--- title: Open Source Is Rocket Fuel for AI Agents | daily.dev description: A developer at Marmelab ran a controlled experiment building the same admin application three ways — from scratch, with shadcn/ui, and with react-admin — using... canonical: https://daily.dev/posts/open-source-is-rocket-fuel-for-ai-agents-hsjzrkgef twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Open Source Is Rocket Fuel for AI Agents | daily.dev og:description: A developer at Marmelab ran a controlled experiment building the same admin application three ways — from scratch, with shadcn/ui, and with react-admin — using... og:url: https://daily.dev/posts/open-source-is-rocket-fuel-for-ai-agents-hsjzrkgef og:image: https://api.daily.dev/og/posts/hSjzRKGef.png og:image:alt: Open Source Is Rocket Fuel for AI Agents 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. # Open Source Is Rocket Fuel for AI Agents **[marmelab](https://daily.dev/sources/Marmelab)** · 10 min read · 6 upvotes · 0 comments ## Summary A developer at Marmelab ran a controlled experiment building the same admin application three ways — from scratch, with shadcn/ui, and with react-admin — using Claude Code agents. Results showed react-admin used ~25% fewer tokens than from-scratch and ~45% fewer than shadcn/ui, translating to direct cost savings. The key insight: well-designed libraries with accessible documentation reduce token consumption because they encapsulate complexity the model doesn't need to rewrite. The post also raises a broader concern: AI agents are quietly undermining open-source sustainability by bypassing documentation sites, forums, and sponsorship flows — as illustrated by Tailwind CSS losing ~80% of revenue despite tripled usage. The author argues that using established open-source libraries is both economically rational for AI-assisted development and essential for preserving the ecosystem AI itself depends on. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Does using a UI library like react-admin reduce the number of tokens an AI coding agent needs to build an app compared to building from scratch? Yes, in a controlled benchmark building the same admin backoffice, react-admin used about 811,000-818,000 tokens with Claude Sonnet and Opus 4.8, versus roughly 1,075,000-1,490,000 tokens for from-scratch or shadcn/ui builds. React-admin's built-in routing, forms, and data management meant the model only wrote business logic and configuration, cutting both token usage and session time versus manually implemented alternatives. _Comparing AI coding agent costs across frameworks is easier when tracking benchmarks like this via daily.dev._ ### Why did using shadcn/ui with an AI coding agent use more tokens than building a UI from scratch? Shadcn/ui generates component code directly inside the project rather than as a node_modules dependency, so Claude had to read and understand all that generated code, then still implement routing and form logic by hand just like a from-scratch build. In the benchmark this produced the highest token counts of the three approaches tested (about 1.2-1.5 million tokens with Claude Sonnet and Opus 4.8). _Weighing UI library trade-offs for AI-assisted development is simpler when following comparisons like this on daily.dev._ ### What is slopsquatting and how does it relate to AI coding assistants hallucinating package names? Slopsquatting is when attackers register package names on npm or PyPI that AI systems are known to hallucinate, then wait for developers whose AI agent suggests that fake package to install it. It exploits the predictability of AI hallucinations the way typosquatting exploits human typing errors, and recent models still hallucinate package names, making unspecified library choices a supply-chain risk. _Staying alert to AI-driven supply chain risks like slopsquatting is easier when tracking security news on daily.dev._ ## Similar posts on daily.dev - [Vibe coding may be hazardous to open source](https://daily.dev/posts/vibe-coding-may-be-hazardous-to-open-source-adodkq3hd) · The Register · 0 upvotes · 0 comments - [How open source projects need to adapt to the AI coding era](https://daily.dev/posts/how-open-source-projects-need-to-adapt-to-the-ai-coding-era-bjikrjdag) · All Things Open · 0 upvotes · 0 comments - [AI is a business model stress test](https://daily.dev/posts/ai-is-a-business-model-stress-test-geg8xcry5) · Hacker