003e\n\u003cli\u003eAPI contracts that will break under versioning\u003c/li\u003e\n\u003cli\u003eAbstractions that trade clarity for cleverness\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThis isn\u0026rsquo;t about memorizing syntax. It\u0026rsquo;s about pattern recognition from seeing thousands of implementations and their production consequences. AI can generate the code. Professionals can predict where it fails before deployment.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"asking-questions-ai-cant-formulate\"\u003e\u003ca href=\"/posts/real-professional-software-engineering-ai-era/#asking-questions-ai-cant-formulate\" title=\"Asking Questions AI Can\u0026rsquo;t Formulate\"\u003eAsking Questions AI Can\u0026rsquo;t Formulate\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eAI optimizes for the prompt it receives. Professionals know which questions to ask before prompting:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eWhat\u0026rsquo;s the failure mode if this service is unavailable?\u003c/li\u003e\n\u003cli\u003eHow does this perform when the dataset grows 100x?\u003c/li\u003e\n\u003cli\u003eWhat happens during partial failures across service boundaries?\u003c/li\u003e\n\u003cli\u003eHow do we roll this back if production deployment reveals problems?\u003c/li\u003e\n\u003cli\u003eWhat operational metrics signal that this implementation is degrading?\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThese questions emerge from production scars, not documentation. They represent thinking that can\u0026rsquo;t be prompted because the prompt itself requires experience to formulate.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"recognizing-when-ai-solutions-are-wrong\"\u003e\u003ca href=\"/posts/real-professional-software-engineering-ai-era/#recognizing-when-ai-solutions-are-wrong\" title=\"Recognizing When AI Solutions Are Wrong\"\u003eRecognizing When AI Solutions Are Wrong\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eAI generates plausible code. Professionals recognize when plausible diverges from correct:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eThe generated caching looks reasonable but introduces race conditions\u003c/li\u003e\n\u003cli\u003eThe suggested refactoring breaks semantic guarantees the original code maintained\u003c/li\u003e\n\u003cli\u003eThe performance optimization trades correctness for speed\u003c/li\u003e\n\u003cli\u003eThe error handling silences failures that should propagate\u003c/li\u003e\n\u003cli\u003eThe abstraction solves the described problem but makes the actual problem harder\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThis skill—recognizing subtle wrongness—requires understanding not just what code does, but what it should do in context. AI has no context beyond the prompt. Professionals carry context from the entire system, the organization\u0026rsquo;s constraints, and production failure history.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"debugging-when-ai-generated-code-fails\"\u003e\u003ca href=\"/posts/real-professional-software-engineering-ai-era/#debugging-when-ai-generated-code-fails\" title=\"Debugging When AI-Generated Code Fails\"\u003eDebugging When AI-Generated Code Fails\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eAI can\u0026rsquo;t debug its own output effectively because it has no execution model. It can suggest changes based on error messages, but it can\u0026rsquo;t reason about:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eWhy the garbage collector is thrashing\u003c/li\u003e\n\u003cli\u003eWhere the memory leak originates across object graphs\u003c/li\u003e\n\u003cli\u003eWhy this specific race condition appears under production load but not in testing\u003c/li\u003e\n\u003cli\u003eHow this performance degradation emerged from the interaction of six separate components\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eProfessionals debug by understanding execution: what the CPU is doing, how memory is managed, where I/O blocking occurs, how the runtime schedules work. This understanding comes from the feedback loop—watching code execute, measuring behavior, correlating symptoms with causes.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"maintaining-code-ai-generated-yesterday\"\u003e\u003ca href=\"/posts/real-professional-software-engineering-ai-era/#maintaining-code-ai-generated-yesterday\" title=\"Maintaining Code AI Generated Yesterday\"\u003eMaintaining Code AI Generated Yesterday\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eThe code AI generates today becomes the maintenance burden of tomorrow. Professionals understand that maintainability isn\u0026rsquo;t syntax elegance—it\u0026rsquo;s whether future developers (including AI-assisted ones) can understand intent, modify behavior safely, and reason about consequences.\u003c/p\u003e\n\u003cp\u003eAI-generated code often optimizes for immediate functionality over long-term maintainability because prompts rarely include \u0026ldquo;make this easy to modify in six months when requirements change.\u0026rdquo; Professionals review AI output through the lens of future maintenance: Does this abstraction clarify or obscure? Will this pattern scale when similar features are added? Can someone unfamiliar with this code understand its failure modes?\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"the-economic-reality-of-ai-accelerated-development\"\u003e\u003ca href=\"/posts/real-professional-software-engineering-ai-era/#the-economic-reality-of-ai-accelerated-development\" title=\"The Economic Reality of AI-Accelerated Development\"\u003eThe Economic Reality of AI-Accelerated Development\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eAI tools make junior developers dramatically more productive at generating code. Sounds like pure upside, right?\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eUntil you measure the total lifecycle cost:\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eFeatures ship faster \u003cstrong\u003ebut\u003c/strong\u003e accumulate technical debt faster\u003c/li\u003e\n\u003cli\u003eCode coverage is high \u003cstrong\u003ebut\u003c/strong\u003e defect rates increase by 25-40%\u003c/li\u003e\n\u003cli\u003eDevelopment velocity looks impressive \u003cstrong\u003euntil\u003c/strong\u003e production incidents spike\u003c/li\u003e\n\u003cli\u003eRefactoring becomes more expensive because no one understands the AI-generated foundations\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003eTwo types of organizations:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eType 1 measures productivity by lines of code generated or features shipped per sprint. They see AI as a massive win.\u003c/p\u003e\n\u003cp\u003eType 2 measures productivity by system reliability, operational cost, and maintenance burden. They see a more complex picture—and higher total cost.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eYour value proposition shifts:\u003c/strong\u003e From \u0026ldquo;can write code\u0026rdquo; to \u0026ldquo;can ensure AI-generated code survives production.\u0026rdquo;\u003c/p\u003e\n\u003cp\u003eThat\u0026rsquo;s not a diminished role. \u003cstrong\u003eIt\u0026rsquo;s a more critical one.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe ability to generate code becomes commoditized. The ability to evaluate, refine, and maintain that code? \u003cstrong\u003eThat becomes your differentiator.\u003c/strong\u003e\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"why-the-feedback-loop-cant-be-automated\"\u003e\u003ca href=\"/posts/real-professional-software-engineering-ai-era/#why-the-feedback-loop-cant-be-automated\" title=\"Why the Feedback Loop Can\u0026rsquo;t Be Automated\"\u003eWhy the Feedback Loop Can\u0026rsquo;t Be Automated\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eAI can participate in parts of the feedback loop:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eIt can suggest implementations based on requirements\u003c/li\u003e\n\u003cli\u003eIt can generate tests based on code\u003c/li\u003e\n\u003cli\u003eIt can propose refactorings based on patterns\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eBut it can\u0026rsquo;t close the loop because closing the loop requires:\u003c/p\u003e\n\u003col\u003e\n\u003cli\u003e\u003cstrong\u003eExecution in realistic conditions\u003c/strong\u003e: Production load, real data volumes, actual failure scenarios\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eMeasurement of consequences\u003c/strong\u003e: Performance under stress, cost implications, operational burden\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eInterpretation of results\u003c/strong\u003e: Understanding why this metric degraded, why this pattern emerged, why this assumption failed\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eRefinement of thinking\u003c/strong\u003e: Updating mental models about what works, what fails, and why\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eApplication to future decisions\u003c/strong\u003e: Recognizing similar patterns in new contexts and avoiding repeated mistakes\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eAI can help with steps 1 and 2. Steps 3, 4, and 5 require human judgment informed by accumulated experience. This is the feedback loop David referenced—the mechanism that sharpens thinking through repeated collision with executable reality.\u003c/p\u003e\n\u003cp\u003eReal professionals master this loop. They write code (or review AI-generated code), watch it execute, measure its behavior, understand its failure modes, and refine their thinking. Each iteration strengthens their ability to recognize what will work before writing it, what will fail before deploying it, and what will cost more than it\u0026rsquo;s worth before building it.\u003c/p\u003e\n\u003cp\u003eThis skill can\u0026rsquo;t be replaced because it\u0026rsquo;s not about having the right answer immediately—it\u0026rsquo;s about knowing how to find the right answer through disciplined iteration between abstract thinking and concrete execution.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"conclusion-code-demands-honest-thinking\"\u003e\u003ca href=\"/posts/real-professional-software-engineering-ai-era/#conclusion-code-demands-honest-thinking\" title=\"Conclusion: Code Demands Honest Thinking\"\u003eConclusion: Code Demands Honest Thinking\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eYes, thinking is hard. Reasoning through constraints, evaluating trade-offs, understanding system dynamics—these require deep intellectual work. \u003cstrong\u003eI\u0026rsquo;ve never disputed this.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBut \u003cstrong\u003ehere\u0026rsquo;s what the \u0026ldquo;thinking is everything\u0026rdquo; narrative misses:\u003c/strong\u003e code is not just the mechanical output of that thinking. Code is the form that \u003cstrong\u003eforces thinking into honesty\u003c/strong\u003e. It\u0026rsquo;s where vague reasoning gets \u003cstrong\u003ebrutally exposed\u003c/strong\u003e, deferred decisions become \u003cstrong\u003eunavoidable\u003c/strong\u003e, and abstract consequences materialize as \u003cstrong\u003eoperational reality that costs real money and wakes you up at 3 AM\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eTreating code as \u0026ldquo;just another language\u0026rdquo; undersells what programming actually does: it transforms thought from abstract possibility into \u003cstrong\u003eexecutable certainty\u003c/strong\u003e. It makes performance \u003cstrong\u003emeasurable\u003c/strong\u003e, correctness \u003cstrong\u003etestable\u003c/strong\u003e, and complexity \u003cstrong\u003evisible\u003c/strong\u003e. It forces precision where thought allows \u003cstrong\u003ecomfortable ambiguity\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSoftware engineering isn\u0026rsquo;t thinking OR programming. It\u0026rsquo;s thinking made rigorous through programming.\u003c/strong\u003e It\u0026rsquo;s the tight feedback loop where abstract reasoning and executable verification sharpen each other iteratively.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOne without the other doesn\u0026rsquo;t scale:\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eThinking without executable form stays untested and \u003cstrong\u003eoften wrong\u003c/strong\u003e\u003c/li\u003e\n\u003cli\u003eCode without thoughtful design becomes \u003cstrong\u003eunmaintainable complexity\u003c/strong\u003e\u003c/li\u003e\n\u003cli\u003eAI-generated code without understanding becomes \u003cstrong\u003etechnical debt that compounds with every sprint\u003c/strong\u003e\u003c/li\u003e\n\u003cli\u003ePrompt engineering without production experience becomes \u003cstrong\u003ea liability dressed as productivity\u003c/strong\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eEngineering quality emerges from the discipline of moving between abstract reasoning and concrete implementation—\u003cstrong\u003erepeatedly, rigorously, honestly\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThat\u0026rsquo;s what makes software engineering difficult.\u003c/strong\u003e Not the typing. Not even just the thinking. But the intellectual discipline of forcing thought into executable form that \u003cstrong\u003esurvives contact with production reality\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eAI can type code faster than you. It can suggest implementations, generate tests, propose refactorings. \u003cstrong\u003eWhat it can\u0026rsquo;t do is learn from watching systems fail in production, understand why they failed, and apply that hard-won knowledge to prevent the next failure.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAnd that discipline, the feedback loop David referenced, cannot be replaced.\u003c/strong\u003e\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"series-summary\"\u003e\u003ca href=\"/posts/real-professional-software-engineering-ai-era/#series-summary\" title=\"Series Summary\"\u003eSeries Summary\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003e\u003cstrong\u003ePart 1: \u003ca href=\"../code-sharpens-thinking/\"\u003eWhy Real Professionals Will Never Be Replaced by AI\u003c/a\u003e\u003c/strong\u003e\u003cbr\u003e\nEstablished that AI-generated code without understanding creates productivity illusions. Vibe coding collapses when code generation becomes trivial and understanding execution, failure modes, and operational cost becomes everything.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePart 2: \u003ca href=\"../feedback-loop-ai-cant-replace/\"\u003eThe Feedback Loop That AI Can\u0026rsquo;t Replace\u003c/a\u003e\u003c/strong\u003e\u003cbr\u003e\nExamined the mechanisms that transform abstract thinking into operational understanding: compilers validate logic, tests expose behavioral gaps, profilers measure performance reality, production reveals deferred decisions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePart 3: Real Professional Software Engineering in the AI Era\u003c/strong\u003e (this article)\u003cbr\u003e\nExplored the irreplaceable professional skillset: recognizing execution characteristics, asking questions AI can\u0026rsquo;t formulate, debugging failures AI can\u0026rsquo;t reason about, maintaining code AI generated yesterday, and understanding the economic reality where \u0026ldquo;AI productivity\u0026rdquo; often means faster technical debt accumulation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThe throughline:\u003c/strong\u003e Real professionals will never be replaced because they\u0026rsquo;ve mastered the feedback loop: the iterative discipline of writing code, watching it fail, understanding why, and refining thinking. AI participates in parts of this loop but can\u0026rsquo;t close it. That\u0026rsquo;s where professionals remain indispensable.\u003c/p\u003e\n","date_modified":"2026-05-25T22:06:34+02:00","date_published":"2026-01-20T17:00:00+01:00","id":"https://daily-devops.net/posts/real-professional-software-engineering-ai-era/","language":"en","summary":"AI generates code instantly. Professionals spot when it is subtly wrong, debug failures AI cannot reason about, and see through the productivity narrative.\n","tags":["softwareengineering","codequality","bestpractices","architecture","dotnet","csharp","technicaldebt","ai-code-assistant","github-copilot"],"title":"Real Professional Software Engineering in the AI Era\n","url":"https://daily-devops.net/posts/real-professional-software-engineering-ai-era/"},{"authors":[{"name":"Martin Stühmer","url":"https://daily-devops.net/authors/martin/"}],"content_html":"\u003cp\u003eIn \u003ca href=\"../code-sharpens-thinking/\"\u003ePart 1 of this series\u003c/a\u003e, we explored why AI code generation creates an illusion of productivity that collapses when \u0026ldquo;vibe coding\u0026rdquo; meets production reality.\u003c/p\u003e\n\u003cp\u003eTyping code is now trivial. AI handles it faster than humans can type.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBut here\u0026rsquo;s the critical skill:\u003c/strong\u003e Understanding what that code costs. Where it fails. Why it breaks under load.\u003c/p\u003e\n\u003cp\u003eThe differentiator between professionals and prompt engineers? \u003cstrong\u003eThe feedback loop.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eYou write code (or review AI-generated code). Watch it execute. Measure its behavior. Understand its failure modes. Refine your thinking. Each iteration sharpens your ability to recognize what will work before implementing it, what will fail before deploying it, and what will cost more than it\u0026rsquo;s worth before building it.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSo what exactly is this feedback loop? And why can\u0026rsquo;t AI replicate it?