--- title: Devs are fleeing Claude Code, and that's the real story description: Developers migrating from Claude Code to Codex are sparking a debate about what the exodus actually signals. Some, like svpino, frame it as proof Anthropic's... canonical: https://daily.dev/posts/devs-are-fleeing-claude-code-and-that-s-the-real-story-yrvvvz79a twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Devs are fleeing Claude Code, and that's the real story | daily.dev og:description: Developers migrating from Claude Code to Codex are sparking a debate about what the exodus actually signals. Some, like svpino, frame it as proof Anthropic's... og:url: https://daily.dev/posts/devs-are-fleeing-claude-code-and-that-s-the-real-story-yrvvvz79a og:image: https://api.daily.dev/og/posts/yrVvvZ79a.png og:image:alt: Devs are fleeing Claude Code, and that's the real story 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. # Devs are fleeing Claude Code, and that's the real story **[Trends](https://daily.dev/sources/trends)** · 2 min read · 4 upvotes · 2 comments ## Summary Developers migrating from Claude Code to Codex are sparking a debate about what the exodus actually signals. Some, like svpino, frame it as proof Anthropic's products lack stickiness and that the switch was painless. Others, notably QuinnyPig, argue the real takeaway isn't Anthropic-specific decline but that coding assistants and their harnesses are becoming interchangeable commodities with near-zero switching costs, which should worry every vendor in the space, not just Anthropic. ## Content The Claude Code vs Codex debate has been running hot for weeks, and the honest answer is: it depends, and everyone has a different "it." The flashpoint is migration. @svpino declared he moved from Claude Code to Codex two weeks ago and "doesn't miss a single thing," adding that "Anthropic's products have zero stickiness. Anyone can migrate away from them in an hour." @QuinnyPig pushed back on the doom-reading: people treating the Claude-to-Codex exodus as proof "Anthropic is doomed" are missing the actual takeaway, which is that "coding assistants and their harnesses are trivially interchangeable." That framing is probably the most useful one here. **What the actual comparisons show** A hands-on benchmark across five identical tasks found GPT-5.6 Soul, Opus 5, and Fable 5 tied at 34.7% on SWE-Suite's "tasteful solve rate." The models are roughly equivalent. The harness is where they diverge. Codex runs 3-4x cheaper and faster. Claude Code produces more thorough, test-heavy output but takes longer. One developer compared both on a Japanese vertical-text layout bug: Codex burned 1.5 million tokens and changed 2,189 lines to produce a mess; Claude used 37,000 tokens, changed one line, and fixed it. Pricing is a real wedge. Codex's $20 plan includes the flagship Soul model. Claude's $20 plan gives you Sonnet 5 — Opus and Fable require the $100 Max plan. @theo's take is the most honest split: "Claude Code CLI > Codex CLI. Codex desktop app > Claude Code desktop app. I will not be taking further questions." A practitioner framework that's getting traction: use Codex for single hard tasks driven to completion, use Claude Code for orchestrating 50-100 parallel subagents per day, since Codex loses track when juggling multiple threads. **The open-source complaint** @yacineMTB landed a pointed shot: "It's honestly insane that Claude Code is closed source. It's a JavaScript bundle that makes API calls and all your competitors are open. Are you a model harness company or an AI company?" Anthropics's team has been shipping fast regardless — CLI startup is now 20% faster, token visibility improved, auto mode rules are easier to edit. @corbtt says Codex compaction has gotten good enough that he keeps the same thread open for weeks instead of manually dumping state to markdown files. Bottom line: the models are close enough that workflow fit and pricing are doing most of the deciding. Pick your harness, not your religion. ## Questions this post answers ### Is it hard to switch from Claude Code to Codex? No, developers report the switch is fast and low-friction. One developer described taking about a day to get used to the change after two weeks of use, with no regrets and nothing missed, calling Anthropic's products low on stickiness since anyone can migrate away within an hour. _Weigh the real