{"slug": "openai-makes-codex-s-auto-review-free-drawing-no-usage-from-your-plan", "title": "OpenAI makes Codex's Auto-review free, drawing no usage from your plan", "summary": "OpenAI's Codex team made Auto-review, the second agent that approves Codex's risky actions, free for ChatGPT sign-ins and no longer drawing usage from a user's plan, as day two of a 28-day pledge posted by Codex team member Tibo Sottiaux. OpenAI had said Auto-review approves around 99% of what it reviews and cuts human-approval stops roughly 200x versus manual mode. Separately, the research newsletter SemiAnalysis measured Opus 5.5 on a Claude plan at about 5x the API-equivalent value of GPT-6.1 Sol on OpenAI's plan, and OpenAI's Codex pricing page lowered its estimate for a typical GPT-5.6 Sol task to 2-15 credits from 5-30 credits as of October 1.", "body_md": "# OpenAI makes Codex's Auto-review free, drawing no usage from your plan\n\nDay two of the Codex team's 28-day pledge: Auto-review, a second agent that approves risky actions, no longer draws on your plan.\n\nTibo Sottiaux of OpenAI's Codex team shipped day two of a 28-day pledge on Tuesday: Auto-review, which lets a second agent approve Codex's risky actions, is now free for ChatGPT sign-ins and draws no usage. Separately, the research newsletter SemiAnalysis says Opus 5.5 on a Claude plan buys about five times the value of GPT-6.1 Sol on OpenAI's, measured at API prices. OpenAI's Codex pricing page also lowered its estimate of how many credits a typical GPT-5.6 Sol task uses. Cognition's coding agent Devin now tidies its memory of how you work once a day. And Claude Code's newest patch fixes a bug from the one before it.\n\n## What changed this week\n\n### OpenAI makes Codex's Auto-review free, drawing no usage from your plan\n\nTibo Sottiaux of OpenAI's Codex team posted day two of the team's 28-day pledge, which promises one clear improvement or a full usage reset each day: Auto-review is now free for anyone signed in with a ChatGPT account and \"does not draw usage from your plan.\" Auto-review hands approval of risky actions to a second agent instead of you. [OpenAI shipped it in April](https://alignment.openai.com/auto-review/?ref=thenewway.ai), saying sessions then stop for human approval \"roughly 200x less often\" than in manual approval mode, and that \"Auto-review approves around 99%\" of what it reviews. [Day one, on Monday](https://x.com/thsottiaux/status/2107158998495748264?ref=thenewway.ai), promised a \"~50% faster\" default speed through the subscription. We found no measurement of that on the subscription side; [Artificial Analysis's speed numbers](https://artificialanalysis.ai/models/gpt-6-1-sol?ref=thenewway.ai) track the API.\n\n### SemiAnalysis says Claude plans give about 5x OpenAI's value on mid-tier models\n\nThe research newsletter SemiAnalysis measured how far each lab's usage meter moves per token and priced the result at API rates. By that measure, Opus 5.5 on a Claude plan is \"an overwhelmingly better deal, offering ~5x the API-equivalent value\" of GPT-6.1 Sol on OpenAI's, under SemiAnalysis's own September usage mix. It also says new buyers of OpenAI's $200 plan get half the value, while older plans keep higher limits \"until October 29th.\" Its comparison of third-party plans like Cursor and Devin is paywalled.\n\n### OpenAI lowers its estimate for a typical GPT-5.6 Sol task to 2-15 credits\n\nOpenAI's Codex pricing page now says \"A typical GPT-5.6 Sol task may use 2-15 credits,\" down from \"5-30 credits\" as recently as October 1. Credits are what you spend once you pass a plan's included limits. The same page says \"GPT-5.6 Sol, Terra, and Luna rates remain unchanged,\" so the lower figure is a new estimate of how many credits a task takes, not a cheaper credit.\n\n### Devin adds a daily \"Dreaming\" pass that edits its own memory\n\nCognition's coding agent Devin now keeps a \"Memory Drive,\" which Cognition calls \"a persistent Git repository of markdown files\" holding what Devin has learned about how you work. \"Dreaming is a daily asynchronous process that improves the index of Devin's memories about you.\" At the start of each session, Devin reads a short MEMORY.md file and searches the rest as needed. Cognition says it has [open-sourced the standard](https://cognition.com/agent-memory-repo?ref=thenewway.ai) behind the system.\n\n### Claude Code's newest patch fixes a bug its last patch caused\n\nClaude Code 2.1.291 fixes \"a regression in 2.1.290 where cloud sessions could drop answers to permission prompts,\" plus one from 2.1.288 that could lose a session's last messages on quit. 2.1.290 itself added more hooks for mods, the plugins that reshape Claude Code's interface, and fixed WebFetch \"silently dropping page text past 100,000 characters.\" It also stopped /loop runs and scheduled reminders from silently failing to come back on resume after a compaction, for compactions made from this version on.\n\n## Also worth your time\n\nKnow someone who'd want this in their inbox? Forward it — that's how this grows. And if we got something wrong, or you think we buried the real story today, hit reply. A person reads every one.\n\n*The New Way is written with AI. It gathers the day's stories, checks them against their sources and drafts every summary. A person decides what runs and reviews every issue before we hit send.*", "url": "https://wpnews.pro/news/openai-makes-codex-s-auto-review-free-drawing-no-usage-from-your-plan", "canonical_source": "https://www.thenewway.ai/openai-makes-codexs-auto-review-free-drawing-no-usage-from-your-plan/", "published_at": "2026-10-06 15:27:56+00:00", "updated_at": "2026-10-06 15:48:11.830340+00:00", "lang": "en", "topics": ["ai-agents", "ai-products", "ai-tools", "large-language-models", "developer-tools"], "entities": ["OpenAI", "Codex", "Tibo Sottiaux", "ChatGPT", "SemiAnalysis", "Claude", "GPT-6.1 Sol", "Cognition"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/openai-makes-codex-s-auto-review-free-drawing-no-usage-from-your-plan", "markdown": "https://wpnews.pro/news/openai-makes-codex-s-auto-review-free-drawing-no-usage-from-your-plan.md", "text": "https://wpnews.pro/news/openai-makes-codex-s-auto-review-free-drawing-no-usage-from-your-plan.txt", "jsonld": "https://wpnews.pro/news/openai-makes-codex-s-auto-review-free-drawing-no-usage-from-your-plan.jsonld"}}