--- title: "Conceptual integrity and counting lines of code" url: https://daily.dev/posts/conceptual-integrity-and-counting-lines-of-code-aiaamupbu source_url: https://simonwillison.net/2026/Aug/19/conceptual-integrity-and-counting-lines-of-code type: article source: "Simon Willison" published: 2026-08-19T22:57:46.557Z updated: 2026-08-20T11:29:42.811Z tags: ["ai", "ai-agents"] reading_time: 4 upvotes: 0 comments: 0 language: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Conceptual integrity and counting lines of code **[Simon Willison](https://daily.dev/sources/simonwillison)** · 4 min read · 0 upvotes · 0 comments ## Summary Reflecting on a Talking Postgres podcast conversation with Claire Giordano, Simon Willison argues that lines of code can still be a meaningful productivity metric with coding agents, since human cognitive capacity to review and maintain code remains the real bottleneck rather than raw output speed. He also discusses how coding agents threaten 'conceptual integrity' (a term from The Mythical Man-Month), comparing sprawling AI-assisted codebases to the Winchester Mystery House, where cheap feature additions erode overall design coherence and discipline. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Is measuring developer productivity in lines of code still a valid approach when using AI coding agents? It can be, according to an argument that a hard limit exists on human-produced code quality: a great engineer might produce around 200 debugged, production-ready lines per day without agents, often just 50-60. If agents raise sustainable output to around a thousand lines of equally maintainable, tested code, that is a meaningful productivity gain, though the real constraint becomes an engineer's cognitive capacity to stay on top of that much code. _Developers debating how to measure AI-assisted output can follow this productivity debate on daily.dev._ ### What is conceptual integrity in software design and why do AI coding agents threaten it? Conceptual integrity, a term from The Mythical Man-Month, describes well-designed software where everything fits together coherently with no surprises. Coding agents threaten it because prompting a new feature into existence takes minutes instead of the week it once required, removing the natural discipline that time constraints used to impose, so software accumulates disjointed additions much like the ever-expanding Winchester Mystery House. _Teams weighing agent-driven feature velocity against long-term maintainability can track this discussion on daily.dev._ ## Similar posts on daily.dev - [claude code is not making your product better](https://daily.dev/posts/claude-code-is-not-making-your-product-better-gmfddhnyw) · Lobsters · 1 upvotes · 0 comments - [Why I Refuse to Scale My Coding Agents](https://daily.dev/posts/why-i-refuse-to-scale-my-coding-agents-nledrbjiu) · Medium · 0 upvotes · 0 comments - [The bottleneck isn't writing code anymore. It's understanding it.](https://daily.dev/posts/the-bottleneck-isn-t-writing-code-anymore-it-s-understanding-it--oqefvqhf8) · Three Dots Labs · 152 upvotes · 31 comments --- Tags: [#ai](https://daily.dev/tags/ai), [#ai-agents](https://daily.dev/tags/ai-agents) [View this post on daily.dev](https://daily.dev/posts/conceptual-integrity-and-counting-lines-of-code-aiaamupbu)