\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis article examines the mechanisms that transform abstract thinking into operational understanding:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e\u003cstrong\u003eCompilers\u003c/strong\u003e that validate logical consistency and force completeness\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003ePerformance profilers\u003c/strong\u003e that expose what abstract analysis defers\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eTesting frameworks\u003c/strong\u003e that reveal behavioral gaps\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eProduction environments\u003c/strong\u003e that materialize every deferred decision\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThese aren\u0026rsquo;t just development tools—they\u0026rsquo;re thinking validators that expose where reasoning was incomplete. AI can participate in parts of this loop, but it can\u0026rsquo;t close it. Understanding why reveals why real professionals remain irreplaceable.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"the-compiler-as-thought-validator\"\u003e\u003ca href=\"/posts/feedback-loop-ai-cant-replace/#the-compiler-as-thought-validator\" title=\"The Compiler as Thought Validator\"\u003eThe Compiler as Thought Validator\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eModern compilers do more than translate syntax—they validate logical consistency. Static analysis, type checking, nullability analysis, and pattern exhaustiveness checks all function as automated reasoning validators. They catch the gaps that pure thought leaves unresolved.\u003c/p\u003e\n\u003cp\u003eConsider exhaustive pattern matching introduced in C# 8:\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-csharp\" data-lang=\"csharp\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"kd\"\u003eenum\u003c/span\u003e \u003cspan class=\"n\"\u003eOrderStatus\u003c/span\u003e \u003cspan class=\"p\"\u003e{\u003c/span\u003e \u003cspan class=\"n\"\u003ePending\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003eConfirmed\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003eShipped\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003eDelivered\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003eCancelled\u003c/span\u003e \u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"kt\"\u003estring\u003c/span\u003e \u003cspan class=\"n\"\u003eGetStatusMessage\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eOrderStatus\u003c/span\u003e \u003cspan class=\"n\"\u003estatus\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"n\"\u003estatus\u003c/span\u003e \u003cspan class=\"k\"\u003eswitch\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"n\"\u003eOrderStatus\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003ePending\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"s\"\u003e\u0026#34;Order is pending\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"n\"\u003eOrderStatus\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eConfirmed\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"s\"\u003e\u0026#34;Order confirmed\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"n\"\u003eOrderStatus\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eShipped\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"s\"\u003e\u0026#34;Order shipped\u0026#34;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"c1\"\u003e// Compiler error CS8509: The switch expression does not handle all possible values\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e};\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eThe compiler refuses to accept incomplete reasoning. In abstract discussion, you might focus on the \u0026ldquo;normal\u0026rdquo; states and unconsciously ignore edge cases. The compiler forces completeness.\u003c/p\u003e\n\u003cp\u003eOr consider cyclomatic complexity analysis built into Visual Studio and available through analyzers. High complexity scores (typically above 10) indicate control flow that\u0026rsquo;s difficult to reason about and test thoroughly. The code analyzer doesn\u0026rsquo;t just flag style violations—it measures cognitive load and highlights where thinking has likely become too tangled to maintain reliably.\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-csharp\" data-lang=\"csharp\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"c1\"\u003e// Complexity: 15 (Warning CS1591)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"kt\"\u003edecimal\u003c/span\u003e \u003cspan class=\"n\"\u003eCalculateDiscount\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eOrder\u003c/span\u003e \u003cspan class=\"n\"\u003eorder\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003eCustomer\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003eDateTime\u003c/span\u003e \u003cspan class=\"n\"\u003eorderDate\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"k\"\u003eif\u003c/span\u003e \u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eIsPremium\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"k\"\u003eif\u003c/span\u003e \u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eorder\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eTotal\u003c/span\u003e \u003cspan class=\"p\"\u003e\u0026gt;\u003c/span\u003e \u003cspan class=\"m\"\u003e1000\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"k\"\u003eif\u003c/span\u003e \u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eorderDate\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eMonth\u003c/span\u003e \u003cspan class=\"p\"\u003e==\u003c/span\u003e \u003cspan class=\"m\"\u003e12\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e                \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"m\"\u003e0.25\u003c/span\u003e\u003cspan class=\"n\"\u003em\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"k\"\u003eelse\u003c/span\u003e \u003cspan class=\"k\"\u003eif\u003c/span\u003e \u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eYearsActive\u003c/span\u003e \u003cspan class=\"p\"\u003e\u0026gt;\u003c/span\u003e \u003cspan class=\"m\"\u003e5\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e                \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"m\"\u003e0.20\u003c/span\u003e\u003cspan class=\"n\"\u003em\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"k\"\u003eelse\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e                \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"m\"\u003e0.15\u003c/span\u003e\u003cspan class=\"n\"\u003em\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"k\"\u003eelse\u003c/span\u003e \u003cspan class=\"k\"\u003eif\u003c/span\u003e \u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eorder\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eTotal\u003c/span\u003e \u003cspan class=\"p\"\u003e\u0026gt;\u003c/span\u003e \u003cspan class=\"m\"\u003e500\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"m\"\u003e0.10\u003c/span\u003e\u003cspan class=\"n\"\u003em\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"k\"\u003eelse\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"m\"\u003e0.05\u003c/span\u003e\u003cspan class=\"n\"\u003em\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"k\"\u003eelse\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"k\"\u003eif\u003c/span\u003e \u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eorder\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eTotal\u003c/span\u003e \u003cspan class=\"p\"\u003e\u0026gt;\u003c/span\u003e \u003cspan class=\"m\"\u003e1000\u003c/span\u003e \u003cspan class=\"p\"\u003e\u0026amp;\u0026amp;\u003c/span\u003e \u003cspan class=\"n\"\u003eorderDate\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eMonth\u003c/span\u003e \u003cspan class=\"p\"\u003e==\u003c/span\u003e \u003cspan class=\"m\"\u003e12\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"m\"\u003e0.15\u003c/span\u003e\u003cspan class=\"n\"\u003em\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"k\"\u003eelse\u003c/span\u003e \u003cspan class=\"k\"\u003eif\u003c/span\u003e \u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eorder\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eTotal\u003c/span\u003e \u003cspan class=\"p\"\u003e\u0026gt;\u003c/span\u003e \u003cspan class=\"m\"\u003e500\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"m\"\u003e0.05\u003c/span\u003e\u003cspan class=\"n\"\u003em\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"m\"\u003e0\u003c/span\u003e\u003cspan class=\"n\"\u003em\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eThis method might make sense in abstract discussion: \u0026ldquo;We give discounts based on customer status, order size, and date.\u0026rdquo; But complexity analysis reveals what abstract thinking hides—the decision tree is convoluted, error-prone, and unmaintainable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLook closer at the logic:\u003c/strong\u003e Business rules state that long-term premium customers (6+ years) should get the highest discount (30%) for high-value orders—even better than the December holiday bonus. But a premium customer with 7 years active ordering €1,500 in December only gets 25%—the \u003ccode\u003eelse if (customer.YearsActive \u0026gt; 5)\u003c/code\u003e branch returning 0.20m is \u003cstrong\u003eunreachable\u003c/strong\u003e because the December check already returned. \u003cstrong\u003eThe nested if-structure makes the bug invisible in code review but obvious when a test fails:\u003c/strong\u003e\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-csharp\" data-lang=\"csharp\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"na\"\u003e[Fact]\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"k\"\u003evoid\u003c/span\u003e \u003cspan class=\"n\"\u003eCalculateDiscount_LoyalPremiumCustomer_December_ShouldGetLoyaltyBonus\u003c/span\u003e\u003cspan class=\"p\"\u003e()\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"c1\"\u003e// Long-term customers should get loyalty discount even in December\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kt\"\u003evar\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"k\"\u003enew\u003c/span\u003e \u003cspan class=\"n\"\u003eCustomer\u003c/span\u003e \u003cspan class=\"p\"\u003e{\u003c/span\u003e \u003cspan class=\"n\"\u003eIsPremium\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"kc\"\u003etrue\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003eYearsActive\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"m\"\u003e7\u003c/span\u003e \u003cspan class=\"p\"\u003e};\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kt\"\u003evar\u003c/span\u003e \u003cspan class=\"n\"\u003eorder\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"k\"\u003enew\u003c/span\u003e \u003cspan class=\"n\"\u003eOrder\u003c/span\u003e \u003cspan class=\"p\"\u003e{\u003c/span\u003e \u003cspan class=\"n\"\u003eTotal\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"m\"\u003e1500\u003c/span\u003e \u003cspan class=\"p\"\u003e};\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kt\"\u003evar\u003c/span\u003e \u003cspan class=\"n\"\u003edate\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"k\"\u003enew\u003c/span\u003e \u003cspan class=\"n\"\u003eDateTime\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"m\"\u003e2025\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"m\"\u003e12\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"m\"\u003e15\u003c/span\u003e\u003cspan class=\"p\"\u003e);\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kt\"\u003evar\u003c/span\u003e \u003cspan class=\"n\"\u003ediscount\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"n\"\u003e_calculator\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eCalculateDiscount\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eorder\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003edate\u003c/span\u003e\u003cspan class=\"p\"\u003e);\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"n\"\u003eAssert\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eEqual\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"m\"\u003e0.30\u003c/span\u003e\u003cspan class=\"n\"\u003em\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003ediscount\u003c/span\u003e\u003cspan class=\"p\"\u003e);\u003c/span\u003e \u003cspan class=\"c1\"\u003e// FAILS: Returns 0.25m instead\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e                                    \u003cspan class=\"c1\"\u003e// The YearsActive\u0026gt;5 branch is unreachable!\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eThe code forces you to confront what clean thinking would have structured differently:\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-csharp\" data-lang=\"csharp\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"c1\"\u003e// Complexity: 4\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"kt\"\u003edecimal\u003c/span\u003e \u003cspan class=\"n\"\u003eCalculateDiscount\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eOrder\u003c/span\u003e \u003cspan class=\"n\"\u003eorder\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003eCustomer\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003eDateTime\u003c/span\u003e \u003cspan class=\"n\"\u003eorderDate\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kt\"\u003evar\u003c/span\u003e \u003cspan class=\"n\"\u003erules\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"k\"\u003enew\u003c/span\u003e \u003cspan class=\"n\"\u003eDiscountRuleEngine\u003c/span\u003e\u003cspan class=\"p\"\u003e()\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eAddRule\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"k\"\u003enew\u003c/span\u003e \u003cspan class=\"n\"\u003ePremiumCustomerRule\u003c/span\u003e\u003cspan class=\"p\"\u003e())\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eAddRule\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"k\"\u003enew\u003c/span\u003e \u003cspan class=\"n\"\u003eHighValueOrderRule\u003c/span\u003e\u003cspan class=\"p\"\u003e())\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eAddRule\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"k\"\u003enew\u003c/span\u003e \u003cspan class=\"n\"\u003eHolidayPromotionRule\u003c/span\u003e\u003cspan class=\"p\"\u003e());\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"n\"\u003erules\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eCalculateDiscount\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eorder\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003eorderDate\u003c/span\u003e\u003cspan class=\"p\"\u003e);\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eRefactoring didn\u0026rsquo;t just clean up syntax—it exposed and resolved structural thinking problems that abstract reasoning missed. The rule engine evaluates all rules and picks the highest discount, making the business logic explicit and the bug impossible.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"performance-where-theory-meets-production-reality\"\u003e\u003ca href=\"/posts/feedback-loop-ai-cant-replace/#performance-where-theory-meets-production-reality\" title=\"Performance: Where Theory Meets Production Reality\"\u003ePerformance: Where Theory Meets Production Reality\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eAlgorithmic complexity feels manageable in theoretical discussion. O(n) sounds reasonable. O(n²) seems acceptable for small datasets. O(n log n) feels efficient. Then production traffic hits, datasets grow larger than anticipated, and theoretical complexity translates into CPU cost, memory pressure, and timeout failures.\u003c/p\u003e\n\u003cp\u003eI\u0026rsquo;ve debugged production incidents where perfectly logical code—code that passed all functional tests—caused cascading performance failures. \u003cstrong\u003eHours wasted.\u003c/strong\u003e Customer complaints. Emergency hotfixes.\u003c/p\u003e\n\u003cp\u003eWhy? Complexity analysis happened in abstract terms rather than executable measurement.