So GPT-6 Cyber Sucks at Coding but is Better at Hacking!?
OpenAI's new GPT-6 Cyber wrote working root exploits for 948 Linux kernel bugs for $1.2M in API costs, and it's what we got this week instead of GPT-6.2. I go through what the Linux run actually shows, who can use it and what it costs, the skeptical side and what to patch, DeepMind's 1024-bit RSA results, METR's report on Anthropic's unreleased models, the Sonnet 5.5 Extra leak, and the Claude Code bug that sent your email to websites. 🔔 SUBSCRIBE TO BYTEFORWARD AI news, coding agents, model comparisons, benchmarks, and the things people are actually building: https://www.youtube.com/channel/UCGGHPOAzUaKsSc2yjsOeAFQ?sub_confirmation=1 CHAPTERS 0:00 Greg Brockman's thousand Linux bugs 1:00 What happened to GPT-6.2 2:07 What GPT-6 Cyber actually did 4:12 Who can use it and what it costs 5:29 The skeptical side and what to patch 6:55 DeepMind and 1024-bit RSA 8:41 Anthropic's cyber program 9:11 METR's report on Anthropic 11:55 The Sonnet 5.5 Extra leak 13:27 Claude Code was sending your email 14:34 What I am watching next week Confirmed vs claimed: the 948 bugs, the costs and the 81 Firefox bugs are OpenAI's own counts; Mozilla's Firefox 159 advisory credits OpenAI on 41 entries. GPT-6.2 names on Google Cloud come from user screenshots and aren't confirmed by OpenAI. Polymarket odds were the single-day market when I checked (Nov 12 at 25%) and they move. The Opus 5.5 index value is read from METR's chart. The Sonnet 5.5 Extra post is a user's claim; the 1,860-line figure is from the post itself. OPENAI A thousand fewer paths to root: https://openai.com/index/a-thousand-fewer-paths-to-root/ Introducing GPT-6 Cyber: https://openai.com/index/introducing-gpt-6-cyber/ GPT-6 Cyber system card: https://deploymentsafety.openai.com/gpt-6-cyber GPT-6 Cyber model and pricing: https://developers.openai.com/api/docs/models/gpt-6-cyber OpenAI models list: https://developers.openai.com/api/docs/models Daybreak, trusted access for cyber: https://help.openai.com/en/articles/20001258-trusted-access-for-cyber Cyber at the Frontier (Nov 12): https://openai.com/cyber-at-the-frontier FIREFOX AND BROTLI Mozilla MFSA 2026-118 (Firefox 159): https://www.mozilla.org/en-US/security/advisories/mfsa2026-118/ Brotli fixed versions (Patch the Planet): https://patchtheplanet.org/brotli GOOGLE DEEPMIND Announcement: https://x.com/GoogleDeepMind/status/2118755418675765656 Signatures paper: https://storage.googleapis.com/deepmind-media/papers/Faster_Affine_Root_and_a_1024bit%20Computation.pdf Batch factoring paper: https://storage.googleapis.com/deepmind-media/papers/Fan_Factories_Multiple_Number_Fields_in_Factorization_Factories_and_Batch_NFS.pdf ANTHROPIC AND CLAUDE METR assessment: https://metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment/ Claude Platform release notes (IP allowlist): https://platform.claude.com/docs/en/release-notes/overview Project Glasswing / Cyber Verification Program: https://www.anthropic.com/glasswing Models overview (knowledge cutoff): https://platform.claude.com/docs/en/about-claude/models/overview Effort docs: https://platform.claude.com/docs/en/build-with-claude/effort Claude Code model configuration (ultrathink): https://code.claude.com/docs/en/model-config Claude Code CHANGELOG: https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md DISCUSSIONS Hacker News, Claude Code email: https://news.ycombinator.com/item?id=50385272 Claude Code team reply: https://news.ycombinator.com/item?id=50388176 Hacker News, security reports comment: https://news.ycombinator.com/item?id=50392523 Polymarket single-day market: https://polymarket.com/event/gpt-astra-6pt1-released-onptptpt POSTS SHOWN @steph_palazzolo https://x.com/steph_palazzolo/status/2117997555778744815 @AiBattle_ https://x.com/AiBattle_/status/2118123362245587208 @Mr_Salio https://x.com/Mr_Salio/status/2118167284237611051 @gdb https://x.com/gdb/status/2118402749817926095 @OpenAI https://x.com/OpenAI/status/2118390190246879554 @atharvabuilds https://x.com/atharvabuilds/status/2118393933235667062 @i0n1c https://x.com/i0n1c/status/2118409855119880354 @METR_Evals https://x.com/METR_Evals/status/2118412183860351084 https://x.com/METR_Evals/status/2118412205662314935 @Sam_McCandlish https://x.com/Sam_McCandlish/status/2118422266669371519 @chetaslua https://x.com/chetaslua/status/2118323489560219791 @karpathy https://x.com/karpathy/status/2056753169888334312 FOOTAGE WSJ Tech Live, Greg Brockman: https://www.youtube.com/watch?v=yfnmAsXqvIs OpenAI, GPT-6 Cyber launch and Cyber at the Frontier clips: https://x.com/OpenAI/status/2118390190246879554 https://x.com/OpenAI/status/2118390195099930906 OpenAI, How Mozilla strengthens Firefox security with GPT-6 Cyber: https://www.youtube.com/watch?v=pXRZyOxbDL4 #GPT6 #OpenAI #Claude
