--- title: The Creator of Claude Code Doesn’t Prompt It Anymore description: Boris Cherny, creator of Claude Code, no longer manually prompts the tool — instead he writes autonomous loops that do the prompting for him. This concept,... canonical: https://daily.dev/posts/the-creator-of-claude-code-doesn-t-prompt-it-anymore-5sxi90nj0 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: The Creator of Claude Code Doesn’t Prompt It Anymore | daily.dev og:description: Boris Cherny, creator of Claude Code, no longer manually prompts the tool — instead he writes autonomous loops that do the prompting for him. This concept,... og:url: https://daily.dev/posts/the-creator-of-claude-code-doesn-t-prompt-it-anymore-5sxi90nj0 og:image: https://api.daily.dev/og/posts/5SXI90nJ0.png og:image:alt: The Creator of Claude Code Doesn’t Prompt It Anymore 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. # The Creator of Claude Code Doesn’t Prompt It Anymore **[Towards Dev](https://daily.dev/sources/towardsdev)** · 12 min read · 18 upvotes · 3 comments ## Summary Boris Cherny, creator of Claude Code, no longer manually prompts the tool — instead he writes autonomous loops that do the prompting for him. This concept, coined 'loop engineering' by Addy Osmani in June 2026, represents a shift from writing individual prompts to designing systems that orchestrate AI agents continuously. The post explains the three-level ladder of prompt → context → loop engineering, defines four outer loop types (heartbeat, cron, hook, goal), and provides a 30-line Python example of a goal loop for code migration. Key engineering concerns covered include stopping conditions, the verification split (the agent that writes code must not approve it), external state management to prevent amnesia across runs, and three guard layers: hard caps, verifiable conditions, and independent verification. The author also warns about quiet human failures like comprehension debt, orchestration tax, and cognitive surrender. A practical five-step guide for building your first loop is included. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is loop engineering in the context of AI coding agents? Loop engineering is designing the system around an AI agent's runs rather than manually prompting it each time. It covers five stages: a trigger that decides when to run, a frame step that assembles the task, the agent's run, an independent verification step, and a record step that writes state externally so the next iteration remembers progress. _Developers building autonomous agent pipelines can track emerging patterns like this on daily.dev._ ### What are the four types of outer loops used for autonomous AI agents? The four types are heartbeat loops that run continuously on a short interval until a budget cap stops them, cron loops that fire on a schedule and end when their batch is done, hook loops that fire once per event like a PR opening or CI failing, and goal loops that iterate until a success condition verifies true, such as all tests passing. _Anyone architecting agent workflows can follow practical breakdowns like this via daily.dev._ ### Why shouldn't an AI coding agent verify its own completed work? An agent that produced a change should not be the one approving it, because models are systematically overconfident about their own completions and will claim a task is done on plausible-looking but incorrect work. The recommended pattern uses one subagent to implement changes in an isolated git worktree and a second subagent with fresh context to review against conventions and run tests before release. _Teams designing reliable agent verification steps can keep up with patterns like this through daily.dev._ ## Community discussion Top comments from developers on daily.dev. **@seyoungcho** · 1 upvotes > Months later: The creator of Claude code doesn't even make loops anymore **@renehoehle** · 1 upvotes > And the next stop is loop in loop? **@kequc** · 1 upvotes > Last time I realized my coding agent and planning agent were both effectively working with me as a middleman, and I just made them talk to each other it fell apart within 30 minutes. > > Yes you still need to be the foreman. ## Similar posts on daily.dev - [The Anthropic leader who built Claude Code says he ditched prompting — now he just writes loops.](https://daily.dev/posts/the-anthropic-leader-who-built-claude-code-says-he-ditched-prompting-now-he-just-writes-loops--p8qgluhaf) · The New Stack · 1 upvotes · 0 comments - [You’re Not Prompting Anymore. You’re Engineering Loops.](https://daily.dev/posts/you-re-not-prompting-anymore-you-re-engineering-loops--aknavecvs) · Medium · 1 upvotes · 0 comments - [Loop Engineering](https://daily.dev/posts/loop-engineering-h663rbtdz) · Elevate · 3 upvotes · 0 comments - [An Introduction to Loop Engineering](https://daily.dev/posts/an-introduction-to-loop-engineering-7ctzwxfxz) · Machine Learning Mastery · 2 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#automation](https://daily.dev/tags/automation), [#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/the-creator-of-claude-code-doesn-t-prompt-it-anymore-5sxi90nj0) ```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":"The Creator of Claude Code Doesn’t Prompt It Anymore","url":"https://daily.dev/posts/the-creator-of-claude-code-doesn-t-prompt-it-anymore-5sxi90nj0","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/the-creator-of-claude-code-doesn-t-prompt-it-anymore-5sxi90nj0"},"datePublished":"2026-07-20T16:10:27.027Z","dateModified":"2026-09-14T07:26:49.267Z","description":"Boris Cherny, creator of Claude Code, no longer manually prompts the tool — instead he writes autonomous loops that do the prompting for him. This concept,...