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExample:\u003c/strong\u003e Nested LINQ queries that looked clean and expressive during development:\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-csharp\" data-lang=\"csharp\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"c1\"\u003e// Looks elegant, reads well\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"n\"\u003eIEnumerable\u003c/span\u003e\u003cspan class=\"p\"\u003e\u0026lt;\u003c/span\u003e\u003cspan class=\"n\"\u003eOrderSummary\u003c/span\u003e\u003cspan class=\"p\"\u003e\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003eGetCustomerOrders\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"kt\"\u003eint\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomerId\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"n\"\u003e_orders\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eWhere\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eo\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003eo\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eCustomerId\u003c/span\u003e \u003cspan class=\"p\"\u003e==\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomerId\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eSelect\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eo\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"k\"\u003enew\u003c/span\u003e \u003cspan class=\"n\"\u003eOrderSummary\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"n\"\u003eOrderId\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"n\"\u003eo\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eId\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"n\"\u003eTotal\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"n\"\u003eo\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eLineItems\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eSum\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eli\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003eli\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003ePrice\u003c/span\u003e \u003cspan class=\"p\"\u003e*\u003c/span\u003e \u003cspan class=\"n\"\u003eli\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eQuantity\u003c/span\u003e\u003cspan class=\"p\"\u003e),\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"n\"\u003eItemCount\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"n\"\u003eo\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eLineItems\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eCount\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"n\"\u003eCategories\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"n\"\u003eo\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eLineItems\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e                \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eSelect\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eli\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003eli\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eProduct\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eCategory\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e                \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eDistinct\u003c/span\u003e\u003cspan class=\"p\"\u003e()\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e                \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eOrderBy\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003ec\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003ec\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eName\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e                \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eToList\u003c/span\u003e\u003cspan class=\"p\"\u003e()\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e})\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eOrderByDescending\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003es\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003es\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eTotal\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eToList\u003c/span\u003e\u003cspan class=\"p\"\u003e();\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eThis code communicates intent clearly. In abstract reasoning, it feels straightforward: \u0026ldquo;Get orders, calculate summaries, sort by total.\u0026rdquo; But execute it with real data and watch database query patterns, memory allocations, and execution time:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eMultiple database round trips per order (N+1 query problem)\u003c/li\u003e\n\u003cli\u003eRepeated calculations over the same collections\u003c/li\u003e\n\u003cli\u003eUnnecessary allocations for intermediate collections\u003c/li\u003e\n\u003cli\u003eLinear scans for categories on every line item\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThe abstract reasoning missed what executable profiling makes obvious:\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-csharp\" data-lang=\"csharp\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"c1\"\u003e// Same intent, different execution characteristics\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"kd\"\u003easync\u003c/span\u003e \u003cspan class=\"n\"\u003eTask\u003c/span\u003e\u003cspan class=\"p\"\u003e\u0026lt;\u003c/span\u003e\u003cspan class=\"n\"\u003eIEnumerable\u003c/span\u003e\u003cspan class=\"p\"\u003e\u0026lt;\u003c/span\u003e\u003cspan class=\"n\"\u003eOrderSummary\u003c/span\u003e\u003cspan class=\"p\"\u003e\u0026gt;\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003eGetCustomerOrders\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"kt\"\u003eint\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomerId\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kt\"\u003evar\u003c/span\u003e \u003cspan class=\"n\"\u003eorders\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"k\"\u003eawait\u003c/span\u003e \u003cspan class=\"n\"\u003e_context\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eOrders\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eWhere\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eo\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003eo\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eCustomerId\u003c/span\u003e \u003cspan class=\"p\"\u003e==\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomerId\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eInclude\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eo\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003eo\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eLineItems\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eThenInclude\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eli\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003eli\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eProduct\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eThenInclude\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003ep\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003ep\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eCategory\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eAsNoTracking\u003c/span\u003e\u003cspan class=\"p\"\u003e()\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eToListAsync\u003c/span\u003e\u003cspan class=\"p\"\u003e();\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"n\"\u003eorders\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eSelect\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eo\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"k\"\u003enew\u003c/span\u003e \u003cspan class=\"n\"\u003eOrderSummary\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"n\"\u003eOrderId\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"n\"\u003eo\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eId\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"n\"\u003eTotal\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"n\"\u003eo\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eLineItems\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eSum\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eli\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003eli\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003ePrice\u003c/span\u003e \u003cspan class=\"p\"\u003e*\u003c/span\u003e \u003cspan class=\"n\"\u003eli\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eQuantity\u003c/span\u003e\u003cspan class=\"p\"\u003e),\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"n\"\u003eItemCount\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"n\"\u003eo\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eLineItems\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eCount\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"n\"\u003eCategories\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"n\"\u003eo\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eLineItems\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e                \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eSelect\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eli\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003eli\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eProduct\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eCategory\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eName\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e                \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eDistinct\u003c/span\u003e\u003cspan class=\"p\"\u003e()\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e                \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eOrder\u003c/span\u003e\u003cspan class=\"p\"\u003e()\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e                \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eToList\u003c/span\u003e\u003cspan class=\"p\"\u003e()\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e})\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eOrderByDescending\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003es\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003es\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eTotal\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eToList\u003c/span\u003e\u003cspan class=\"p\"\u003e();\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eCode forced the performance implications into measurable form. Profiling revealed what abstract thought deferred—database round trips, allocation patterns, execution cost. Without writing and measuring executable code, these consequences remain invisible.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"the-feedback-loop-programming-provides\"\u003e\u003ca href=\"/posts/feedback-loop-ai-cant-replace/#the-feedback-loop-programming-provides\" title=\"The Feedback Loop Programming Provides\"\u003eThe Feedback Loop Programming Provides\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eProgramming isn\u0026rsquo;t just thinking\u0026rsquo;s output—it\u0026rsquo;s thinking\u0026rsquo;s verification mechanism. The discipline of translating thought into executable form exposes inconsistencies, reveals missing decisions, and surfaces consequences that abstract reasoning defers.\u003c/p\u003e\n\u003cp\u003eThis feedback loop operates at multiple levels:\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"compilation-immediate-logical-feedback\"\u003e\u003ca href=\"/posts/feedback-loop-ai-cant-replace/#compilation-immediate-logical-feedback\" title=\"Compilation: Immediate Logical Feedback\"\u003eCompilation: Immediate Logical Feedback\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eThe compiler catches type mismatches, null reference possibilities, exhaustiveness gaps, and logical inconsistencies within seconds. No mental review provides this consistency and speed.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"testing-behavioral-verification\"\u003e\u003ca href=\"/posts/feedback-loop-ai-cant-replace/#testing-behavioral-verification\" title=\"Testing: Behavioral Verification\"\u003eTesting: Behavioral Verification\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eUnit tests, integration tests, and property-based tests validate that your mental model of system behavior matches actual execution. I\u0026rsquo;ve written tests expecting specific behavior only to discover the code does something entirely different—not because implementation was wrong, but because reasoning was incomplete.\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-csharp\" data-lang=\"csharp\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"na\"\u003e[Fact]\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"k\"\u003evoid\u003c/span\u003e \u003cspan class=\"n\"\u003eCalculateDiscount_PremiumCustomer_HighValue_December_Returns25Percent\u003c/span\u003e\u003cspan class=\"p\"\u003e()\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"c1\"\u003e// Test reveals the logic we thought we implemented doesn\u0026#39;t match what we actually coded\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kt\"\u003evar\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"k\"\u003enew\u003c/span\u003e \u003cspan class=\"n\"\u003eCustomer\u003c/span\u003e \u003cspan class=\"p\"\u003e{\u003c/span\u003e \u003cspan class=\"n\"\u003eIsPremium\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"kc\"\u003etrue\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003eYearsActive\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"m\"\u003e3\u003c/span\u003e \u003cspan class=\"p\"\u003e};\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kt\"\u003evar\u003c/span\u003e \u003cspan class=\"n\"\u003eorder\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"k\"\u003enew\u003c/span\u003e \u003cspan class=\"n\"\u003eOrder\u003c/span\u003e \u003cspan class=\"p\"\u003e{\u003c/span\u003e \u003cspan class=\"n\"\u003eTotal\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"m\"\u003e1500\u003c/span\u003e \u003cspan class=\"p\"\u003e};\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kt\"\u003evar\u003c/span\u003e \u003cspan class=\"n\"\u003edate\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"k\"\u003enew\u003c/span\u003e \u003cspan class=\"n\"\u003eDateTime\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"m\"\u003e2025\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"m\"\u003e12\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"m\"\u003e15\u003c/span\u003e\u003cspan class=\"p\"\u003e);\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kt\"\u003evar\u003c/span\u003e \u003cspan class=\"n\"\u003ediscount\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"n\"\u003e_calculator\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eCalculateDiscount\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eorder\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003edate\u003c/span\u003e\u003cspan class=\"p\"\u003e);\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"n\"\u003eAssert\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eEqual\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"m\"\u003e0.25\u003c/span\u003e\u003cspan class=\"n\"\u003em\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003ediscount\u003c/span\u003e\u003cspan class=\"p\"\u003e);\u003c/span\u003e \u003cspan class=\"c1\"\u003e// Fails: Returns 0.15m instead\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eThe test didn\u0026rsquo;t catch a bug in isolation—it caught incomplete thinking that manifested as unexpected behavior. Without executable code and explicit testing, that gap stays hidden until production.\u003c/p\u003e\n\n\n\n\n\u003ch4 id=\"the-ai-testing-trap\"\u003e\u003ca href=\"/posts/feedback-loop-ai-cant-replace/#the-ai-testing-trap\" title=\"The AI Testing Trap\"\u003eThe AI Testing Trap\u003c/a\u003e\u003c/h4\u003e\n\u003cp\u003eAI can generate tests as easily as it generates implementations. Ask for unit tests, and you\u0026rsquo;ll get methods that exercise code paths and verify outputs. \u003cstrong\u003eThis creates a dangerous illusion: high code coverage with low confidence.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAI-generated tests typically verify \u003cstrong\u003ehappy paths\u003c/strong\u003e—the scenarios explicitly described in prompts. They \u003cstrong\u003erarely\u003c/strong\u003e test:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eEdge cases that emerge from domain understanding\u003c/li\u003e\n\u003cli\u003eConcurrency issues that only appear under load\u003c/li\u003e\n\u003cli\u003eError propagation through system boundaries\u003c/li\u003e\n\u003cli\u003eIntegration failures when dependencies behave unexpectedly\u003c/li\u003e\n\u003cli\u003ePerformance degradation with realistic data volumes\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003eI\u0026rsquo;ve reviewed codebases with 90%+ test coverage where AI generated both implementation and tests.\u003c/strong\u003e Every test passed. Yet production revealed critical bugs because the tests verified that the code did \u003cstrong\u003ewhat it was written to do\u003c/strong\u003e, not that it \u003cstrong\u003esolved the actual problem correctly\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThe professional\u0026rsquo;s advantage:\u003c/strong\u003e knowing what to test comes from understanding how systems fail in production. That knowledge \u003cstrong\u003ecan\u0026rsquo;t be prompted\u003c/strong\u003e—it must be experienced, internalized, and applied deliberately.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"profiling-performance-reality-check\"\u003e\u003ca href=\"/posts/feedback-loop-ai-cant-replace/#profiling-performance-reality-check\" title=\"Profiling: Performance Reality Check\"\u003eProfiling: Performance Reality Check\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eProfilers measure actual CPU consumption, memory allocation patterns, I/O bottlenecks, and threading contention. Abstract complexity analysis (Big-O notation) provides theoretical bounds. Profiling provides operational reality.\u003c/p\u003e\n\u003cp\u003eVisual Studio\u0026rsquo;s .NET Object Allocation tool shows exactly which code paths allocate memory and how much. BenchmarkDotNet provides precise execution timing with statistical analysis. These tools don\u0026rsquo;t just measure code—they validate or invalidate reasoning about performance characteristics.