--- title: Anthropic Says Nobody Needs a Mental Model of Claude... description: Anthropic's Claude Code team says nobody keeps a complete mental model of Claude Code anymore, now that Claude writes about 90% of it. Instead, engineers ask... canonical: https://daily.dev/posts/anthropic-says-nobody-needs-a-mental-model-of-claude-code-practitioners-are-testing-that-pa5017zey twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Anthropic Says Nobody Needs a Mental Model of Claude Code. Practitioners Are Testing That | daily.dev og:description: Anthropic's Claude Code team says nobody keeps a complete mental model of Claude Code anymore, now that Claude writes about 90% of it. Instead, engineers ask... og:url: https://daily.dev/posts/anthropic-says-nobody-needs-a-mental-model-of-claude-code-practitioners-are-testing-that-pa5017zey og:image: https://api.daily.dev/og/posts/Pa5017ZEY.png og:image:alt: Anthropic Says Nobody Needs a Mental Model of Claude Code. Practitioners Are Testing That 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. # Anthropic Says Nobody Needs a Mental Model of Claude Code. Practitioners Are Testing That **[InfoQ](https://daily.dev/sources/infoq)** · 5 min read · 0 upvotes · 0 comments ## Summary Anthropic's Claude Code team says nobody keeps a complete mental model of Claude Code anymore, now that Claude writes about 90% of it. Instead, engineers ask Claude to generate an explainer for whatever slice they need. A risk-tier classifier decides who reviews a change: low-risk changes merge with no human approval, medium-risk ones need one person, and an agent's approval never counts as a person's. One pull request went through more than fifty rounds of review and fixes overnight before its author stopped it, and about 59% of what the reviewer kept flagging was code that had been there since its first pass, not new bugs. The team admits it lacks circuit breakers to tighten approval automatically when review metrics worsen. A Hacker News commenter questioned whether an explainer-on-demand replaces a causal model in someone's head for debugging. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### what risk tiers does Anthropic use to decide when a human must review a Claude Code pull request Anthropic uses a classifier-assigned risk tier where the change itself sets the tier, not the author. Low risk, covering docs, copy, and additive flagged changes, clears with no human approval. Medium risk, including caching, the auto-updater, and system prompts, needs one human reviewer. An agent's approval never counts as a required human reviewer. _Teams weighing how much to automate code review can follow real-world risk-tiering examples like this on daily.dev._ ### how many review rounds did a Claude Code pull request go through before a human stopped it One continuous integration pull request went through more than fifty rounds of automated review and fixes overnight before its human author stopped it. Anthropic's engineering team found that running unlimited review rounds by default made outcomes worse, so they capped the number of automatic rounds allowed per pull request. _Developers tuning agent review loops can track emerging guardrails like this on daily.dev._ ### what percentage of flagged issues in Claude Code's AI reviewer were pre-existing bugs versus new ones About 59% of what Claude Code's automated reviewer flagged during repeated review rounds was code that had already existed since the reviewer's first pass, not new bugs introduced by the latest change. The reviewer was mostly surfacing a few additional old issues each round rather than catching fresh regressions. _Anyone deciding whether to trust AI code reviewers can follow findings like this on daily.dev._ --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#architecture](https://daily.dev/tags/architecture), [#anthropic](https://daily.dev/tags/anthropic), [#claude-code](https://daily.dev/tags/claude-code), [#code-review](https://daily.dev/tags/code-review) [View this post on daily.dev](https://daily.dev/posts/anthropic-says-nobody-needs-a-mental-model-of-claude-code-practitioners-are-testing-that-pa5017zey) ```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":"Anthropic Says Nobody Needs a Mental Model of Claude Code. Practitioners Are Testing That","url":"https://daily.dev/posts/anthropic-says-nobody-needs-a-mental-model-of-claude-code-practitioners-are-testing-that-pa5017zey","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/anthropic-says-nobody-needs-a-mental-model-of-claude-code-practitioners-are-testing-that-pa5017zey"},"datePublished":"2026-10-29T10:01:52.066Z","dateModified":"2026-10-30T15:48:41.313Z","description":"Anthropic's Claude Code team says nobody keeps a complete mental model of Claude Code anymore, now that Claude writes about 90% of it. Instead, engineers ask...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/3923edadc33857f1c2b60b6f23a7c298?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/3923edadc33857f1c2b60b6f23a7c298?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"InfoQ","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":"InfoQ","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/afc3bced3e1e4b188dd9127017a60e0c","url":"https://daily.dev/sources/infoq"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/anthropic-says-nobody-needs-a-mental-model-of-claude-code-practitioners-are-testing-that-pa5017zey","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,architecture,anthropic,claude-code,code-review","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"InfoQ","item":"https://daily.dev/sources/infoq"},{"@type":"ListItem","position":3,"name":"Anthropic Says Nobody Needs a Mental Model of Claude Code. Practitioners Are Testing That"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/anthropic-says-nobody-needs-a-mental-model-of-claude-code-practitioners-are-testing-that-pa5017zey#faq","mainEntity":[{"@type":"Question","name":"what risk tiers does Anthropic use to decide when a human must review a Claude Code pull request","acceptedAnswer":{"@type":"Answer","text":"Anthropic uses a classifier-assigned risk tier where the change itself sets the tier, not the author. Low risk, covering docs, copy, and additive flagged changes, clears with no human approval. Medium risk, including caching, the auto-updater, and system prompts, needs one human reviewer. An agent's approval never counts as a required human reviewer. Teams weighing how much to automate code review can follow real-world risk-tiering examples like this on daily.dev."}},{"@type":"Question","name":"how many review rounds did a Claude Code pull request go through before a human stopped it","acceptedAnswer":{"@type":"Answer","text":"One continuous integration pull request went through more than fifty rounds of automated review and fixes overnight before its human author stopped it. Anthropic's engineering team found that running unlimited review rounds by default made outcomes worse, so they capped the number of automatic rounds allowed per pull request. Developers tuning agent review loops can track emerging guardrails like this on daily.dev."}},{"@type":"Question","name":"what percentage of flagged issues in Claude Code's AI reviewer were pre-existing bugs versus new ones","acceptedAnswer":{"@type":"Answer","text":"About 59% of what Claude Code's automated reviewer flagged during repeated review rounds was code that had already existed since the reviewer's first pass, not new bugs introduced by the latest change. The reviewer was mostly surfacing a few additional old issues each round rather than catching fresh regressions. Anyone deciding whether to trust AI code reviewers can follow findings like this on daily.dev."}}]} ```