switching costs between AI coding tools using daily.dev before committing to one._ ### Why are developers leaving Claude Code for Codex? Many are switching because the tools feel largely interchangeable rather than because Claude Code got dramatically worse. Commentary framed the exodus not as proof Anthropic is doomed, but as evidence that coding assistants and their underlying harnesses are trivially interchangeable, meaning low switching costs make user churn normal rather than fatal for any single vendor. _Compare shifting opinions on AI coding assistants like Claude Code and Codex through daily.dev._ ## Community take How the wider developer community reacted, aggregated from 2 discussions and 49 comments across x (as of 2026-08-31). **TL;DR:** The available discussion barely touches the Claude Code vs Codex comparison itself, instead sprawling into tangential arguments about a Codex-focused book's value and about long-running context/session management in agentic coding tools. **Sentiment:** 15% positive · 45% mixed · 40% skeptical **The case for** - Some report that long-running pinned threads with newer models hold up well over weeks without noticeable context degradation. - A few see workflow patterns and skill-building around tools like Codex as valuable beyond any single trick. **The pushback** - Several argue technical books on fast-moving AI tools go obsolete almost immediately and prefer updatable digital sources. - Skeptics accuse the poster of shilling products for referral clicks rather than genuine recommendations. - Some warn that automatic context compaction hides what the model 'thinks it knows,' making failures harder to audit than a manually written handoff file. - One commenter argues bloated context windows quietly degrade output quality even when users don't notice. **By community** - x (mixed): Replies split between hype for a Codex workflow book and pointed skepticism about promotional motives and the durability of printed AI advice, plus a separate thread arguing over whether long-context sessions and auto-compaction help or quietly erode reliability. **Hottest debate:** Whether relying on long-running threads with automatic context compaction is a genuine productivity win or a hidden reliability risk compared to manual handoff notes. **Open questions** - Does context compaction reliably preserve important early constraints, or does it silently drop them over long sessions? - Is there real evidence that book-style guides on a specific harness stay useful given how fast these tools change? **Highlights** > @corbtt what you gave up is being able to see what it thinks it knows. a handoff file you wrote yourself is auditable. a compaction youve never read isnt, so a wrong call from week one just quietly stays in the summary > — [AIQuanting on x](https://x.com/AIQuanting/status/2094507278313705821) > @corbtt AI is so smart nowadays that people don't notice the IQ penalty they're paying by letting the context window get too bloated, but that doesn't mean the IQ didn't genuinely decline. So yeah, if you're fine with that and with the not knowing, then don't worry at all about context. > — [\_quanta\_ on x](https://x.com/_quanta_/status/2094550307422224585) > @svpino Shilling something every other week. Not a problem to say things he does not believe and suggest things he didn’t use or read. All he cares is the count of clicks on that referral link to make some extra bucks on his followers. Miss the days he used to share hardcore projects > — [orcunozyurt on x · 1 points](https://x.com/orcunozyurt/status/2094427698298786124) > @Chahatusharma With gpt 5.6 sol I never see this anymore. > — [corbtt on x](https://x.com/corbtt/status/2094465242793435257) > @svpino The alternative IMO is digital sources that can be updated as things change. Instead of a bunch of printed sheets of paper. I'm not against reading. I'm talking about how not durable the printed medium is for these topics. > — [gnapse on x · 1 comments](https://x.com/gnapse/status/2094465638102090180) **Source threads** - [x](https://x.com/svpino/status/2094402835521966117) · 0 points · 36 comments - [x](https://x.com/corbtt/status/2094447806295797958) · 0 points · 13 comments ## Community discussion Top comments from developers on daily.dev. **@tokstesla** · 0 upvotes > This was a brilliant read. I think most users are loyal to affordability