\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-csharp\" data-lang=\"csharp\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"na\"\u003e[MemoryDiagnoser]\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"k\"\u003eclass\u003c/span\u003e \u003cspan class=\"nc\"\u003eStringBuildingBenchmark\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"na\"\u003e    [Benchmark]\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"kt\"\u003estring\u003c/span\u003e \u003cspan class=\"n\"\u003eConcatenationInLoop\u003c/span\u003e\u003cspan class=\"p\"\u003e()\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"kt\"\u003estring\u003c/span\u003e \u003cspan class=\"n\"\u003eresult\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"s\"\u003e\u0026#34;\u0026#34;\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"k\"\u003efor\u003c/span\u003e \u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"kt\"\u003eint\u003c/span\u003e \u003cspan class=\"n\"\u003ei\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"m\"\u003e0\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e \u003cspan class=\"n\"\u003ei\u003c/span\u003e \u003cspan class=\"p\"\u003e\u0026lt;\u003c/span\u003e \u003cspan class=\"m\"\u003e1000\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e \u003cspan class=\"n\"\u003ei\u003c/span\u003e\u003cspan class=\"p\"\u003e++)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"n\"\u003eresult\u003c/span\u003e \u003cspan class=\"p\"\u003e+=\u003c/span\u003e \u003cspan class=\"n\"\u003ei\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eToString\u003c/span\u003e\u003cspan class=\"p\"\u003e();\u003c/span\u003e \u003cspan class=\"c1\"\u003e// Abstract: \u0026#34;Should be fine for 1000 iterations\u0026#34;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"n\"\u003eresult\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"na\"\u003e    \n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"na\"\u003e    [Benchmark]\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"kt\"\u003estring\u003c/span\u003e \u003cspan class=\"n\"\u003eStringBuilderInLoop\u003c/span\u003e\u003cspan class=\"p\"\u003e()\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"kt\"\u003evar\u003c/span\u003e \u003cspan class=\"n\"\u003ebuilder\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"k\"\u003enew\u003c/span\u003e \u003cspan class=\"n\"\u003eStringBuilder\u003c/span\u003e\u003cspan class=\"p\"\u003e();\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"k\"\u003efor\u003c/span\u003e \u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"kt\"\u003eint\u003c/span\u003e \u003cspan class=\"n\"\u003ei\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"m\"\u003e0\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e \u003cspan class=\"n\"\u003ei\u003c/span\u003e \u003cspan class=\"p\"\u003e\u0026lt;\u003c/span\u003e \u003cspan class=\"m\"\u003e1000\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e \u003cspan class=\"n\"\u003ei\u003c/span\u003e\u003cspan class=\"p\"\u003e++)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"n\"\u003ebuilder\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eAppend\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003ei\u003c/span\u003e\u003cspan class=\"p\"\u003e);\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"n\"\u003ebuilder\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eToString\u003c/span\u003e\u003cspan class=\"p\"\u003e();\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"c1\"\u003e// Results expose reality abstract thinking missed:\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"c1\"\u003e// ConcatenationInLoop:  3,450 μs,  allocated: 2,031,616 B\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"c1\"\u003e// StringBuilderInLoop:     45 μs,  allocated:     24,624 B\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eThe difference between \u0026ldquo;seems reasonable\u0026rdquo; and \u0026ldquo;actually performs\u0026rdquo; is 75x execution time and 80x memory allocation. Abstract reasoning deferred these consequences. Executable code and measurement made them visible.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"production-ultimate-reality-validation\"\u003e\u003ca href=\"/posts/feedback-loop-ai-cant-replace/#production-ultimate-reality-validation\" title=\"Production: Ultimate Reality Validation\"\u003eProduction: Ultimate Reality Validation\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eProduction exposes every assumption abstract thinking made: scale, concurrency, failure modes, dependency availability, network latency, operational complexity. Code that worked flawlessly in development reveals hidden assumptions when deployed at scale with real users, real data, and real failure conditions.\u003c/p\u003e\n\u003cp\u003eMonitoring, telemetry, and distributed tracing provide feedback about system behavior under actual conditions. Without executable code running in production, all architectural reasoning remains theoretical.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"programming-and-thinking-inseparable-not-sequential\"\u003e\u003ca href=\"/posts/feedback-loop-ai-cant-replace/#programming-and-thinking-inseparable-not-sequential\" title=\"Programming and Thinking: Inseparable, Not Sequential\"\u003eProgramming and Thinking: Inseparable, Not Sequential\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eThe original framing positioned thinking and programming sequentially: think first (the hard part), then program (the easy translation). This model fundamentally misrepresents the relationship.\u003c/p\u003e\n\u003cp\u003eProgramming and thinking are inseparable, iterative, and mutually reinforcing:\u003c/p\u003e\n\u003col\u003e\n\u003cli\u003e\u003cstrong\u003eAbstract thinking\u003c/strong\u003e identifies problems, explores solution spaces, and proposes approaches.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eCode writing\u003c/strong\u003e forces abstraction into precise, executable form, exposing gaps and inconsistencies.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eExecution and measurement\u003c/strong\u003e reveal consequences—performance, resource consumption, failure modes—that abstract thought deferred.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eRefinement\u003c/strong\u003e incorporates execution reality back into thinking, improving the mental model.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eRepeat\u003c/strong\u003e until thinking and execution align.\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eNeither operates effectively alone. Thinking without code stays vague and unvalidated. Code without thinking becomes mechanical translation without understanding. High-quality software emerges from tight iteration between abstract reasoning and executable verification.\u003c/p\u003e\n\u003cp\u003eThis isn\u0026rsquo;t pedantry about implementation details. This is recognition that software engineering is fundamentally about managing complexity in executable systems. Complexity that can\u0026rsquo;t be reasoned about produces brittle, unmaintainable systems. Complexity that remains purely abstract never confronts operational reality.\u003c/p\u003e\n\u003cp\u003eThe discipline of programming (writing code, measuring behavior, refactoring based on feedback) is how abstract thinking becomes operational reality. It\u0026rsquo;s not the easy part that follows hard thinking. It\u0026rsquo;s the verification mechanism that sharpens thinking and exposes where reasoning was incomplete.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"what-makes-professionals-irreplaceable\"\u003e\u003ca href=\"/posts/feedback-loop-ai-cant-replace/#what-makes-professionals-irreplaceable\" title=\"What Makes Professionals Irreplaceable\"\u003eWhat Makes Professionals Irreplaceable\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eCompilers validate logic. Tests reveal behavioral gaps. Profilers measure performance reality. Production exposes every deferred decision. These tools generate feedback constantly.\u003c/p\u003e\n\u003cp\u003eBut feedback is worthless without interpretation. And interpretation requires experience.\u003c/p\u003e\n\u003cp\u003eWhen a profiler shows 75x performance degradation, the junior developer sees a red flag. The senior engineer sees a memory allocation pattern they\u0026rsquo;ve debugged before, recognizes the architectural constraint it reveals, and knows three ways to fix it based on context. When production monitoring shows intermittent timeout spikes, AI suggests retry logic. The experienced architect recognizes a connection pool exhaustion pattern and addresses the root cause.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThe irreplaceable skill isn\u0026rsquo;t generating code. It\u0026rsquo;s closing the feedback loop.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThat means watching code fail, understanding \u003cem\u003ewhy\u003c/em\u003e it fails, and refining your mental model until your intuition predicts failure modes before they manifest. AI participates in generating code and even in analyzing errors. But it can\u0026rsquo;t internalize the lessons. It can\u0026rsquo;t build the judgment that comes from years of production incidents, debugging sessions, and architectural decisions that played out over time.\u003c/p\u003e\n\u003cp\u003eIn the \u003ca href=\"../real-professional-software-engineering-ai-era/\"\u003efinal part of this series\u003c/a\u003e, we\u0026rsquo;ll examine what this means for professional development. When code generation is commoditized, what skills actually matter? How do you build the cognitive architecture that AI can\u0026rsquo;t replicate?\u003c/p\u003e\n\u003cp\u003eThe answer shapes how we train developers, evaluate expertise, and define what \u0026ldquo;senior engineer\u0026rdquo; means in an AI-augmented world.\u003c/p\u003e\n","date_modified":"2026-05-26T10:22:03+02:00","date_published":"2026-01-15T17:00:00+01:00","id":"https://daily-devops.net/posts/feedback-loop-ai-cant-replace/","language":"en","summary":"Compilers validate logic, profilers expose performance lies, and production reveals every deferred decision. AI cannot close that feedback loop for you.\n","tags":["softwareengineering","codequality","bestpractices","architecture","dotnet","csharp","technicaldebt","ai-code-assistant","github-copilot"],"title":"The Feedback Loop That AI Can't Replace\n","url":"https://daily-devops.net/posts/feedback-loop-ai-cant-replace/"},{"authors":[{"name":"Martin Stühmer","url":"https://daily-devops.net/authors/martin/"}],"content_html":"\u003cp\u003eA \u003ca href=\"https://www.linkedin.com/posts/davideguida_i-feel-its-time-to-make-something-clear-activity-7411391768283271168-naE4\" target=\"_blank\" rel=\"noopener external noreferrer\"\u003eLinkedIn post by David Guida\u003c/a\u003e sparked a discussion that cuts to the bone: \u003cstrong\u003eIs software engineering about thinking or typing?\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDavid argued forcefully that \u0026ldquo;software engineering is NOT about writing code\u0026rdquo;—that code is merely mechanical output, the easy part, just another language. The hard part, he wrote, is thinking: \u0026ldquo;Programming is a byproduct of the thinking process. And that one, my friends, is the hard part.\u0026rdquo;\u003c/p\u003e\n\u003cp\u003eI responded with a point that needed more space than a LinkedIn comment allows:\u003c/p\u003e\n\u003cblockquote\u003e\n\u003cp\u003e\u0026ldquo;Strong point, but it slightly overcorrects. Yes, typing code is the easy, mechanical part. The hard part is reasoning, trade-offs, and understanding constraints. Agreed. \u003cstrong\u003eBut dismissing code as \u0026lsquo;just another language\u0026rsquo; undersells its impact.\u003c/strong\u003e Code is not only expression, it is execution, cost, failure modes, and long-term operational risk. Thinking without being forced into precise, executable form often stays vague. Writing code is where weak thinking gets exposed. Programming is a byproduct of thinking, \u003cstrong\u003ebut it is also the feedback loop that sharpens that thinking.\u003c/strong\u003e One without the other does not scale.\u0026rdquo;\u003c/p\u003e\n\u003c/blockquote\u003e\n\u003cp\u003eDavid\u0026rsquo;s response captured what I\u0026rsquo;m exploring here: \u0026ldquo;I totally agree! I must have oversimplified my thoughts. Your closing note on the feedback loop \u003cstrong\u003ecaptures the reason why real professionals will never be replaced.\u003c/strong\u003e\u0026rdquo;\u003c/p\u003e\n\u003cp\u003eThat last phrase wasn\u0026rsquo;t casual. It addresses the elephant everyone sees but few acknowledge: \u003cstrong\u003eAI code assistants everywhere.\u003c/strong\u003e GitHub Copilot. ChatGPT generating entire applications from prompts. The emerging \u0026ldquo;vibe coding\u0026rdquo; trend where developers describe vibes and let AI handle the dirty work.\u003c/p\u003e\n\u003cp\u003eThe timing matters. We\u0026rsquo;re in an era where typing code has \u003cstrong\u003enever been easier\u003c/strong\u003e. AI generates syntactically correct implementations \u003cstrong\u003efaster than any human can type\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eYet here\u0026rsquo;s what David and I both realized: This makes the feedback loop between thinking and code \u003cstrong\u003emore critical, not less\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAsk yourself:\u003c/strong\u003e When code generation becomes trivial, what separates you from a prompt engineer who thinks they\u0026rsquo;re building software?\u003c/p\u003e\n\u003cp\u003eThe answer: Understanding what that code actually \u003cstrong\u003edoes\u003c/strong\u003e. What it \u003cstrong\u003ecosts\u003c/strong\u003e. Where it \u003cstrong\u003efails\u003c/strong\u003e. Why it \u003cstrong\u003ebreaks under load\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eThis article expands on that feedback loop—the relationship between thinking and code that AI can\u0026rsquo;t replicate. It explores why AI-generated code without deep understanding creates an \u003cstrong\u003eillusion of productivity that collapses catastrophically under production load\u003c/strong\u003e.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"where-vague-thinking-hides\"\u003e\u003ca href=\"/posts/code-sharpens-thinking/#where-vague-thinking-hides\" title=\"Where Vague Thinking Hides\"\u003eWhere Vague Thinking Hides\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eWalk through any architecture review where diagrams look perfect, responsibilities seem clear, and everyone nods in agreement. Then watch what happens when someone starts writing the actual implementation. Suddenly, the clean boundaries blur. The \u0026ldquo;simple\u0026rdquo; abstraction requires five parameters. The proposed interface doesn\u0026rsquo;t fit half the use cases. The design that felt obvious in discussion becomes ambiguous when translated to executable code.\u003c/p\u003e\n\u003cp\u003eThis isn\u0026rsquo;t implementation failing design. This is design revealing itself to be incomplete.\u003c/p\u003e\n\u003cp\u003eI\u0026rsquo;ve sat through countless discussions where proposed solutions felt reasonable until we asked: \u0026ldquo;Show me the code.\u0026rdquo; Not production code—just a sketch. Suddenly, implicit assumptions surface. Missing responsibilities become visible. Performance implications emerge. The architecture that seemed solid in abstract terms crumbles when forced into compilable form.\u003c/p\u003e\n\u003cp\u003eCode demands precision that thought alone doesn\u0026rsquo;t require. When you think through a problem, your mind fills gaps unconsciously, papers over inconsistencies, and substitutes intuition for rigor. When you write code, the compiler—and eventually production—refuses to cooperate with vague intent.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"how-the-compiler-exposes-vague-thinking\"\u003e\u003ca href=\"/posts/code-sharpens-thinking/#how-the-compiler-exposes-vague-thinking\" title=\"How The Compiler Exposes Vague Thinking\"\u003eHow The Compiler Exposes Vague Thinking\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eConsider nullable reference types in C#. Without explicit declaration, you can mentally handwave nullability concerns:\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-csharp\" data-lang=\"csharp\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"c1\"\u003e// Vague thinking: \u0026#34;customer will always have a name\u0026#34;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"k\"\u003eclass\u003c/span\u003e \u003cspan class=\"nc\"\u003eCustomer\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"kt\"\u003estring\u003c/span\u003e \u003cspan class=\"n\"\u003eName\u003c/span\u003e \u003cspan class=\"p\"\u003e{\u003c/span\u003e \u003cspan class=\"k\"\u003eget\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e \u003cspan class=\"k\"\u003eset\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e \u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eEnable nullable reference types, and the compiler forces you to confront reality:\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-csharp\" data-lang=\"csharp\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"cp\"\u003e#nullable\u003c/span\u003e \u003cspan class=\"n\"\u003eenable\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"k\"\u003eclass\u003c/span\u003e \u003cspan class=\"nc\"\u003eCustomer\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"kt\"\u003estring\u003c/span\u003e \u003cspan class=\"n\"\u003eName\u003c/span\u003e \u003cspan class=\"p\"\u003e{\u003c/span\u003e \u003cspan class=\"k\"\u003eget\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e \u003cspan class=\"k\"\u003eset\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e \u003cspan class=\"p\"\u003e}\u003c/span\u003e \u003cspan class=\"c1\"\u003e// Warning CS8618: Non-nullable property \u0026#39;Name\u0026#39; must contain \u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e                                      \u003cspan class=\"c1\"\u003e// a non-null value when exiting constructor\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eThis isn\u0026rsquo;t pedantry. This is thinking being forced into honest, executable form. Either you guarantee initialization, accept nullability explicitly, or redesign the constructor contract:\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-csharp\" data-lang=\"csharp\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"k\"\u003eclass\u003c/span\u003e \u003cspan class=\"nc\"\u003eCustomer\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"n\"\u003eCustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"kt\"\u003estring\u003c/span\u003e \u003cspan class=\"n\"\u003ename\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"n\"\u003eName\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"n\"\u003ename\u003c/span\u003e \u003cspan class=\"p\"\u003e??