--- title: AI didn't solve programming, it just moved where the... description: A recurring critique of AI coding hype argues that AI tools have shifted, not eliminated, the hard parts of software engineering. Cory House points out that... canonical: https://daily.dev/posts/ai-didn-t-solve-programming-it-just-moved-where-the-mess-lives-mzcj9ni6m twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: AI didn't solve programming, it just moved where the mess lives | daily.dev og:description: A recurring critique of AI coding hype argues that AI tools have shifted, not eliminated, the hard parts of software engineering. Cory House points out that... og:url: https://daily.dev/posts/ai-didn-t-solve-programming-it-just-moved-where-the-mess-lives-mzcj9ni6m og:image: https://api.daily.dev/og/posts/mzcJ9Ni6M.png og:image:alt: AI didn't solve programming, it just moved where the mess lives og:image:width: 1200 og:image:height: 630 og:locale: 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. # AI didn't solve programming, it just moved where the mess lives **[Trends](https://daily.dev/sources/trends)** · 3 min read · 1 upvotes · 0 comments ## Summary A recurring critique of AI coding hype argues that AI tools have shifted, not eliminated, the hard parts of software engineering. Cory House points out that teams using AI still face the same old problems: bugs in production, unclear requirements, flaky tests, manual deploys, weak security, and poor accessibility. Uno Platform's line 'AI demos are the easy part' and Zach Lloyd's call for a 'true engineering mindset' toward coding agents echo the same theme: maintenance, edge cases, and production rigor remain unsolved by generation speed. The conclusion is that AI raises the ceiling of individual output but not the floor of what teams reliably ship, since that floor depends on process discipline and organizational culture rather than tooling. ## Content The productivity story everyone's selling right now goes like this: AI writes code faster, engineers ship more, everyone wins. The data is more complicated. Faros AI telemetry across 22,000 developers shows PRs grew 154% larger in 2025 and another 51% in 2026. Developers are juggling 67% more PR contexts daily. One team went from 80 merged PRs per week to 154 — but only after they stopped reviewing "low-risk" PRs entirely. That's not a productivity gain. That's a risk transfer. The verification gap is the real story. AI adoption sits at 40% for code generation, 38% for review assistance — and 6% for deployment decisions. The stuff that actually matters before code hits production is almost entirely untouched. A DX study of 500+ engineering orgs found AI saves developers 6 hours a week, but that time doesn't convert into more innovation. The "innovation ratio" — share of effort on new features vs. maintenance — barely moved. Median quarterly AI spend jumped from $1,500 to $44,000 in a year. Feature work stayed flat at 58%. Boris Cherny, who built Claude Code, frames it with a 1996 HBR study on computerization: companies that just bolted computers onto existing paper processes saw no gains. The ones that redesigned around computers did. Most teams right now are doing the former — using AI to write code faster while leaving the review queue, CI pipeline, approval gates, and unclear requirements exactly where they were. The failure modes are specific and documented. One developer's 30-day AI-only experiment produced a Stripe webhook that acknowledged payment events before persisting them — meaning a DB blip could revoke a paying customer's access. It looked correct. It passed manual review. The agent had no way to know it was wrong because nobody wrote the behavioral spec that said "don't do that." Some teams are adapting. One legal software company tripled output per developer over 18 months — but the gains came from eliminating handoffs between product, QA, security, and deployment, not from generating code faster. Governance gates went in *before* scaling AI adoption. Another team got cost-per-PR down to $30 by benchmarking models on their own data and building proper observability. The pattern in every success story is the same: treat verification as a first-class engineering problem, not an afterthought. AGENTS.md files, behavioral specs written before implementation, mandatory evidence trails, separate author and reviewer agent roles with human merge approval. The teams skipping this step are accumulating debt that doesn't show up in velocity metrics until something breaks in production. The bottleneck moved. Most teams haven't noticed yet. ## Questions this post answers ### Why did PR sizes grow so much after teams adopted AI coding tools, according to any recent industry data? Telemetry from Faros AI across 22,000 developers found PRs grew 154% larger in one year and another 51% the following year, with developers juggling 67% more PR contexts daily. One team's merged PR count rose from 80 to 154 per week only after they stopped reviewing PRs they classified as low-risk, shifting risk rather than genuinely gaining productivity. _Teams weighing AI-driven velocity against review debt can track this tension through daily.dev._ ### What is the AI adoption rate for deployment decisions compared to code generation? AI adoption for deployment decisions sits at only 6%, compared to 40% for code generation and 38% for review