--- title: AI agents pass tests. Can they write idiomatic Laravel? description: AI coding agents using Laravel Boost can now pass all 17 Laravel eval tasks at or near 100% accuracy, effectively saturating the existing benchmark. With... canonical: https://daily.dev/posts/ai-agents-pass-tests-can-they-write-idiomatic-laravel--yvqlevrwo twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: AI agents pass tests. Can they write idiomatic Laravel? | daily.dev og:description: AI coding agents using Laravel Boost can now pass all 17 Laravel eval tasks at or near 100% accuracy, effectively saturating the existing benchmark. With... og:url: https://daily.dev/posts/ai-agents-pass-tests-can-they-write-idiomatic-laravel--yvqlevrwo og:image: https://api.daily.dev/og/posts/yVQleVRWO.png og:image:alt: AI agents pass tests. Can they write idiomatic Laravel? 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 agents pass tests. Can they write idiomatic Laravel? **[Laravel](https://daily.dev/sources/laravel)** · 8 min read · 0 upvotes · 0 comments ## Summary AI coding agents using Laravel Boost can now pass all 17 Laravel eval tasks at or near 100% accuracy, effectively saturating the existing benchmark. With correctness becoming table stakes, the focus is shifting to two new metrics: correct code per token (efficiency) and idiomatic Laravel quality. The post outlines how Boost will evolve to measure whether agents write code that not only works but follows Laravel conventions — using form requests, eager loading, Route::resource(), and other best practices — while minimizing the context tokens needed to get there. A new scoring layer using LLM-as-judge and a 19-point best-practices rubric is being developed alongside token/cost reporting as headline metrics. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Can AI coding agents like GPT-5.6, Claude Fable 5, or Gemini 3.x write correct Laravel code that passes tests? Yes, frontier models now pass at or near 100% of Laravel Boost's 17-eval benchmark suite when supplied with Laravel Boost's framework context, up from clearing 16 of 17 evals with 99.4% test accuracy in an earlier round of testing. This means basic framework correctness in Laravel is now considered a solved problem for leading AI models. _Developers weighing which AI coding agent to trust with a Laravel codebase can track these shifting benchmarks on daily.dev._ ### Why did OpenAI stop reporting SWE-bench Verified scores? An audit found that a large share of the SWE-bench Verified problems had flawed test cases, undermining confidence in the benchmark's results. Separately, Cursor demonstrated that many models solved tasks by finding known answers in GitHub or .git history rather than genuinely working out fixes, and scores dropped sharply once that history was sealed off. _Anyone benchmarking coding agents can follow how flawed test suites and reward hacking get exposed via daily.dev._ ### What makes Laravel code idiomatic rather than merely correct according to Laravel Boost's best-practices skill? Idiomatic Laravel code follows 19 concrete conventions defined in Boost's best-practices skill, such as using Cache::remember() instead of hand-rolled caching, form requests instead of inline validation, with() for eager loading, and Route::resource() for resourceful controllers. Its first rule, 'Consistency First,' says to match the patterns already used in sibling files before reaching for a supposedly better approach. _Teams reviewing AI-generated Laravel code for style consistency can find these conventions surfaced on daily.dev._ ## Similar posts on daily.dev - [Which AI Model Is Best for Laravel?](https://daily.dev/posts/which-ai-model-is-best-for-laravel--vpvv9iypi) · Laravel · 14 upvotes · 0 comments - [Evals in Laravel: How to Prove Your AI Output Is Actually Good](https://daily.dev/posts/evals-in-laravel-how-to-prove-your-ai-output-is-actually-good-v8xwwmodq) · FREEK.DEV · 1 upvotes · 0 comments - [Laravel raised money and now injects ads directly into your agent](https://daily.dev/posts/laravel-raised-money-and-now-injects-ads-directly-into-your-agent-itsfvv4ge) · Hacker News · 0 upvotes · 2 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#php](https://daily.dev/tags/php), [#laravel](https://daily.dev/tags/laravel) [View this post on daily.dev](https://daily.dev/posts/ai-agents-pass-tests-can-they-write-idiomatic-laravel--yvqlevrwo) ```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 agents pass tests. Can they write idiomatic Laravel?","url":"https://daily.dev/posts/ai-agents-pass-tests-can-they-write-idiomatic-laravel--yvqlevrwo","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ai-agents-pass-tests-can-they-write-idiomatic-laravel--yvqlevrwo"},"datePublished":"2026-07-24T11:43:44.383Z","dateModified":"2026-09-13T21:44:14.193Z","description":"AI coding agents using Laravel Boost can now pass all 17 Laravel eval tasks at or near 100% accuracy, effectively saturating the existing benchmark. With...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/16a5f03e5ccac950b05a10292ada476f?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/16a5f03e5ccac950b05a10292ada476f?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Laravel","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":"Laravel","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/1f77002e78e1486ab699e1b8e03d7fd2","url":"https://daily.dev/sources/laravel"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/ai-agents-pass-tests-can-they-write-idiomatic-laravel--yvqlevrwo","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,php,laravel","timeRequired":"PT8M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Laravel","item":"https://daily.dev/sources/laravel"},{"@type":"ListItem","position":3,"name":"AI agents pass tests. Can they write idiomatic Laravel?"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/ai-agents-pass-tests-can-they-write-idiomatic-laravel--yvqlevrwo#faq","mainEntity":[{"@type":"Question","name":"Can AI coding agents like GPT-5.6, Claude Fable 5, or Gemini 3.x write correct Laravel code that passes tests?","acceptedAnswer":{"@type":"Answer","text":"Yes, frontier models now pass at or near 100% of Laravel Boost's 17-eval benchmark suite when supplied with Laravel Boost's framework context, up from clearing 16 of 17 evals with 99.4% test accuracy in an earlier round of testing. This means basic framework correctness in Laravel is now considered a solved problem for leading AI models. Developers weighing which AI coding agent to trust with a Laravel codebase can track these shifting benchmarks on daily.dev."}},{"@type":"Question","name":"Why did OpenAI stop reporting SWE-bench Verified scores?","acceptedAnswer":{"@type":"Answer","text":"An audit found that a large share of the SWE-bench Verified problems had flawed test cases, undermining confidence in the benchmark's results. Separately, Cursor demonstrated that many models solved tasks by finding known answers in GitHub or .git history rather than genuinely working out fixes, and scores dropped sharply once that history was sealed off. Anyone benchmarking coding agents can follow how flawed test suites and reward hacking get exposed via daily.dev."}},{"@type":"Question","name":"What makes Laravel code idiomatic rather than merely correct according to Laravel Boost's best-practices skill?","acceptedAnswer":{"@type":"Answer","text":"Idiomatic Laravel code follows 19 concrete conventions defined in Boost's best-practices skill, such as using Cache::remember() instead of hand-rolled caching, form requests instead of inline validation, with() for eager loading, and Route::resource() for resourceful controllers. Its first rule, 'Consistency First,' says to match the patterns already used in sibling files before reaching for a supposedly better approach. Teams reviewing AI-generated Laravel code for style consistency can find these conventions surfaced on daily.dev."}}]} ```