and speed of execution. **@dmytriis** · 0 upvotes > Moving [CLAUDE.md](http://CLAUDE.md), hooks and MCP config is cheap. Moving a _live_ session is not. Reasoning state, tool-call identity and provider-native replay semantics leak through the harness. > > Static config portability is mostly solved. Session portability is where the lock-in moved. --- Tags: [#anthropic](https://daily.dev/tags/anthropic), [#claude-code](https://daily.dev/tags/claude-code), [#vibe-coding](https://daily.dev/tags/vibe-coding), [#openai-codex](https://daily.dev/tags/openai-codex) [View this post on daily.dev](https://daily.dev/posts/devs-are-fleeing-claude-code-and-that-s-the-real-story-yrvvvz79a) ```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":"Devs are fleeing Claude Code, and that's the real story","url":"https://daily.dev/posts/devs-are-fleeing-claude-code-and-that-s-the-real-story-yrvvvz79a","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/devs-are-fleeing-claude-code-and-that-s-the-real-story-yrvvvz79a"},"datePublished":"2026-08-24T19:16:30.049Z","dateModified":"2026-08-31T22:36:20.313Z","description":"Developers migrating from Claude Code to Codex are sparking a debate about what the exodus actually signals. Some, like svpino, frame it as proof Anthropic's...","isAccessibleForFree":true,"articleSection":"Trends","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":"Trends","logo":"https://media.daily.dev/image/upload/s--ZfSp3asX--/f_auto,q_auto/v1780996004/logos/trends?_a=BAMAMiWQ0","url":"https://daily.dev/sources/trends"},"commentCount":2,"discussionUrl":"https://daily.dev/posts/devs-are-fleeing-claude-code-and-that-s-the-real-story-yrvvvz79a","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":4},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":2}],"keywords":"anthropic,claude-code,vibe-coding,openai-codex","timeRequired":"PT2M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Trends","item":"https://daily.dev/sources/trends"},{"@type":"ListItem","position":3,"name":"Devs are fleeing Claude Code, and that's the real story"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/devs-are-fleeing-claude-code-and-that-s-the-real-story-yrvvvz79a","comment":[{"@type":"Comment","text":"This was a brilliant read. I think most users are loyal to affordability and speed of execution.","datePublished":"2026-08-27T16:01:49.975Z","url":"https://daily.dev/posts/yrVvvZ79a#c-yNwHeHCDc","author":{"@type":"Person","name":"Tochukwu Okeh","url":"https://daily.dev/tokstesla","image":"https://lh3.googleusercontent.com/a/ALm5wu0M9zC8KvFRBWLakb25xsuWq6eewxwbl6yGQ8Gp=s96-c"}},{"@type":"Comment","text":"Moving CLAUDE.md, hooks and MCP config is cheap. Moving a live session is not. Reasoning state, tool-call identity and provider-native replay semantics leak through the harness.\nStatic config portability is mostly solved. Session portability is where the lock-in moved.","datePublished":"2026-08-31T14:31:35.966Z","url":"https://daily.dev/posts/yrVvvZ79a#c-6i1GB4Ek4","author":{"@type":"Person","name":"Dmytrii S.","url":"https://daily.dev/dmytriis","image":"https://avatars.githubusercontent.com/u/946519?v=4"}}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/devs-are-fleeing-claude-code-and-that-s-the-real-story-yrvvvz79a#faq","mainEntity":[{"@type":"Question","name":"Is it hard to switch from Claude Code to Codex?","acceptedAnswer":{"@type":"Answer","text":"No, developers report the switch is fast and low-friction. One developer described taking about a day to get used to the change after two weeks of use, with no regrets and nothing missed, calling Anthropic's products low on stickiness since anyone can migrate away within an hour. Weigh the real switching costs between AI coding tools using daily.dev before committing to one."}},{"@type":"Question","name":"Why are developers leaving Claude Code for Codex?","acceptedAnswer":{"@type":"Answer","text":"Many are switching because the tools feel largely interchangeable rather than because Claude Code got dramatically worse. Commentary framed the exodus not as proof Anthropic is doomed, but as evidence that coding assistants and their underlying harnesses are trivially interchangeable, meaning low switching costs make user churn normal rather than fatal for any single vendor. Compare shifting opinions on AI coding assistants like Claude Code and Codex through daily.dev."}}]} ```