assistance. This gap means the step that actually gates what reaches production is largely untouched by AI, even as code creation and review get heavy AI involvement, leaving verification as the weak link in the pipeline. _Engineers rethinking where AI belongs in their release pipeline can follow this debate on daily.dev._ ### Does using AI coding assistants actually increase the share of time spent on new feature work versus maintenance? No. A DX study of over 500 engineering organizations found AI saves developers about 6 hours a week, but the innovation ratio, the share of effort spent on new features versus maintenance, barely moved, staying flat around 58% for feature work even as median quarterly AI spend jumped from $1,500 to $44,000 in a year. _Leaders justifying AI tooling budgets against real feature output gains can dig deeper on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 1 discussion and 1 comments across x (as of 2026-09-13). **TL;DR:** There's essentially no substantive discussion available yet; the lone reply is a low-signal expression of agreement rather than a developed take. **Sentiment:** 0% positive · 100% mixed · 0% skeptical **By community** - x (mixed): The single reply is a brief expression of agreement with no elaborated argument. **Source threads** - [x](https://x.com/mattpocockuk/status/2097692962553549064) · 0 points · 1 comments ## Similar posts on daily.dev - [Why AI Code Reviews Are Slowing Down PRs](https://daily.dev/posts/why-ai-code-reviews-are-slowing-down-prs-yubjqxjqr) · Aviator Blog · 0 upvotes · 1 comments - [AI Code Review Bottleneck 2026: Why PRs Are Stuck in Review](https://daily.dev/posts/ai-code-review-bottleneck-2026-why-prs-are-stuck-in-review-o7qcycgq0) · Alex CloudStar · 2 upvotes · 1 comments --- Tags: [#testing](https://daily.dev/tags/testing), [#ai-agents](https://daily.dev/tags/ai-agents), [#ai-coding](https://daily.dev/tags/ai-coding), [#technical-debt](https://daily.dev/tags/technical-debt) [View this post on daily.dev](https://daily.dev/posts/ai-didn-t-solve-programming-it-just-moved-where-the-mess-lives-mzcj9ni6m) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"AI didn't solve programming, it just moved where the mess lives","url":"https://daily.dev/posts/ai-didn-t-solve-programming-it-just-moved-where-the-mess-lives-mzcj9ni6m","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ai-didn-t-solve-programming-it-just-moved-where-the-mess-lives-mzcj9ni6m"},"datePublished":"2026-09-03T17:57:09.552Z","dateModified":"2026-09-13T19:55:49.070Z","description":"A recurring critique of AI coding hype argues that AI tools have shifted, not eliminated, the hard parts of software engineering. Cory House points out that...","isAccessibleForFree":true,"articleSection":"Trends","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Trends","logo":"https://media.daily.dev/image/upload/s--ZfSp3asX--/f_auto,q_auto/v1780996004/logos/trends?_a=BAMAMiWQ0","url":"https://daily.dev/sources/trends"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/ai-didn-t-solve-programming-it-just-moved-where-the-mess-lives-mzcj9ni6m","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"testing,ai-agents,ai-coding,technical-debt","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Trends","item":"https://daily.dev/sources/trends"},{"@type":"ListItem","position":3,"name":"AI didn't solve programming, it just moved where the mess lives"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/ai-didn-t-solve-programming-it-just-moved-where-the-mess-lives-mzcj9ni6m#faq","mainEntity":[{"@type":"Question","name":"Why did PR sizes grow so much after teams adopted AI coding tools, according to any recent industry data?","acceptedAnswer":{"@type":"Answer","text":"Telemetry from Faros AI across 22,000 developers found PRs grew 154% larger in one year and another 51% the following year, with developers juggling 67% more PR contexts daily. One team's merged PR count rose from 80 to 154 per week only after they stopped reviewing PRs they classified as low-risk, shifting risk rather than genuinely gaining productivity. Teams weighing AI-driven velocity against review debt can track this tension through daily.dev."}},{"@type":"Question","name":"What is the AI adoption rate for deployment decisions compared to code generation?","acceptedAnswer":{"@type":"Answer","text":"AI adoption for deployment decisions sits at only 6%, compared to 40% for code generation and 38% for review assistance. This gap means the step that actually gates what reaches production is largely untouched by AI, even as code creation and review get heavy AI involvement, leaving verification as the weak link in the pipeline. Engineers rethinking where AI belongs in their release pipeline can follow this debate on daily.dev."}},{"@type":"Question","name":"Does using AI coding assistants actually increase the share of time spent on new feature work versus maintenance?","acceptedAnswer":{"@type":"Answer","text":"No. A DX study of over 500 engineering organizations found AI saves developers about 6 hours a week, but the innovation ratio, the share of effort spent on new features versus maintenance, barely moved, staying flat around 58% for feature work even as median quarterly AI spend jumped from $1,500 to $44,000 in a year. Leaders justifying AI tooling budgets against real feature output gains can dig deeper on daily.dev."}}]} ```