\u003c/span\u003e \u003cspan class=\"k\"\u003ethrow\u003c/span\u003e \u003cspan class=\"k\"\u003enew\u003c/span\u003e \u003cspan class=\"n\"\u003eArgumentNullException\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003enameof\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003ename\u003c/span\u003e\u003cspan class=\"p\"\u003e));\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"kt\"\u003estring\u003c/span\u003e \u003cspan class=\"n\"\u003eName\u003c/span\u003e \u003cspan class=\"p\"\u003e{\u003c/span\u003e \u003cspan class=\"k\"\u003eget\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e \u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eThe act of writing code exposed a decision that pure thought glossed over. Was \u003ccode\u003eName\u003c/code\u003e required or optional? The compiler didn\u0026rsquo;t care about your mental model—it demanded an explicit answer.\u003c/p\u003e\n\u003cp\u003eThis happens at every level: API contracts, concurrency assumptions, resource ownership, error propagation. Abstract thinking lets you defer these decisions indefinitely. Code forces resolution.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"the-vibe-coding-illusion-when-ai-generates-code-without-understanding\"\u003e\u003ca href=\"/posts/code-sharpens-thinking/#the-vibe-coding-illusion-when-ai-generates-code-without-understanding\" title=\"The Vibe Coding Illusion: When AI Generates Code Without Understanding\"\u003eThe Vibe Coding Illusion: When AI Generates Code Without Understanding\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eAI code assistants accelerate the mechanical part—the typing—\u003cstrong\u003eextraordinarily well\u003c/strong\u003e. Describe a function in natural language, and GitHub Copilot suggests an implementation within seconds. Ask ChatGPT to build a REST API, and it generates hundreds of lines of code that compile and often run.\u003c/p\u003e\n\u003cp\u003eThis feels like \u003cstrong\u003emagic\u003c/strong\u003e until you ask the critical question: \u003cem\u003edoes the generated code do what you actually need, not what you described?\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eI\u0026rsquo;ve reviewed pull requests where developers used AI to generate complete features. The code compiled. Tests passed. The PR description matched the implementation. Everything looked \u003cstrong\u003efine\u003c/strong\u003e. Until production deployment revealed that the AI had:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eGenerated \u003cstrong\u003ethread-unsafe code\u003c/strong\u003e for concurrent scenarios the prompt didn\u0026rsquo;t mention\u003c/li\u003e\n\u003cli\u003eAllocated memory in hot paths \u003cstrong\u003ewithout consideration\u003c/strong\u003e for garbage collection pressure\u003c/li\u003e\n\u003cli\u003eImplemented \u003cstrong\u003eO(n²) algorithms\u003c/strong\u003e where O(n) solutions existed\u003c/li\u003e\n\u003cli\u003eCreated database queries that worked with test data but \u003cstrong\u003efailed catastrophically\u003c/strong\u003e with production scale\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eIgnored error handling\u003c/strong\u003e edge cases that weren\u0026rsquo;t in the prompt\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003eThe AI didn\u0026rsquo;t fail—it did exactly what was asked.\u003c/strong\u003e The developer failed by not understanding that code generated from a prompt is a starting point, \u003cstrong\u003enot a solution\u003c/strong\u003e. The feedback loop—write code, measure behavior, understand consequences, refine thinking—got \u003cstrong\u003eshort-circuited\u003c/strong\u003e.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"why-vibe-coding-collapses-under-load\"\u003e\u003ca href=\"/posts/code-sharpens-thinking/#why-vibe-coding-collapses-under-load\" title=\"Why Vibe Coding Collapses Under Load\"\u003eWhy Vibe Coding Collapses Under Load\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003e\u0026ldquo;\u003cstrong\u003eVibe coding\u003c/strong\u003e\u0026rdquo; is the term emerging for this pattern: describe the vibe of what you want, let AI generate implementation, ship it if it passes basic tests. It treats code as expression divorced from execution reality. It assumes that if code compiles and handles the happy path, \u003cstrong\u003eit\u0026rsquo;s correct\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThis assumption works until it doesn\u0026rsquo;t.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAnd when it doesn\u0026rsquo;t? You\u0026rsquo;re stuck. No foundation for debugging. Can\u0026rsquo;t reason about performance. Can\u0026rsquo;t identify where implementation diverges from requirements. Can\u0026rsquo;t refactor intelligently.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWhy?\u003c/strong\u003e Because you don\u0026rsquo;t understand what the code actually does.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHere\u0026rsquo;s your professional advantage:\u003c/strong\u003e You recognize what the compiler \u003cstrong\u003ecan\u0026rsquo;t\u003c/strong\u003e tell you:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eThat the generated code works for the described case but \u003cstrong\u003efails for the dozen edge cases\u003c/strong\u003e you didn\u0026rsquo;t think to mention\u003c/li\u003e\n\u003cli\u003eThat the algorithm performs acceptably with 100 records but \u003cstrong\u003ecollapses with 100,000\u003c/strong\u003e\u003c/li\u003e\n\u003cli\u003eThat the abstraction looks clean but \u003cstrong\u003ecreates maintenance nightmares\u003c/strong\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003eAI generates syntax. Professionals understand semantics, performance characteristics, failure modes, and operational implications.\u003c/strong\u003e The gap between these is where \u0026ldquo;real professionals will never be replaced.\u0026rdquo;\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"code-materializes-consequences\"\u003e\u003ca href=\"/posts/code-sharpens-thinking/#code-materializes-consequences\" title=\"Code Materializes Consequences\"\u003eCode Materializes Consequences\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eCode isn\u0026rsquo;t just structured thought—it\u0026rsquo;s thought with operational consequences. When you design an architecture, you\u0026rsquo;re reasoning about responsibilities and boundaries. When you implement it, you\u0026rsquo;re creating CPU consumption patterns, memory allocation profiles, I/O bottlenecks, and long-term maintenance burdens.\u003c/p\u003e\n\u003cp\u003eThese aren\u0026rsquo;t secondary concerns. They\u0026rsquo;re the actual impact of your decisions.\u003c/p\u003e\n\u003cp\u003eTake a straightforward example: caching. In discussion, caching sounds simple—\u0026ldquo;we\u0026rsquo;ll cache frequently accessed data.\u0026rdquo; The thinking feels complete. Then you implement it:\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-csharp\" data-lang=\"csharp\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"c1\"\u003e// Looks reasonable in isolation\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003eprivate\u003c/span\u003e \u003cspan class=\"k\"\u003ereadonly\u003c/span\u003e \u003cspan class=\"n\"\u003eDictionary\u003c/span\u003e\u003cspan class=\"p\"\u003e\u0026lt;\u003c/span\u003e\u003cspan class=\"kt\"\u003eint\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003eCustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003e_cache\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"k\"\u003enew\u003c/span\u003e\u003cspan class=\"p\"\u003e();\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"n\"\u003eCustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e?\u003c/span\u003e \u003cspan class=\"n\"\u003eGetCustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"kt\"\u003eint\u003c/span\u003e \u003cspan class=\"n\"\u003eid\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"k\"\u003eif\u003c/span\u003e \u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003e_cache\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eTryGetValue\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eid\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"k\"\u003eout\u003c/span\u003e \u003cspan class=\"kt\"\u003evar\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e))\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"n\"\u003e_repository\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eLoad\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eid\u003c/span\u003e\u003cspan class=\"p\"\u003e);\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"k\"\u003eif\u003c/span\u003e \u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e \u003cspan class=\"p\"\u003e!=\u003c/span\u003e \u003cspan class=\"kc\"\u003enull\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"n\"\u003e_cache\u003c/span\u003e\u003cspan class=\"p\"\u003e[\u003c/span\u003e\u003cspan class=\"n\"\u003eid\u003c/span\u003e\u003cspan class=\"p\"\u003e]\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eThis code compiles. It runs. It even passes basic functional tests.\u003c/p\u003e\n\u003cp\u003eThen production hits.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"what-abstract-thinking-defers\"\u003e\u003ca href=\"/posts/code-sharpens-thinking/#what-abstract-thinking-defers\" title=\"What Abstract Thinking Defers\"\u003eWhat Abstract Thinking Defers\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003eWhat abstract thinking missed:\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e\u003cstrong\u003eMemory\u003c/strong\u003e: Cache grows unbounded. No eviction policy. Memory consumption increases until the process crashes or triggers garbage collection storms that degrade response times by 300%.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eConcurrency\u003c/strong\u003e: No synchronization. Multiple threads corrupt dictionary state, causing crashes or silent data corruption that costs hours of incident response time.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eConsistency\u003c/strong\u003e: Cache never invalidates. Stale data persists indefinitely, creating subtle bugs that customer support escalates—costing reputation and revenue.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eObservability\u003c/strong\u003e: No metrics. You can\u0026rsquo;t tell if caching helps or hurts performance without instrumenting separately.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eEach of these issues represents thinking that felt complete in abstract terms but was fundamentally incomplete in executable reality. The \u0026ldquo;simple\u0026rdquo; caching decision materialized as:\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-csharp\" data-lang=\"csharp\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003eprivate\u003c/span\u003e \u003cspan class=\"k\"\u003ereadonly\u003c/span\u003e \u003cspan class=\"n\"\u003eConcurrentDictionary\u003c/span\u003e\u003cspan class=\"p\"\u003e\u0026lt;\u003c/span\u003e\u003cspan class=\"kt\"\u003eint\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003eCacheEntry\u003c/span\u003e\u003cspan class=\"p\"\u003e\u0026lt;\u003c/span\u003e\u003cspan class=\"n\"\u003eCustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e\u0026gt;\u0026gt;\u003c/span\u003e \u003cspan class=\"n\"\u003e_cache\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"k\"\u003enew\u003c/span\u003e\u003cspan class=\"p\"\u003e();\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003eprivate\u003c/span\u003e \u003cspan class=\"k\"\u003ereadonly\u003c/span\u003e \u003cspan class=\"n\"\u003eTimeSpan\u003c/span\u003e \u003cspan class=\"n\"\u003e_expirationWindow\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"n\"\u003eTimeSpan\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eFromMinutes\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"m\"\u003e5\u003c/span\u003e\u003cspan class=\"p\"\u003e);\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003eprivate\u003c/span\u003e \u003cspan class=\"k\"\u003ereadonly\u003c/span\u003e \u003cspan class=\"kt\"\u003eint\u003c/span\u003e \u003cspan class=\"n\"\u003e_maxCacheSize\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"m\"\u003e10000\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"kd\"\u003epublic\u003c/span\u003e \u003cspan class=\"n\"\u003eCustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e?\u003c/span\u003e \u003cspan class=\"n\"\u003eGetCustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"kt\"\u003eint\u003c/span\u003e \u003cspan class=\"n\"\u003eid\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"k\"\u003eif\u003c/span\u003e \u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003e_cache\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eTryGetValue\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eid\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"k\"\u003eout\u003c/span\u003e \u003cspan class=\"kt\"\u003evar\u003c/span\u003e \u003cspan class=\"n\"\u003eentry\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e \u003cspan class=\"p\"\u003e\u0026amp;\u0026amp;\u003c/span\u003e \u003cspan class=\"p\"\u003e!\u003c/span\u003e\u003cspan class=\"n\"\u003eentry\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eIsExpired\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003e_expirationWindow\u003c/span\u003e\u003cspan class=\"p\"\u003e))\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"n\"\u003eMetrics\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eCacheHitCounter\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eIncrement\u003c/span\u003e\u003cspan class=\"p\"\u003e();\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"n\"\u003eentry\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eValue\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"n\"\u003eMetrics\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eCacheMissCounter\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eIncrement\u003c/span\u003e\u003cspan class=\"p\"\u003e();\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"kt\"\u003evar\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"n\"\u003e_repository\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eLoad\u003c/span\u003e\u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003eid\u003c/span\u003e\u003cspan class=\"p\"\u003e);\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"k\"\u003eif\u003c/span\u003e \u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e \u003cspan class=\"p\"\u003e!=\u003c/span\u003e \u003cspan class=\"kc\"\u003enull\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e{\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"k\"\u003eif\u003c/span\u003e \u003cspan class=\"p\"\u003e(\u003c/span\u003e\u003cspan class=\"n\"\u003e_cache\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eCount\u003c/span\u003e \u003cspan class=\"p\"\u003e\u0026gt;=\u003c/span\u003e \u003cspan class=\"n\"\u003e_maxCacheSize\u003c/span\u003e\u003cspan class=\"p\"\u003e)\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e            \u003cspan class=\"n\"\u003eEvictOldestEntry\u003c/span\u003e\u003cspan class=\"p\"\u003e();\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e        \u003cspan class=\"n\"\u003e_cache\u003c/span\u003e\u003cspan class=\"p\"\u003e[\u003c/span\u003e\u003cspan class=\"n\"\u003eid\u003c/span\u003e\u003cspan class=\"p\"\u003e]\u003c/span\u003e \u003cspan class=\"p\"\u003e=\u003c/span\u003e \u003cspan class=\"k\"\u003enew\u003c/span\u003e \u003cspan class=\"n\"\u003eCacheEntry\u003c/span\u003e\u003cspan class=\"p\"\u003e\u0026lt;\u003c/span\u003e\u003cspan class=\"n\"\u003eCustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e\u0026gt;(\u003c/span\u003e\u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e,\u003c/span\u003e \u003cspan class=\"n\"\u003eDateTimeOffset\u003c/span\u003e\u003cspan class=\"p\"\u003e.\u003c/span\u003e\u003cspan class=\"n\"\u003eUtcNow\u003c/span\u003e\u003cspan class=\"p\"\u003e);\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e    \u003cspan class=\"k\"\u003ereturn\u003c/span\u003e \u003cspan class=\"n\"\u003ecustomer\u003c/span\u003e\u003cspan class=\"p\"\u003e;\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"p\"\u003e}\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eCode didn\u0026rsquo;t complicate a simple idea—it revealed that the idea was never actually simple. Abstract thinking deferred decisions about memory, concurrency, staleness, observability, and eviction. Code forced those decisions into concrete form where consequences become visible and measurable.\u003c/p\u003e\n\u003cp\u003eThis is not implementation detail obscuring elegant design. This is reality asserting itself.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"what-comes-next-the-feedback-loop-ai-cannot-replicate\"\u003e\u003ca href=\"/posts/code-sharpens-thinking/#what-comes-next-the-feedback-loop-ai-cannot-replicate\" title=\"What Comes Next: The Feedback Loop AI Cannot Replicate\"\u003eWhat Comes Next: The Feedback Loop AI Cannot Replicate\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eAI-generated code without understanding creates productivity illusions that collapse in production. Code forces abstract thinking into executable form, exposing gaps that pure reasoning glosses over. That much is clear.\u003c/p\u003e\n\u003cp\u003eBut understanding the problem doesn\u0026rsquo;t answer the deeper question: What exactly is this feedback loop between code and reality, and why can\u0026rsquo;t AI replicate it? What mechanisms transform vague reasoning into concrete understanding?\u003c/p\u003e\n\u003cp\u003eThe answer lies in the tools we use every day: compilers, profilers, tests, production environments. These aren\u0026rsquo;t just validation gates. They\u0026rsquo;re \u003cstrong\u003ereality engines\u003c/strong\u003e that do something AI fundamentally cannot: they execute your assumptions against actual constraints and report back with unfiltered truth.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eIn the next part of this series, we\u0026rsquo;ll explore how these mechanisms form a cognitive feedback loop that sharpens professional thinking in ways no AI prompt can simulate.