--- title: AI-Native Engineering Culture from a CTO | daily.dev description: A CTO at a startup shares how his five-person team has moved to fully AI-native engineering where no one writes or reads code manually. The conversation covers... canonical: https://daily.dev/posts/ai-native-engineering-culture-from-a-cto-dd6qg7q4h twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: AI-Native Engineering Culture from a CTO | daily.dev og:description: A CTO at a startup shares how his five-person team has moved to fully AI-native engineering where no one writes or reads code manually. The conversation covers... og:url: https://daily.dev/posts/ai-native-engineering-culture-from-a-cto-dd6qg7q4h og:image: https://api.daily.dev/og/posts/dD6qg7Q4H.png og:image:alt: AI-Native Engineering Culture from a CTO 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-Native Engineering Culture from a CTO **[Bret Fisher](https://daily.dev/sources/bretfisher)** · 64 min read · 4 upvotes · 0 comments ## Summary A CTO at a startup shares how his five-person team has moved to fully AI-native engineering where no one writes or reads code manually. The conversation covers their agentic development workflow using Claude Code, how they spin up full production copies for every PR, their testing infrastructure (unit, E2E, Playwright), risk assessment philosophy for shipping fast, and how engineering speed has outpaced sales, design, and product. Key themes include using judgment over tickets, letting junior engineers fail and learn, token cost management, harness optimization, and why this approach only works for web/mobile apps — not healthcare or embedded systems. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What did a Stanford paper find about how much the agent harness affects LLM coding performance? A Stanford paper found that a well-designed harness can improve model performance by up to 6x. This is cited as the reason to prioritize harness engineering over model selection once models reach a baseline capability level, since even the best available harness, such as Claude Code, still leaves significant room for optimization. _Teams weighing harness design against model upgrades can track agentic engineering practices on daily.dev._ ### How can you reduce LLM token costs when running autonomous coding agents in production? One approach is restricting the model to expressing itself only through writing Python code rather than direct tool calls, packaging tool access into an agent SDK. This lets agents reuse previously written code instead of regenerating it each time, and allows small models to handle Python generation reliably, cutting both token usage and cost as agents accumulate reusable code over time. _Anyone budgeting for agent-heavy pipelines can follow cost-optimization approaches like this on daily.dev._ ### How risky is it to let AI agents commit and deploy code to production without any human review? It can work if failure states are recoverable quickly and testing infrastructure is exhaustive: a described setup used a full production copy per pull request (including the database), 15-16 automated workflows, unit tests, end-to-end tests, Playwright browser tests, and linters. Even so, two catastrophic mistakes forced a two-week system regression, illustrating that errors in this style of engineering tend to be severe rather than minor. _Engineers weighing autonomous deployment risk versus recoverability can dig into real-world accounts on daily.dev._ ## Similar posts on daily.dev - [When the cost of code approaches zero, what does engineering leadership look like?](https://daily.dev/posts/when-the-cost-of-code-approaches-zero-what-does-engineering-leadership-look-like--fcoig8gfg) · Stack Overflow Blog · 3 upvotes · 0 comments - [From MCP and Vibe Coding to Harness Engineering: How Did AI Native Engineering Evolve in One Year](https://daily.dev/posts/from-mcp-and-vibe-coding-to-harness-engineering-how-did-ai-native-engineering-evolve-in-one-year-7maibegpy) · InfoQ · 0 upvotes · 0 comments - [Revised rules of engineering leadership.](https://daily.dev/posts/revised-rules-of-engineering-leadership--mcfnko3s5) · Irrational Exuberance · 5 upvotes · 0 comments - [My First Two Weeks as an Agentic Engineer](https://daily.dev/posts/my-first-two-weeks-as-an-agentic-engineer-ibtc3m6h6) · The Miners · 0 upvotes · 0 comments - [Coding Is No Longer the Constraint: Scaling Developer Experience to Teams and Agents at