--- title: Prototyping as Leadership: How a CTO Ships with AI... description: The CTO and co-founder of The Browser Company describes how autonomous coding agents let him ship 2-10 PRs a week despite a packed manager schedule and a... canonical: https://daily.dev/posts/prototyping-as-leadership-how-a-cto-ships-with-ai-agents-hursh-agrawal-the-browser-company-yzcwxhrwj twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Prototyping as Leadership: How a CTO Ships with AI Agents — Hursh Agrawal, The Browser Company | daily.dev og:description: The CTO and co-founder of The Browser Company describes how autonomous coding agents let him ship 2-10 PRs a week despite a packed manager schedule and a... og:url: https://daily.dev/posts/prototyping-as-leadership-how-a-cto-ships-with-ai-agents-hursh-agrawal-the-browser-company-yzcwxhrwj og:image: https://api.daily.dev/og/posts/YzcwxhRWJ.png og:image:alt: Prototyping as Leadership: How a CTO Ships with AI Agents — Hursh Agrawal, The Browser Company 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. # Prototyping as Leadership: How a CTO Ships with AI Agents — Hursh Agrawal, The Browser Company **[AI Engineer](https://daily.dev/sources/aidotengineer)** · 18 min read · 1 upvotes · 0 comments ## Summary The CTO and co-founder of The Browser Company describes how autonomous coding agents let him ship 2-10 PRs a week despite a packed manager schedule and a toddler at home. He runs an overnight loop: gathering business context via a Slack/Notion-connected co-work agent in the evening, kicking off long autonomous runs with Claude Code or similar tools overnight (building features, hill-climbing AI evals, or even training custom ML models like a BERT PII classifier), and reviewing results each morning. He argues leaders are well positioned to prototype because they hold the most business context, and that building lets them personally understand each new frontier model's capabilities every three months rather than relying on secondhand opinions. He stresses supporting infrastructure (AI code reviewers, agents.md hygiene, feature flags, a prototype branch) and discipline (writing tests first, small readable PRs, reading code before adding reviewers) to avoid the production incidents he admits he has caused. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How can I set up an overnight workflow with Claude Code so it works autonomously for hours without me babysitting it? Give the agent maximum business context upfront rather than breaking the task into micro-prompts, since it will run unsupervised for 4-8 hours. Have it write tests first, verify the end-to-end flow with computer use, split work into reviewer-friendly PRs, run an AI code review skill on itself and fix the findings, keep CI green, and leave a morning report on trade-offs made. _Developers building overnight agent workflows compare real setups and prompting patterns on daily.dev._ ### How do you build an eval set to hill climb an AI feature when you only have a small amount of user feedback? Add a lightweight feedback button to the feature, collect as few as four to thirty feedback entries saved as JSON dumps with the system prompt and inputs, then have a coding agent turn those into a local eval set (SQLite or Markdown works) and build a harness that runs the call against the evals repeatedly, explicitly instructing it not to overfit on the small sample. _Teams tuning AI feature quality can trade eval and hill-climbing approaches on daily.dev._ ### Can modern coding agents actually train a custom ML model like a PII classifier overnight without much ML expertise? Yes, a modernBERT PII classifier was trained overnight after Opus and Haiku proved too expensive and slow with insufficient precision-recall, by giving a coding agent business context, training data, open AI and Anthropic API keys, and sandboxed AWS access to pick a model class, train two candidate models, evaluate, and deprovision resources, returning trained models with a morning report. _Engineers exploring agent-driven model training can follow similar experiments and results on daily.dev._ ## Similar posts on daily.dev - [Revised rules of engineering leadership.](https://daily.dev/posts/revised-rules-of-engineering-leadership--mcfnko3s5) · Irrational Exuberance · 5 upvotes · 0 comments - [Why I code as a CTO](https://daily.dev/posts/why-i-code-as-a-cto-zsgdnxclg) · Hacker News · 42 upvotes · 5 comments - [How The Browser Company moves so fast: 3 habits you can borrow](https://daily.dev/posts/how-the-browser-company-moves-so-fast-3-habits-you-can-borrow-a6clqmv0m) · Atlassian · 0 upvotes · 0 comments - [When code is abundant](https://daily.dev/posts/when-code-is-abundant-gezbqtizx) · GitLab · 0 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#leadership](https://daily.dev/tags/leadership), [#claude-code](https://daily.dev/tags/claude-code), [#prompt-engineering](https://daily.dev/tags/prompt-engineering) [View this post on daily.dev](https://daily.dev/posts/prototyping-as-leadership-how-a-cto-ships-with-ai-agents-hursh-agrawal-the-browser-company-yzcwxhrwj) ```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":"Prototyping as Leadership: How a CTO Ships with AI Agents — Hursh Agrawal, The Browser Company","url":"https://daily.dev/posts/prototyping-as-leadership-how-a-cto-ships-with-ai-agents-hursh-agrawal-the-browser-company-yzcwxhrwj","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/prototyping-as-leadership-how-a-cto-ships-with-ai-agents-hursh-agrawal-the-browser-company-yzcwxhrwj"},"datePublished":"2026-08-20T14:28:37.671Z","dateModified":"2026-09-14T09:13:32.562Z","description":"The CTO and co-founder of The Browser Company describes how autonomous coding agents let him ship 2-10 PRs a week despite a packed manager schedule and a...","image":"https://i.ytimg.com/vi/bdHaOXZOhcM/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/bdHaOXZOhcM/sddefault.jpg","isAccessibleForFree":true,"articleSection":"AI Engineer","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":"AI Engineer","logo":"https://media.daily.dev/image/upload/s--u5PucxNT--/f_auto/v1724338940/logos/aidotengineer","url":"https://daily.dev/sources/aidotengineer"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/prototyping-as-leadership-how-a-cto-ships-with-ai-agents-hursh-agrawal-the-browser-company-yzcwxhrwj","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,ai-agents,leadership,claude-code,prompt-engineering","timeRequired":"PT18M","video":{"@type":"VideoObject","name":"Prototyping as Leadership: How a CTO Ships with AI Agents — Hursh Agrawal, The Browser Company","description":"The CTO and co-founder of The Browser Company describes how autonomous coding agents let him ship 2-10 PRs a week despite a packed manager schedule and a...","thumbnailUrl":"https://i.ytimg.com/vi/bdHaOXZOhcM/sddefault.jpg","uploadDate":"2026-08-20T14:28:37.671Z","duration":"PT18M","url":"https://api.daily.dev/r/YzcwxhRWJ","embedUrl":"https://www.youtube.com/embed/bdHaOXZOhcM"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"AI Engineer","item":"https://daily.dev/sources/aidotengineer"},{"@type":"ListItem","position":3,"name":"Prototyping as Leadership: How a CTO Ships with AI Agents — Hursh Agrawal, The Browser Company"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/prototyping-as-leadership-how-a-cto-ships-with-ai-agents-hursh-agrawal-the-browser-company-yzcwxhrwj#faq","mainEntity":[{"@type":"Question","name":"How can I set up an overnight workflow with Claude Code so it works autonomously for hours without me babysitting it?","acceptedAnswer":{"@type":"Answer","text":"Give the agent maximum business context upfront rather than breaking the task into micro-prompts, since it will run unsupervised for 4-8 hours. Have it write tests first, verify the end-to-end flow with computer use, split work into reviewer-friendly PRs, run an AI code review skill on itself and fix the findings, keep CI green, and leave a morning report on trade-offs made. Developers building overnight agent workflows compare real setups and prompting patterns on daily.dev."}},{"@type":"Question","name":"How do you build an eval set to hill climb an AI feature when you only have a small amount of user feedback?","acceptedAnswer":{"@type":"Answer","text":"Add a lightweight feedback button to the feature, collect as few as four to thirty feedback entries saved as JSON dumps with the system prompt and inputs, then have a coding agent turn those into a local eval set (SQLite or Markdown works) and build a harness that runs the call against the evals repeatedly, explicitly instructing it not to overfit on the small sample. Teams tuning AI feature quality can trade eval and hill-climbing approaches on daily.dev."}},{"@type":"Question","name":"Can modern coding agents actually train a custom ML model like a PII classifier overnight without much ML expertise?","acceptedAnswer":{"@type":"Answer","text":"Yes, a modernBERT PII classifier was trained overnight after Opus and Haiku proved too expensive and slow with insufficient precision-recall, by giving a coding agent business context, training data, open AI and Anthropic API keys, and sandboxed AWS access to pick a model class, train two candidate models, evaluate, and deprovision resources, returning trained models with a morning report. Engineers exploring agent-driven model training can follow similar experiments and results on daily.dev."}}]} ```