\u003c/em\u003e\u003c/p\u003e\n","date_modified":"2026-05-26T10:22:03+02:00","date_published":"2026-01-06T17:00:00+01:00","id":"https://daily-devops.net/posts/code-sharpens-thinking/","language":"en","summary":"Typing code is trivial now—AI does it instantly. So why will real professionals never be replaced? Because vibe coding collapses under production reality.\n","tags":["softwareengineering","codequality","bestpractices","architecture","dotnet","csharp","technicaldebt","ai-code-assistant","github-copilot"],"title":"Why Real Professionals Will Never Be Replaced by AI\n","url":"https://daily-devops.net/posts/code-sharpens-thinking/"},{"authors":[{"name":"Martin Stühmer","url":"https://daily-devops.net/authors/martin/"}],"content_html":"\u003cp\u003eArchitectural Decision Records (ADRs) capture the \u0026ldquo;why\u0026rdquo; behind your technical choices—documenting decisions, rationale, and context for future reference. They should guide teams through complex landscapes, inform new decisions, and provide clarity during audits or onboarding.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBut here\u0026rsquo;s the problem:\u003c/strong\u003e Most ADRs become digital dust collectors.\u003c/p\u003e\n\u003cp\u003eThey sit in repositories, referenced only during crisis meetings or compliance audits. Developers bypass them during daily work, and automation tools ignore them completely. The gap between architectural intent and daily practice grows wider every sprint.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"understanding-the-problem\"\u003e\u003ca href=\"/posts/instruction-by-design/#understanding-the-problem\" title=\"Understanding the Problem\"\u003eUnderstanding the Problem\u003c/a\u003e\u003c/h2\u003e\n\n\n\n\n\u003ch3 id=\"the-core-challenge\"\u003e\u003ca href=\"/posts/instruction-by-design/#the-core-challenge\" title=\"The Core Challenge\"\u003eThe Core Challenge\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eTraditional ADRs are \u003cstrong\u003epassive documentation\u003c/strong\u003e—they record what happened, but don\u0026rsquo;t actively shape what happens next:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e\u003cstrong\u003eDiscovery friction:\u003c/strong\u003e New team members must hunt through scattered documents to understand current standards\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eEnforcement gaps:\u003c/strong\u003e Build systems and linters operate independently of architectural decisions\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eConsistency drift:\u003c/strong\u003e Without active reinforcement, even well-documented standards gradually erode across teams\u003c/li\u003e\n\u003c/ul\u003e\n\n\n\n\n\u003ch3 id=\"the-vision-adrs-that-actually-work\"\u003e\u003ca href=\"/posts/instruction-by-design/#the-vision-adrs-that-actually-work\" title=\"The Vision: ADRs That Actually Work\"\u003eThe Vision: ADRs That Actually Work\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eImagine ADRs that don\u0026rsquo;t just document decisions—they \u003cstrong\u003edrive\u003c/strong\u003e them:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eEvery architectural choice directly influences code suggestions from AI Code Assistant\u003c/li\u003e\n\u003cli\u003eNew developers instantly understand current standards through integrated guidance\u003c/li\u003e\n\u003cli\u003eAutomation systems enforce architectural decisions in real time\u003c/li\u003e\n\u003cli\u003eTeams work with consistent, up-to-date guidance embedded in their daily tools\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThis isn\u0026rsquo;t just better documentation—it\u0026rsquo;s \u003cstrong\u003eoperational architecture\u003c/strong\u003e. By making ADRs machine-consumable and embedding clear instructions, they become the single source of truth that powers both human understanding and automated enforcement.\u003c/p\u003e\n\u003cp\u003eThe result? Development environments where architectural intent is always clear, actionable, and automatically aligned across every team member and tool.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"why-traditional-adrs-fall-short\"\u003e\u003ca href=\"/posts/instruction-by-design/#why-traditional-adrs-fall-short\" title=\"Why Traditional ADRs Fall Short\"\u003eWhy Traditional ADRs Fall Short\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eThe gap between architectural intent and daily practice is where most projects struggle. Traditional ADRs capture decisions brilliantly but fail to integrate them into the development workflow where they matter most.\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e\u003cstrong\u003ePassive documentation:\u003c/strong\u003e ADRs become historical artifacts that developers consult only during crisis or retrospectives—if at all.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eDisconnected from automation:\u003c/strong\u003e Build systems, linters, and AI tools operate independently of architectural decisions, missing opportunities to enforce standards automatically.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eOnboarding friction:\u003c/strong\u003e New team members must manually discover and interpret scattered decisions, slowing their ability to contribute effectively.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eInconsistent application:\u003c/strong\u003e Without active reinforcement, even well-documented decisions gradually drift or get forgotten across different teams and projects.\u003c/li\u003e\n\u003c/ul\u003e\n\n\n\n\n\u003ch2 id=\"the-solution-instruction-by-design\"\u003e\u003ca href=\"/posts/instruction-by-design/#the-solution-instruction-by-design\" title=\"The Solution: Instruction by Design\"\u003eThe Solution: Instruction by Design\u003c/a\u003e\u003c/h2\u003e\n\n\n\n\n\u003ch3 id=\"instruction-by-design-from-records-to-directives\"\u003e\u003ca href=\"/posts/instruction-by-design/#instruction-by-design-from-records-to-directives\" title=\"Instruction by Design: From Records to Directives\"\u003eInstruction by Design: From Records to Directives\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eThe transformation begins when we stop thinking of ADRs as documentation and start treating them as executable specifications for both human and AI behavior.\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e\u003cstrong\u003eMachine-consumable structure:\u003c/strong\u003e Every ADR includes structured metadata and clear instructions that AI Code Assistant can parse and apply immediately.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eOperational states with meaning:\u003c/strong\u003e \u0026ldquo;Accepted\u0026rdquo; decisions become mandatory requirements, \u0026ldquo;proposed\u0026rdquo; become considerations, while \u0026ldquo;deprecated\u0026rdquo; and \u0026ldquo;superseded\u0026rdquo; trigger active avoidance patterns.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eDirect workflow integration:\u003c/strong\u003e Decisions automatically influence code suggestions, review processes, and validation pipelines without manual intervention.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eSingle source of truth:\u003c/strong\u003e Both developers and AI agents reference the same authoritative guidance, eliminating interpretation gaps and ensuring consistent application.\u003c/li\u003e\n\u003c/ul\u003e\n\n\n\n\n\u003ch3 id=\"the-ai-enforcement-layer\"\u003e\u003ca href=\"/posts/instruction-by-design/#the-ai-enforcement-layer\" title=\"The AI Enforcement Layer\"\u003eThe AI Enforcement Layer\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eWhen architectural decisions become machine-readable, they can drive intelligent automation throughout your development process:\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-markdown\" data-lang=\"markdown\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"gu\"\u003e## Decision References\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e*\u003c/span\u003e MUST document all decisions in \u003cspan class=\"sb\"\u003e`decisions/`\u003c/span\u003e folder using \u003cspan class=\"sb\"\u003e`templates/architecture-decision.md`\u003c/span\u003e format.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e*\u003c/span\u003e MUST treat \u0026#34;accepted\u0026#34; decisions as mandatory requirements with highest precedence.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e*\u003c/span\u003e MUST respect decision states:\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e  \u003cspan class=\"k\"\u003e-\u003c/span\u003e **accepted**: mandatory requirements\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e  \u003cspan class=\"k\"\u003e-\u003c/span\u003e **proposed**: optional considerations\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e  \u003cspan class=\"k\"\u003e-\u003c/span\u003e **deprecated**: avoid in new implementations\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e  \u003cspan class=\"k\"\u003e-\u003c/span\u003e **superseded**: forbidden, follow superseding decision instead\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e*\u003c/span\u003e MUST use the \u003cspan class=\"sb\"\u003e`instructions`\u003c/span\u003e frontmatter property as primary AI guidance for each decision.\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eThese rules make every ADR actionable. Human or AI, your team always knows what matters most. Now the journey with your AI buddy begins with clear, actionable guidance. Without the day-to-day friction of interpreting static documents, your team can focus on what really matters: building great software.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"the-enhanced-adr-template-built-for-action\"\u003e\u003ca href=\"/posts/instruction-by-design/#the-enhanced-adr-template-built-for-action\" title=\"The Enhanced ADR Template: Built for Action\"\u003eThe Enhanced ADR Template: Built for Action\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eThe template itself is full of helpful instructions and required fields, for clarity and standardization:\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-markdown\" data-lang=\"markdown\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u0026lt;!-- List of authors who contributed to this decision. Include full names and roles if applicable. --\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eauthors:\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e Name Surname \u0026lt;!-- Replace with actual name --\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e Another Name Surname \u0026lt;!-- Add more authors as needed --\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u0026lt;!--\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eThe patterns this decision applies to. Each entry is a glob pattern that matches files affected by this decision.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eExample:\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eapplyTo:\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e \u0026#34;**/*.cs\u0026#34;          # Applies to all C# files\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e \u0026#34;src/**/*.razor\u0026#34;   # Applies to all Blazor components in src folder\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e \u0026#34;tests/**/*.sql\u0026#34;   # Applies to all SQL files in tests folder\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e--\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eapplyTo:\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e \u0026#34;**/*\u0026#34; \u0026lt;!-- Replace with specific glob patterns --\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u0026lt;!-- The date this ADR was initially created in YYYY-MM-DD format. --\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003ecreated: YYYY-MM-DD\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u0026lt;!--\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eThe most recent date this ADR was updated in YYYY-MM-DD format.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eIMPORTANT: Update this field whenever the decision is modified.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e--\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003elastModified: YYYY-MM-DD\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u0026lt;!--\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eThe current state of this ADR. If superseded, include references to the superseding ADR.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eValid values: proposed, accepted, deprecated, superseded\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e--\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003estate: proposed\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u0026lt;!--\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eA compact AI LLM compatible definition of this decision.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eThis should be a precise, structured description that AI systems can easily parse and understand.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eInclude the core decision, key rationale, and primary impact in 1-2 concise sentences.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e--\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003einstructions: |\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e  Compact definition of the decision made and its core purpose.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e  Key rationale and primary impact on the project or development process.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e---\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u0026lt;!-- REQUIRED: Filename MUST follow the format: YYYY-MM-DD-Title (replace all spaces with hyphens) --\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"gh\"\u003e# Title \u0026lt;!-- A concise title that summarizes the decision. Use a format like \u0026#34;Decision: [Short Description of Decision]\u0026#34;. --\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u0026lt;!--\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eA brief summary of the decision. This should be a short paragraph that captures the essence of the decision made.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e--\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"gu\"\u003e## Context\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u0026lt;!--\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eProvide a detailed explanation of the problem or issue that led to this decision. Include background information, constraints, and any relevant context to help readers understand why this decision was necessary.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e--\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"gu\"\u003e## Decision\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u0026lt;!--\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eClearly state the decision made. Describe the chosen solution or approach in detail, including any specific technologies, tools, or methods involved.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e--\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"gu\"\u003e## Consequences\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u0026lt;!--\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eExplain the implications of this decision. What are the expected benefits, trade-offs, and potential risks? How will this decision impact the project or organization?\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e--\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"gu\"\u003e## Alternatives Considered\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u0026lt;!--\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eList and describe other options that were considered. For each alternative, explain why it was not chosen. Include pros and cons, feasibility, and any other relevant factors.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e--\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"gu\"\u003e## Related Decisions (Optional)\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u0026lt;!--\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eProvide links or references to other ADRs that are related to this decision. Explain how they are connected and why they are relevant.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eUse markdown link syntax to reference other decisions: [\u003cspan class=\"nt\"\u003eDecision Title\u003c/span\u003e](\u003cspan class=\"na\"\u003e./YYYY-MM-DD-decision-filename.md\u003c/span\u003e)\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eIf there are no related decisions, this section may be omitted or include a note stating \u0026#34;None at this time.\u0026#34;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eExample:\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e [\u003cspan class=\"nt\"\u003eCentralized Package Version Management\u003c/span\u003e](\u003cspan class=\"na\"\u003e./2025-07-10-centralized-package-version-management.md\u003c/span\u003e) - Related because this decision impacts how we manage dependencies\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e [\u003cspan class=\"nt\"\u003eConventional Commits\u003c/span\u003e](\u003cspan class=\"na\"\u003e./2025-07-10-conventional-commits.md\u003c/span\u003e) - This decision affects our commit message format which impacts versioning\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e--\u0026gt;\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003e\u003cstrong\u003eWhy this structure matters:\u003c/strong\u003e\nEvery field drives your team toward actionable clarity. No more vague rationale, ambiguous decisions, or documentation drift. The template itself becomes machine-readable—AI Code Assistant can parse every element directly, while humans get the structure they need to make consistent, enforceable decisions.