Spotify](https://daily.dev/posts/coding-is-no-longer-the-constraint-scaling-developer-experience-to-teams-and-agents-at-spotify-fas1vty9a) · Spotify Labs · 1 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#cicd](https://daily.dev/tags/cicd), [#ai-coding](https://daily.dev/tags/ai-coding), [#claude-code](https://daily.dev/tags/claude-code) [View this post on daily.dev](https://daily.dev/posts/ai-native-engineering-culture-from-a-cto-dd6qg7q4h) ```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-Native Engineering Culture from a CTO","url":"https://daily.dev/posts/ai-native-engineering-culture-from-a-cto-dd6qg7q4h","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ai-native-engineering-culture-from-a-cto-dd6qg7q4h"},"datePublished":"2026-08-01T04:43:18.018Z","dateModified":"2026-09-13T19:43:19.512Z","description":"A CTO at a startup shares how his five-person team has moved to fully AI-native engineering where no one writes or reads code manually. The conversation covers...","image":"https://i.ytimg.com/vi/m1Wz6dbzRrc/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/m1Wz6dbzRrc/sddefault.jpg","isAccessibleForFree":true,"articleSection":"Bret Fisher","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":"Bret Fisher","logo":"https://media.daily.dev/image/upload/s--8RH_C-vW--/f_auto,q_auto/v1780213225/logos/bretfisher?_a=BAMAMiWQ0","url":"https://daily.dev/sources/bretfisher"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/ai-native-engineering-culture-from-a-cto-dd6qg7q4h","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":4},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,cicd,ai-coding,claude-code","timeRequired":"PT64M","video":{"@type":"VideoObject","name":"AI-Native Engineering Culture from a CTO","description":"A CTO at a startup shares how his five-person team has moved to fully AI-native engineering where no one writes or reads code manually. The conversation covers...","thumbnailUrl":"https://i.ytimg.com/vi/m1Wz6dbzRrc/sddefault.jpg","uploadDate":"2026-08-01T04:43:18.018Z","duration":"PT64M","url":"https://api.daily.dev/r/dD6qg7Q4H","embedUrl":"https://www.youtube.com/embed/m1Wz6dbzRrc"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Bret Fisher","item":"https://daily.dev/sources/bretfisher"},{"@type":"ListItem","position":3,"name":"AI-Native Engineering Culture from a CTO"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/ai-native-engineering-culture-from-a-cto-dd6qg7q4h#faq","mainEntity":[{"@type":"Question","name":"What did a Stanford paper find about how much the agent harness affects LLM coding performance?","acceptedAnswer":{"@type":"Answer","text":"A Stanford paper found that a well-designed harness can improve model performance by up to 6x. This is cited as the reason to prioritize harness engineering over model selection once models reach a baseline capability level, since even the best available harness, such as Claude Code, still leaves significant room for optimization. Teams weighing harness design against model upgrades can track agentic engineering practices on daily.dev."}},{"@type":"Question","name":"How can you reduce LLM token costs when running autonomous coding agents in production?","acceptedAnswer":{"@type":"Answer","text":"One approach is restricting the model to expressing itself only through writing Python code rather than direct tool calls, packaging tool access into an agent SDK. This lets agents reuse previously written code instead of regenerating it each time, and allows small models to handle Python generation reliably, cutting both token usage and cost as agents accumulate reusable code over time. Anyone budgeting for agent-heavy pipelines can follow cost-optimization approaches like this on daily.dev."}},{"@type":"Question","name":"How risky is it to let AI agents commit and deploy code to production without any human review?","acceptedAnswer":{"@type":"Answer","text":"It can work if failure states are recoverable quickly and testing infrastructure is exhaustive: a described setup used a full production copy per pull request (including the database), 15-16 automated workflows, unit tests, end-to-end tests, Playwright browser tests, and linters. Even so, two catastrophic mistakes forced a two-week system regression, illustrating that errors in this style of engineering tend to be severe rather than minor. Engineers weighing autonomous deployment risk versus recoverability can dig into real-world accounts on daily.dev."}}]} ```