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/489c3ba3fedcf4480c6fcada52d907f6?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/489c3ba3fedcf4480c6fcada52d907f6?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Towards Dev","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":"Towards Dev","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/2c710db2750a4a70835862534cb7c9d4","url":"https://daily.dev/sources/towardsdev"},"commentCount":3,"discussionUrl":"https://daily.dev/posts/the-creator-of-claude-code-doesn-t-prompt-it-anymore-5sxi90nj0","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":18},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":3}],"keywords":"ai-agents,automation,claude-code,prompt-engineering","timeRequired":"PT12M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Towards Dev","item":"https://daily.dev/sources/towardsdev"},{"@type":"ListItem","position":3,"name":"The Creator of Claude Code Doesn’t Prompt It Anymore"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/the-creator-of-claude-code-doesn-t-prompt-it-anymore-5sxi90nj0","comment":[{"@type":"Comment","text":"Months later: The creator of Claude code doesn’t even make loops anymore","datePublished":"2026-07-28T01:03:08.241Z","url":"https://daily.dev/posts/5SXI90nJ0#c-K93XeYRfH","author":{"@type":"Person","name":"SeyoungCho","url":"https://daily.dev/seyoungcho","image":"https://media.daily.dev/image/upload/s---QniJFEK--/f_auto/v1724901542/avatars/avatar_O3ucZSadxc0iLADSOH3rX"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1}},{"@type":"Comment","text":"And the next stop is loop in loop?","datePublished":"2026-08-02T14:46:34.999Z","url":"https://daily.dev/posts/5SXI90nJ0#c-9vicNnZ2c","author":{"@type":"Person","name":"René H.","url":"https://daily.dev/renehoehle","image":"https://lh3.googleusercontent.com/a/ACg8ocJ8hDDIeMWSluQCPM4cpeozwunDxUYydfDudbgcatmFJLbBF8IH=s96-c"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1}},{"@type":"Comment","text":"Last time I realized my coding agent and planning agent were both effectively working with me as a middleman, and I just made them talk to each other it fell apart within 30 minutes.\nYes you still need to be the foreman.","datePublished":"2026-07-25T01:21:00.524Z","url":"https://daily.dev/posts/5SXI90nJ0#c-TNULBlkAm","author":{"@type":"Person","name":"Kequc","url":"https://daily.dev/kequc","image":"https://avatars.githubusercontent.com/u/584075?v=4"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1}}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/the-creator-of-claude-code-doesn-t-prompt-it-anymore-5sxi90nj0#faq","mainEntity":[{"@type":"Question","name":"What is loop engineering in the context of AI coding agents?","acceptedAnswer":{"@type":"Answer","text":"Loop engineering is designing the system around an AI agent's runs rather than manually prompting it each time. It covers five stages: a trigger that decides when to run, a frame step that assembles the task, the agent's run, an independent verification step, and a record step that writes state externally so the next iteration remembers progress. Developers building autonomous agent pipelines can track emerging patterns like this on daily.dev."}},{"@type":"Question","name":"What are the four types of outer loops used for autonomous AI agents?","acceptedAnswer":{"@type":"Answer","text":"The four types are heartbeat loops that run continuously on a short interval until a budget cap stops them, cron loops that fire on a schedule and end when their batch is done, hook loops that fire once per event like a PR opening or CI failing, and goal loops that iterate until a success condition verifies true, such as all tests passing. Anyone architecting agent workflows can follow practical breakdowns like this via daily.dev."}},{"@type":"Question","name":"Why shouldn't an AI coding agent verify its own completed work?","acceptedAnswer":{"@type":"Answer","text":"An agent that produced a change should not be the one approving it, because models are systematically overconfident about their own completions and will claim a task is done on plausible-looking but incorrect work. The recommended pattern uses one subagent to implement changes in an isolated git worktree and a second subagent with fresh context to review against conventions and run tests before release. Teams designing reliable agent verification steps can keep up with patterns like this through daily.dev."}}]} ```