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"concrete-example-english-as-project-language\"\u003e\u003ca href=\"/posts/instruction-by-design/#concrete-example-english-as-project-language\" title=\"Concrete Example: English as Project Language\"\u003eConcrete Example: English as Project Language\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eLet’s see how this works with a real, high-impact example.\u003c/p\u003e\n\u003cdiv class=\"highlight\"\u003e\u003cpre tabindex=\"0\" class=\"chroma\"\u003e\u003ccode class=\"language-markdown\" data-lang=\"markdown\"\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eauthors:\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e Jane Doe, Solution Architect\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e John Smith, Lead Developer\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eapplyTo:\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e \u0026#34;**/*\u0026#34;\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003ecreated: 2025-07-15\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003elastModified: 2025-07-15\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003estate: accepted\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003einstructions: |\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e  Establish English as the mandatory language for all code, documentation, comments, commit messages, and written content to ensure consistency and global accessibility.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e  Applies to all identifiers, configuration files, database objects, and communication using clear, professional English standards.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e---\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"gh\"\u003e# Decision: English as Project Language\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eAll project artifacts, including code, docs, configs, database objects, and commit messages, must use clear, professional English. This enables global collaboration, faster onboarding, and consistent reviews—by both people and AI.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"gu\"\u003e## Context\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eFragmented language use has slowed down onboarding, increased misunderstandings, and made collaboration harder across regions. A single language standard solves these problems.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"gu\"\u003e## Decision\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003eAll content—code, comments, documentation, configs, and communication—must be in English, using clear and professional standards.\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"gu\"\u003e## Consequences\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"gs\"\u003e**Benefits:**\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e Global teams onboard faster and communicate better\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e Automated tools and AI Code Assistant can parse, review, and generate content reliably\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e Fewer mistakes, less rework, and smoother audits\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"gs\"\u003e**Trade-offs/Risks:**\u003c/span\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e Non-native English speakers may need support\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e Existing teams may need time to adapt\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"gu\"\u003e## Alternatives Considered\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e Local language flexibility: Increased confusion and audit risk\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e Bilingual documentation: High maintenance, likely to get out of sync\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"gu\"\u003e## Related Decisions\n\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\n\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"line\"\u003e\u003cspan class=\"cl\"\u003e\u003cspan class=\"k\"\u003e-\u003c/span\u003e [\u003cspan class=\"nt\"\u003eCentralized Documentation Standards\u003c/span\u003e](\u003cspan class=\"na\"\u003e./2025-07-10-centralized-documentation-standards.md\u003c/span\u003e)\n\u003c/span\u003e\u003c/span\u003e\u003c/code\u003e\u003c/pre\u003e\u003c/div\u003e\u003cp\u003eThis ADR is a perfect example of \u003cstrong\u003eInstruction by Design\u003c/strong\u003e. It’s not just a record of a decision; it’s a directive that shapes how your team works every day. By specifying that all project artifacts must be in English, it sets clear expectations for both human developers and AI Code Assistant.\u003c/p\u003e\n\u003cp\u003eIt eliminates ambiguity, reduces friction, and ensures that everyone—regardless of their native language—can contribute effectively.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"why-this-matters\"\u003e\u003ca href=\"/posts/instruction-by-design/#why-this-matters\" title=\"Why This Matters\"\u003eWhy This Matters\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eThis ADR is more than just a decision; it’s a \u003cstrong\u003estandard\u003c/strong\u003e that your team can rely on. It’s clear, actionable, and enforceable. By using this template, you ensure that every architectural decision is not only documented but also actively shapes your development process.\u003c/p\u003e\n\u003cp\u003eIt’s a living document that evolves with your project, guiding both human and AI agents toward consistent, high-quality outcomes.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"how-ai-and-automation-put-this-to-work\"\u003e\u003ca href=\"/posts/instruction-by-design/#how-ai-and-automation-put-this-to-work\" title=\"How AI and Automation Put This to Work\"\u003eHow AI and Automation Put This to Work\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eWith Instruction by Design, your ADRs become living documents that AI Code Assistant can use to guide development. Here’s how it works in practice:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eAutomatic code and docs checks:\u003c/strong\u003e\nAI Code Assistant flag any non-English content and suggest improvements in real time. Always considering the ADRs as the source of truth.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eGuided pull requests:\u003c/strong\u003e\nEvery reviewer can refer to the ADR for clear, objective decisions. Without friction, they can align on expectations and requirements quickly.\u003c/p\u003e\n\u003c/li\u003e\n\u003cli\u003e\n\u003cp\u003e\u003cstrong\u003eFast onboarding:\u003c/strong\u003e\nNew developers and AI agents see the language policy immediately—and know it’s enforced.\u003c/p\u003e\n\u003c/li\u003e\n\u003c/ul\u003e\n\n\n\n\n\u003ch2 id=\"the-bigger-picture-operationalizing-architecture\"\u003e\u003ca href=\"/posts/instruction-by-design/#the-bigger-picture-operationalizing-architecture\" title=\"The Bigger Picture: Operationalizing Architecture\"\u003eThe Bigger Picture: Operationalizing Architecture\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eInstruction by Design is about more than just better ADRs. It’s about \u003cstrong\u003eoperationalizing architecture\u003c/strong\u003e—making your architectural decisions active participants in your development process.\nBy embedding clear, actionable instructions into every ADR, you create a system where architectural intent is always clear, enforceable, and aligned with your team’s daily work.\u003c/p\u003e\n\u003cp\u003eThis approach transforms ADRs from passive records into active guides that shape both human and AI behavior. It ensures that every decision is not just documented but also \u003cstrong\u003eoperationalized\u003c/strong\u003e—driving consistent, high-quality outcomes across your development teams.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"benefits-of-instruction-by-design\"\u003e\u003ca href=\"/posts/instruction-by-design/#benefits-of-instruction-by-design\" title=\"Benefits of Instruction by Design\"\u003eBenefits of Instruction by Design\u003c/a\u003e\u003c/h2\u003e\n\u003cul\u003e\n\u003cli\u003e\u003cstrong\u003eConsistency:\u003c/strong\u003e Every team member and AI agent follows the same standards, reducing drift and confusion\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eClarity:\u003c/strong\u003e Clear, actionable instructions eliminate ambiguity and ensure everyone knows what’s expected\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eAutomation:\u003c/strong\u003e AI Code Assistant can enforce decisions in real time, catching issues before they become problems\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eEvolution:\u003c/strong\u003e As projects grow, ADRs evolve with them—ensuring that architectural intent remains clear and actionable\u003c/li\u003e\n\u003c/ul\u003e\n\n\n\n\n\u003ch2 id=\"conclusion-transform-your-adrs-today\"\u003e\u003ca href=\"/posts/instruction-by-design/#conclusion-transform-your-adrs-today\" title=\"Conclusion: Transform Your ADRs Today\"\u003eConclusion: Transform Your ADRs Today\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eInstruction by Design is a game-changer for how we think about architectural decision records. By transforming ADRs into actionable, AI-ready guidance, we bridge the gap between architectural intent and daily practice.\nNo more passive documentation—now your ADRs actively shape how your teams work, ensuring consistency, clarity, and quality across every project.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDon’t just document. Operationalize.\u003c/strong\u003e - Turn your ADRs into active guides for your people and your tools—because real progress comes from decisions you actually use. \u003cstrong\u003eReady to transform your ADRs?\u003c/strong\u003e\u003c/p\u003e\n","date_modified":"2026-05-26T10:22:03+02:00","date_published":"2025-07-15T10:30:00+02:00","id":"https://daily-devops.net/posts/instruction-by-design/","language":"en","summary":"Transform architectural decision records (ADRs) into actionable AI guidance for enhanced team consistency, streamlined onboarding, and automated workflows.","tags":["ai-code-assistant","architecture","bestpractices","github","github-copilot","rcda","softwareengineering","technicaldebt"],"title":"Instruction by Design: Transforming ADRs into Actionable AI Guidance","url":"https://daily-devops.net/posts/instruction-by-design/"},{"authors":[{"name":"Martin Stühmer","url":"https://daily-devops.net/authors/martin/"}],"content_html":"\u003cp\u003eIn a previous article, we laid it out – unfiltered: \u003ca href=\"https://daily-devops.net/posts/copilot-turns-junior-devs-into-syntax-secretaries/\" target=\"_blank\" rel=\"noopener external noreferrer\"\u003e\u003cstrong\u003eCopilot turns junior devs into syntax secretaries.\u003c/strong\u003e\u003c/a\u003e\u003c/p\u003e\n\u003cp\u003eNot because it’s evil. But because it \u003cstrong\u003eremoves friction before understanding\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eIt gives you working code before you know what \u003cem\u003eworking\u003c/em\u003e even means. It creates the illusion of progress, while slowly eroding the very skills that define a software engineer: reasoning, decision-making, and technical ownership.\u003c/p\u003e\n\u003cp\u003eThat critique still stands.\u003c/p\u003e\n\u003cp\u003eBut here’s the catch: \u003cstrong\u003eThe same tool that disables junior developers can empower senior engineers – if they know what they’re doing.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe key isn’t the tool. It’s \u003cstrong\u003ewho’s holding the keyboard\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eThis article is about reclaiming Copilot – not as a crutch, but as a force multiplier.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"seven-steps-to-master-copilot\"\u003e\u003ca href=\"/posts/copilot-without-becoming-its-puppet/#seven-steps-to-master-copilot\" title=\"Seven Steps to Master Copilot\"\u003eSeven Steps to Master Copilot\u003c/a\u003e\u003c/h2\u003e\n\n\n\n\n\u003ch3 id=\"step-1-stop-worshipping-the-output\"\u003e\u003ca href=\"/posts/copilot-without-becoming-its-puppet/#step-1-stop-worshipping-the-output\" title=\"Step 1: Stop Worshipping the Output\"\u003eStep 1: Stop Worshipping the Output\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eLet’s get one thing straight: \u003cstrong\u003eCopilot is not \u0026ldquo;AI\u0026rdquo;.\u003c/strong\u003e It’s a token prediction engine trained on millions of public repositories – including the bad ones.\u003c/p\u003e\n\u003cp\u003eIt doesn’t:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eUnderstand your business logic\u003c/li\u003e\n\u003cli\u003eKnow your system constraints\u003c/li\u003e\n\u003cli\u003eRespect your architecture\u003c/li\u003e\n\u003cli\u003eOr care if your code silently corrupts production data\u003c/li\u003e\n\u003c/ul\u003e\n\u003cbr\u003e\n\u003cp\u003eCopilot doesn’t think. It \u003cem\u003eguesses\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eThat means every suggestion it makes should be treated as \u003cstrong\u003eguilty until proven useful\u003c/strong\u003e.\u003c/p\u003e\n\u003cbr\u003e\n\u003cblockquote\u003e\n\u003cp\u003eWould you deploy code written by a clueless intern who sounds confident?\n\u003cem\u003eThen don’t blindly accept Copilot output either.\u003c/em\u003e\u003c/p\u003e\n\u003c/blockquote\u003e\n\n\n\n\n\u003ch3 id=\"step-2-dont-let-the-tool-set-the-pace\"\u003e\u003ca href=\"/posts/copilot-without-becoming-its-puppet/#step-2-dont-let-the-tool-set-the-pace\" title=\"Step 2: Don’t Let the Tool Set the Pace\"\u003eStep 2: Don’t Let the Tool Set the Pace\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eOne of the most subtle traps Copilot sets for senior devs is \u003cstrong\u003evelocity addiction\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eIt gives you a dopamine rush: you type three letters, and a full method appears.\nIt’s seductive. It feels efficient. It feels productive.\u003c/p\u003e\n\u003cp\u003eBut here’s the reality:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eDid you just skip the error-handling strategy?\u003c/li\u003e\n\u003cli\u003eDid you consider testability?\u003c/li\u003e\n\u003cli\u003eDid you choose the right abstraction layer?\u003c/li\u003e\n\u003cli\u003eDid you even \u003cem\u003ename\u003c/em\u003e things meaningfully?\u003c/li\u003e\n\u003c/ul\u003e\n\u003cbr\u003e\n\u003cp\u003eIf you didn’t stop and ask those questions, Copilot didn’t make you faster.\nIt made you \u003cem\u003elazy\u003c/em\u003e.\nAnd lazy senior engineers are more dangerous than clueless juniors – because they ship code that looks trustworthy.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"step-3-use-it-to-offload--not-outsource--thinking\"\u003e\u003ca href=\"/posts/copilot-without-becoming-its-puppet/#step-3-use-it-to-offload--not-outsource--thinking\" title=\"Step 3: Use It to Offload – Not Outsource – Thinking\"\u003eStep 3: Use It to Offload – Not Outsource – Thinking\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eThe real value of Copilot begins when you \u003cstrong\u003ealready know\u003c/strong\u003e what you’re doing.\u003c/p\u003e\n\u003cp\u003eIf you’ve built a hundred layered service implementations in .NET Core – by all means, let Copilot generate the scaffolding.\u003c/p\u003e\n\u003cp\u003eIf you’re writing a test fixture with tedious mocking boilerplate – fine, autocomplete away.\u003c/p\u003e\n\u003cp\u003eBut when you’re:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eDesigning a concurrency model\u003c/li\u003e\n\u003cli\u003eCrafting a DSL\u003c/li\u003e\n\u003cli\u003eBuilding a distributed system component\u003c/li\u003e\n\u003cli\u003eDefining a new domain abstraction\u003c/li\u003e\n\u003c/ul\u003e\n\u003cbr\u003e\n\u003cp\u003eThen \u003cstrong\u003eCopilot has no business driving.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eUse it for:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eLow-brainpower scaffolding\u003c/li\u003e\n\u003cli\u003eRepetitive composition\u003c/li\u003e\n\u003cli\u003eAuto-generating test stubs\u003c/li\u003e\n\u003cli\u003eExploring syntactic variations\u003c/li\u003e\n\u003c/ul\u003e\n\u003cbr\u003e\n\u003cp\u003eBut \u003cem\u003eyou\u003c/em\u003e decide the design. Not the suggestion.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"step-4-prompt-like-a-professional-not-a-prompt-engineer\"\u003e\u003ca href=\"/posts/copilot-without-becoming-its-puppet/#step-4-prompt-like-a-professional-not-a-prompt-engineer\" title=\"Step 4: Prompt Like a Professional, Not a Prompt Engineer\"\u003eStep 4: Prompt Like a Professional, Not a Prompt Engineer\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eCopilot doesn’t \u0026ldquo;understand\u0026rdquo; context. It responds to \u003cstrong\u003epatterns\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eSo don’t write vague half-sentences like \u0026ldquo;make this better\u0026rdquo; and expect miracles.\u003c/p\u003e\n\u003cp\u003eInstead, treat Copilot prompts like function signatures:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eBe explicit\u003c/li\u003e\n\u003cli\u003eBe scoped\u003c/li\u003e\n\u003cli\u003eAssume ambiguity is punished\u003c/li\u003e\n\u003c/ul\u003e\n\u003cbr\u003e\n\u003cblockquote\u003e\n\u003cp\u003e\u003cstrong\u003e✅ Example prompt:\u003c/strong\u003e Write a thread-safe async method in C# that wraps a third-party API call with exponential backoff using Polly, and logs all non-transient failures via Serilog.\u003c/p\u003e\n\u003c/blockquote\u003e\n\u003cbr\u003e\n\u003cp\u003eThat prompt gets you \u003cem\u003eleverage\u003c/em\u003e.\nBecause you’re setting the architectural contract.\nCopilot just fills in the boring parts.\u003c/p\u003e\n\u003cbr\u003e\n\u003cblockquote\u003e\n\u003cp\u003e\u003cstrong\u003e🚫 What not to do:\u003c/strong\u003e Call API with retries.