News · 3 upvotes · 0 comments - [Many tokens make all bugs shallow & open source’s new maintainers \| Chainguard's Dan Lorenc](https://daily.dev/posts/many-tokens-make-all-bugs-shallow-open-source-s-new-maintainers-chainguard-s-dan-lorenc-3bcpdnraf) · Dev Interrupted · 0 upvotes · 0 comments --- Tags: [#open-source](https://daily.dev/tags/open-source), [#ai-agents](https://daily.dev/tags/ai-agents), [#claude-code](https://daily.dev/tags/claude-code), [#context-engineering](https://daily.dev/tags/context-engineering) [View this post on daily.dev](https://daily.dev/posts/open-source-is-rocket-fuel-for-ai-agents-hsjzrkgef) ```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":"Open Source Is Rocket Fuel for AI Agents","url":"https://daily.dev/posts/open-source-is-rocket-fuel-for-ai-agents-hsjzrkgef","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/open-source-is-rocket-fuel-for-ai-agents-hsjzrkgef"},"datePublished":"2026-07-13T09:40:00.768Z","dateModified":"2026-09-14T08:52:16.411Z","description":"A developer at Marmelab ran a controlled experiment building the same admin application three ways — from scratch, with shadcn/ui, and with react-admin — using...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/52192450af0056129b723ee91b6e8dca?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/52192450af0056129b723ee91b6e8dca?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"marmelab","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":"marmelab","logo":"https://media.daily.dev/image/upload/s--AeMXoaI9--/f_auto/v1716621144/logos/Marmelab","url":"https://daily.dev/sources/Marmelab"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/open-source-is-rocket-fuel-for-ai-agents-hsjzrkgef","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":6},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"open-source,ai-agents,claude-code,context-engineering","timeRequired":"PT10M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"marmelab","item":"https://daily.dev/sources/Marmelab"},{"@type":"ListItem","position":3,"name":"Open Source Is Rocket Fuel for AI Agents"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/open-source-is-rocket-fuel-for-ai-agents-hsjzrkgef#faq","mainEntity":[{"@type":"Question","name":"Does using a UI library like react-admin reduce the number of tokens an AI coding agent needs to build an app compared to building from scratch?","acceptedAnswer":{"@type":"Answer","text":"Yes, in a controlled benchmark building the same admin backoffice, react-admin used about 811,000-818,000 tokens with Claude Sonnet and Opus 4.8, versus roughly 1,075,000-1,490,000 tokens for from-scratch or shadcn/ui builds. React-admin's built-in routing, forms, and data management meant the model only wrote business logic and configuration, cutting both token usage and session time versus manually implemented alternatives. Comparing AI coding agent costs across frameworks is easier when tracking benchmarks like this via daily.dev."}},{"@type":"Question","name":"Why did using shadcn/ui with an AI coding agent use more tokens than building a UI from scratch?","acceptedAnswer":{"@type":"Answer","text":"Shadcn/ui generates component code directly inside the project rather than as a node_modules dependency, so Claude had to read and understand all that generated code, then still implement routing and form logic by hand just like a from-scratch build. In the benchmark this produced the highest token counts of the three approaches tested (about 1.2-1.5 million tokens with Claude Sonnet and Opus 4.8). Weighing UI library trade-offs for AI-assisted development is simpler when following comparisons like this on daily.dev."}},{"@type":"Question","name":"What is slopsquatting and how does it relate to AI coding assistants hallucinating package names?","acceptedAnswer":{"@type":"Answer","text":"Slopsquatting is when attackers register package names on npm or PyPI that AI systems are known to hallucinate, then wait for developers whose AI agent suggests that fake package to install it. It exploits the predictability of AI hallucinations the way typosquatting exploits human typing errors, and recent models still hallucinate package names, making unspecified library choices a supply-chain risk. Staying alert to AI-driven supply chain risks like slopsquatting is easier when tracking security news on daily.dev."}}]} ```