\u003c/p\u003e\n\u003c/blockquote\u003e\n\u003cbr\u003e\n\u003cp\u003eThat’s how you end up with retry-on-404 garbage logic that silently fails in production.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"step-5-copilot-is-a-mirror--train-it-to-reflect-quality\"\u003e\u003ca href=\"/posts/copilot-without-becoming-its-puppet/#step-5-copilot-is-a-mirror--train-it-to-reflect-quality\" title=\"Step 5: Copilot Is a Mirror – Train It to Reflect Quality\"\u003eStep 5: Copilot Is a Mirror – Train It to Reflect Quality\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eCopilot doesn’t invent style – it reflects what it sees.\u003c/p\u003e\n\u003cp\u003eWhich means:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eIf your codebase is clean, expressive, well-factored – Copilot suggestions improve.\u003c/li\u003e\n\u003cli\u003eIf it’s inconsistent, under-tested, or polluted with lazy shortcuts – Copilot amplifies that rot.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cbr\u003e\n\u003cp\u003eIn other words:\n\u003cstrong\u003eYour discipline teaches the tool what \u0026ldquo;normal\u0026rdquo; looks like.\u003c/strong\u003e\u003c/p\u003e\n\u003cbr\u003e\n\u003cp\u003eSo:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eUse explicit, descriptive method names\u003c/li\u003e\n\u003cli\u003eKeep test coverage high\u003c/li\u003e\n\u003cli\u003eEnforce boundaries\u003c/li\u003e\n\u003cli\u003eWrite proper failure paths\u003c/li\u003e\n\u003cli\u003eMaintain clear separation of concerns\u003c/li\u003e\n\u003c/ul\u003e\n\u003cbr\u003e\n\u003cp\u003eIf your code smells like engineering, Copilot will start to autocomplete \u003cem\u003eengineering\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eIf it smells like Stack Overflow duct tape – well, you already know the result.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"step-6-review-it-like-it-was-written-by-a-liar\"\u003e\u003ca href=\"/posts/copilot-without-becoming-its-puppet/#step-6-review-it-like-it-was-written-by-a-liar\" title=\"Step 6: Review It Like It Was Written by a Liar\"\u003eStep 6: Review It Like It Was Written by a Liar\u003c/a\u003e\u003c/h3\u003e\n\u003cblockquote\u003e\n\u003cp\u003e\u003cstrong\u003eHere’s a non-negotiable rule:\u003c/strong\u003e Every Copilot suggestion must be reviewed as if it came from someone trying to impress you without understanding the problem.\u003c/p\u003e\n\u003c/blockquote\u003e\n\u003cbr\u003e\n\u003cp\u003eThat means:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eCheck for race conditions\u003c/li\u003e\n\u003cli\u003eExamine exception handling\u003c/li\u003e\n\u003cli\u003eValidate parameter boundaries\u003c/li\u003e\n\u003cli\u003eWatch for leaky abstractions\u003c/li\u003e\n\u003cli\u003eAssess performance under real-world constraints\u003c/li\u003e\n\u003c/ul\u003e\n\u003cbr\u003e\n\u003cp\u003eJust because the code compiles doesn’t mean it’s correct.\u003c/p\u003e\n\u003cp\u003eJust because it runs doesn’t mean it scales.\u003c/p\u003e\n\u003cp\u003eJust because it works doesn’t mean it’s safe.\u003c/p\u003e\n\u003cp\u003eYou’re the engineer. Copilot is just a code monkey with good grammar.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"step-7-know-when-to-walk-away\"\u003e\u003ca href=\"/posts/copilot-without-becoming-its-puppet/#step-7-know-when-to-walk-away\" title=\"Step 7: Know When to Walk Away\"\u003eStep 7: Know When to Walk Away\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eGreat developers know \u003cstrong\u003ewhen not to automate\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eDo not use Copilot:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eIn security-sensitive logic (encryption, auth flows, claims handling)\u003c/li\u003e\n\u003cli\u003eWhen designing new public interfaces\u003c/li\u003e\n\u003cli\u003eWhile writing infrastructure-as-code\u003c/li\u003e\n\u003cli\u003eFor architecture decisions\u003c/li\u003e\n\u003cli\u003eWhen you’re unclear about the problem domain\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eIn these cases, Copilot isn’t helpful.\u003c/p\u003e\n\u003cp\u003eIt’s noise. Distraction. \u003cem\u003eA confident liar offering false shortcuts.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAnd your job as a senior engineer is to guard against shortcuts that violate long-term quality.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"final-word-you-dont-need-copilot-thats-why-you-can-use-it\"\u003e\u003ca href=\"/posts/copilot-without-becoming-its-puppet/#final-word-you-dont-need-copilot-thats-why-you-can-use-it\" title=\"Final Word: You Don’t Need Copilot. That’s Why You Can Use It\"\u003eFinal Word: You Don’t Need Copilot. That’s Why You Can Use It\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eIf you still think Copilot is \u003cem\u003eyour coding assistant,\u003c/em\u003e you’ve missed the point.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIt’s not your peer. It’s your tool.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eYou already know how to:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eWrite expressive code\u003c/li\u003e\n\u003cli\u003eDesign reliable systems\u003c/li\u003e\n\u003cli\u003eReview and refactor ruthlessly\u003c/li\u003e\n\u003cli\u003eQuestion defaults\u003c/li\u003e\n\u003cli\u003eOwn the outcome\u003c/li\u003e\n\u003c/ul\u003e\n\u003cbr\u003e\n\u003cp\u003eThat’s why you can use Copilot \u003cem\u003ewithout losing yourself in it\u003c/em\u003e.\u003c/p\u003e\n\u003cbr\u003e\n\u003cp\u003eBecause at the end of the day: \u003cem\u003e\u003cstrong\u003eTools don’t build software. Engineers do.\u003c/strong\u003e\u003c/em\u003e\u003c/p\u003e\n\u003cbr\u003e\n\u003cp\u003eAnd if we want to preserve the quality of our craft in a world of AI-assisted mediocrity, we need to lead by example.\u003c/p\u003e\n\u003cbr\u003e\n\u003cp\u003eCopilot can help you go faster – but only \u003cstrong\u003eafter\u003c/strong\u003e you’ve done the work to know where you\u0026rsquo;re going.\u003c/p\u003e\n","date_modified":"2026-05-20T21:28:40+02:00","date_published":"2025-05-14T17:30:00+02:00","id":"https://daily-devops.net/posts/copilot-without-becoming-its-puppet/","language":"en","summary":"Master GitHub Copilot as a productivity tool while maintaining your coding skills, critical thinking abilities, and commitment to software craftsmanship.","tags":["ai-code-assistant","bestpractices","dotnet","github","github-copilot","testing","visualstudio"],"title":"How to Use Copilot Without Becoming Its Puppet","url":"https://daily-devops.net/posts/copilot-without-becoming-its-puppet/"},{"authors":[{"name":"Martin Stühmer","url":"https://daily-devops.net/authors/martin/"}],"content_html":"\u003cp\u003eThe hype around GitHub Copilot (or any other AI code assistant) is deafening. AI-assisted coding. Effortless automation. \u003cem\u003e\u003cstrong\u003e10x productivity.\u003c/strong\u003e\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eBut here’s the harsh truth: \u003cstrong\u003eCopilot isn’t empowering junior developers – it’s deskilling them.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eNo shortcuts. No sugarcoating. Just software done right.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIt’s not making them engineers. It’s turning them into \u003cem\u003esyntax secretaries\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eThey type. The tool fills in the blanks. They deploy.\nNo understanding. No design thinking. No learning.\u003c/p\u003e\n\u003cp\u003eLet’s break this down – in code, in context, and in consequence.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"it-writes-code-but-youre-supposed-to-write-systems\"\u003e\u003ca href=\"/posts/copilot-turns-junior-devs-into-syntax-secretaries/#it-writes-code-but-youre-supposed-to-write-systems\" title=\"It Writes Code. But You’re Supposed to Write Systems\"\u003eIt Writes Code. But You’re Supposed to Write Systems\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eProgramming isn’t about writing lines of code. It’s about designing systems.\nIt’s about thinking through:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eState transitions\u003c/li\u003e\n\u003cli\u003eFault tolerance\u003c/li\u003e\n\u003cli\u003eTrade-offs between readability, performance, and maintainability\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eCopilot bypasses all of that.\u003c/p\u003e\n\n\n\n\n\u003ch3 id=\"example-from-net-build-a-caching-layer-that-wraps-a-call-to-an-external-api\"\u003e\u003ca href=\"/posts/copilot-turns-junior-devs-into-syntax-secretaries/#example-from-net-build-a-caching-layer-that-wraps-a-call-to-an-external-api\" title=\"Example from .NET: Build a caching layer that wraps a call to an external API\"\u003eExample from .NET: Build a caching layer that wraps a call to an external API\u003c/a\u003e\u003c/h3\u003e\n\u003cp\u003eCopilot gives you:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ccode\u003eIMemoryCache\u003c/code\u003e\u003c/li\u003e\n\u003cli\u003e\u003ccode\u003eHttpClient\u003c/code\u003e\u003c/li\u003e\n\u003cli\u003easync/await\u003c/li\u003e\n\u003cli\u003emaybe some exception handling\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eBut it \u003cstrong\u003edoesn’t\u003c/strong\u003e make you think about:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eTransient fault handling (\u003ccode\u003ePolly\u003c/code\u003e)\u003c/li\u003e\n\u003cli\u003eHttpClient reuse and DNS updates\u003c/li\u003e\n\u003cli\u003eCache invalidation semantics\u003c/li\u003e\n\u003cli\u003eTimezone-safe comparisons\u003c/li\u003e\n\u003cli\u003eWhat happens if the API fails silently for two hours\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eIt gives you code that \u003cem\u003elooks right\u003c/em\u003e – and that’s the most dangerous kind of code.\nBecause junior developers don’t know what to doubt.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"youre-not-learning-the-why-youre-memorizing-the-what\"\u003e\u003ca href=\"/posts/copilot-turns-junior-devs-into-syntax-secretaries/#youre-not-learning-the-why-youre-memorizing-the-what\" title=\"You\u0026rsquo;re Not Learning the Why. You\u0026rsquo;re Memorizing the What\"\u003eYou\u0026rsquo;re Not Learning the Why. You\u0026rsquo;re Memorizing the What\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eWhen Copilot gives you an answer, there’s no struggle.\nNo exploration. No doc-reading. No architecture discussions. No trade-offs.\u003c/p\u003e\n\u003cp\u003eIt’s just output.\u003c/p\u003e\n\u003cp\u003eAnd juniors, understandably, assume the output is correct.\nSo they learn \u003cem\u003epatterns\u003c/em\u003e, but not \u003cem\u003eprinciples\u003c/em\u003e.\nThey learn \u003cem\u003esolutions\u003c/em\u003e, but not \u003cem\u003eproblems\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eAsk them:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eWhy \u003ccode\u003eConfigureAwait(false)\u003c/code\u003e appears in library code?\u003c/li\u003e\n\u003cli\u003eWhy \u003ccode\u003eAddScoped\u003c/code\u003e and not \u003ccode\u003eAddSingleton\u003c/code\u003e for that service?\u003c/li\u003e\n\u003cli\u003eWhy \u003ccode\u003easync void\u003c/code\u003e is almost always wrong?\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThey won’t know.\nBecause Copilot doesn’t teach that. And they never had to care.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"copilot-rewards-passivity\"\u003e\u003ca href=\"/posts/copilot-turns-junior-devs-into-syntax-secretaries/#copilot-rewards-passivity\" title=\"Copilot Rewards Passivity\"\u003eCopilot Rewards Passivity\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eLet’s be brutally honest: \u003cstrong\u003eMost junior developers don’t need help writing more code.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThey need help:\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eDebugging\u003c/li\u003e\n\u003cli\u003eReasoning\u003c/li\u003e\n\u003cli\u003eModeling\u003c/li\u003e\n\u003cli\u003eTesting\u003c/li\u003e\n\u003cli\u003eReading logs\u003c/li\u003e\n\u003cli\u003eNaming things\u003c/li\u003e\n\u003cli\u003eUnderstanding the system outside their function\u003c/li\u003e\n\u003c/ul\u003e\n\u003cbr\u003e\nCopilot doesn’t train that. It rewards passivity.\n\u003cp\u003eIt turns proactive engineering into reactive prompting.\u003c/p\u003e\n\u003cp\u003eIt’s no longer: \u003cem\u003eHow should I design this?\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eIt’s: \u003cem\u003eHow do I phrase my prompt to get something close enough?\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThat mindset might produce a feature. But it will never produce a \u003cstrong\u003edeveloper\u003c/strong\u003e.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"youre-not-building-knowledge-youre-outsourcing-it\"\u003e\u003ca href=\"/posts/copilot-turns-junior-devs-into-syntax-secretaries/#youre-not-building-knowledge-youre-outsourcing-it\" title=\"You\u0026rsquo;re Not Building Knowledge. You\u0026rsquo;re Outsourcing It\"\u003eYou\u0026rsquo;re Not Building Knowledge. You\u0026rsquo;re Outsourcing It\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eHere’s the most perverse irony of all:\u003c/p\u003e\n\u003cp\u003eThe developer who uses Copilot the most ends up knowing the least.\u003c/p\u003e\n\u003cp\u003eThey’ve outsourced:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eSyntax\u003c/li\u003e\n\u003cli\u003eControl flow\u003c/li\u003e\n\u003cli\u003eAPI usage\u003c/li\u003e\n\u003cli\u003eDesign patterns\u003c/li\u003e\n\u003cli\u003eException handling\u003c/li\u003e\n\u003cli\u003eTest coverage\u003c/li\u003e\n\u003c/ul\u003e\n\u003cbr\u003e\nEverything but the typing.\n\u003cp\u003eWhich means they become \u003cstrong\u003edependent\u003c/strong\u003e on Copilot to do their job.\u003c/p\u003e\n\u003cp\u003eRemove the tool – and the facade collapses.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"it-looks-like-productivity--until-something-breaks\"\u003e\u003ca href=\"/posts/copilot-turns-junior-devs-into-syntax-secretaries/#it-looks-like-productivity--until-something-breaks\" title=\"It Looks Like Productivity – Until Something Breaks\"\u003eIt Looks Like Productivity – Until Something Breaks\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eA junior dev with Copilot might look fast.\nThey push code. Close tickets. Move features.\u003c/p\u003e\n\u003cp\u003eBut fast code isn’t good code.\nAnd when things break – and they will – they have no idea where to start.\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eLogs? Unfamiliar.\u003c/li\u003e\n\u003cli\u003eThreading? Scary.\u003c/li\u003e\n\u003cli\u003eSystem behavior under load? Never even considered.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eBecause they’ve never been forced to reason about any of it.\nCopilot took that friction away.\nAnd with it, all the growth.\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"whats-the-alternative\"\u003e\u003ca href=\"/posts/copilot-turns-junior-devs-into-syntax-secretaries/#whats-the-alternative\" title=\"What’s the Alternative?\"\u003eWhat’s the Alternative?\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eIt’s not about banning Copilot.\nIt’s about \u003cstrong\u003etiming\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eCopilot is fine when you:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eUnderstand the abstraction\u003c/li\u003e\n\u003cli\u003eKnow the system constraints\u003c/li\u003e\n\u003cli\u003eCan review the output critically\u003c/li\u003e\n\u003cli\u003eWould’ve written the same code manually anyway\u003c/li\u003e\n\u003c/ul\u003e\n\u003cbr\u003e\n\u003cp\u003eIn other words:\nUse it \u003cstrong\u003eafter\u003c/strong\u003e you’ve learned to think.\u003c/p\u003e\n\u003cp\u003eUntil then? \u003cstrong\u003eIf you can’t write it without Copilot, don’t write it with Copilot.\u003c/strong\u003e\u003c/p\u003e\n\n\n\n\n\u003ch2 id=\"final-word\"\u003e\u003ca href=\"/posts/copilot-turns-junior-devs-into-syntax-secretaries/#final-word\" title=\"Final Word\"\u003eFinal Word\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003eEarly-career developers need friction.\nThey need confusion.\nThey need mistakes that teach lessons Copilot will never explain.\u003c/p\u003e\n\u003cp\u003eBecause real engineering starts where Copilot stops.\u003c/p\u003e\n\u003cp\u003eSo to every junior developer out there:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003e\u003cstrong\u003eClose the AI tab.\u003c/strong\u003e\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eOpen the docs.\u003c/strong\u003e\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eStruggle a bit.\u003c/strong\u003e\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eBreak things.\u003c/strong\u003e\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eFix them.\u003c/strong\u003e\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eUnderstand why.\u003c/strong\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eAnd remember: \u003cstrong\u003eNo shortcuts. No sugarcoating. Just software done right.\u003c/strong\u003e\u003c/p\u003e","date_modified":"2026-05-05T17:06:04+02:00","date_published":"2025-05-13T17:30:00+02:00","id":"https://daily-devops.net/posts/copilot-turns-junior-devs-into-syntax-secretaries/","language":"en","summary":"Explore how GitHub Copilot and AI assistants impact junior developer growth, focusing on learning fundamentals beyond syntax completion and automation.","tags":["ai-code-assistant","bestpractices","dotnet","github","github-copilot","testing","visualstudio"],"title":"Copilot Turns Junior Devs Into Syntax Secretaries","url":"https://daily-devops.net/posts/copilot-turns-junior-devs-into-syntax-secretaries/"}],"language":"en","title":"GitHub Copilot AI Code Assistant on Daily DevOps \u0026 .NET","version":"https://jsonfeed.org/version/1.1"}