Daily Spotlight — Nov 6, 2026: Jakub Pachocki

Big AI news

• OpenAI launches GPT-6 Cyber for authorized security research · OpenAI

  OpenAI released GPT-6 Cyber, a cybersecurity model that finds and validates exploits while staying inside the scope researchers set. It scores 48.5% on ExploitGym (up from 32.1% for GPT-5.6 Cyber), helped uncover 81 novel high-severity Firefox bugs and develop exploits for 948 Linux privilege-escalation issues, and is rated Critical for cybersecurity under OpenAI’s Preparedness Framework. Initial access is limited to approved Daybreak Red customers.

  OpenAI

• METR says Anthropic’s AI progress is roughly twice as fast · METR

  Independent evaluator METR’s Nov 5 assessment finds Anthropic is likely improving model capabilities on well-specified tasks at about twice the March 2024–October 2025 rate, with most of today’s acceleration attributed to AI-assisted R&D rather than faster compute scaling. Anthropic redacted much of the most informative evidence, so the public has to lean on METR’s judgment more than the authors would like.

  METR

• Claude Personal Finance connects bank accounts via Plaid · Anthropic

  Anthropic opened a US-only Personal Finance beta for Claude Pro, Max, and Ultra that links bank and card accounts through Plaid with read-only access. Claude can summarize spending and up to two years of transactions, but cannot move or invest money, and chats in the Personal Finance hub are excluded from model training.

  Anthropic

• Google’s Gemini Omni 1.2 Flash makes 15-second video cheaper · Google

  Google generally available’d Gemini Omni 1.2 Flash for video generation and editing, extending native clips to 15 seconds (from 10), adding multi-turn conversational edits, and cutting price 30% to $0.07 per second at 720p. It is live on the Gemini API and Google Cloud.

  Google

• Thinking Machines Lab ships Inkling 2 Small · Thinking Machines Lab

  Mira Murati’s Thinking Machines Lab released Inkling 2 Small, a multimodal Mixture-of-Experts model with 594B total parameters and 17B active, available now on Tinker with closed weights. The lab plans an Apache 2.0 open-weight release next week after external safety testing.

  Thinking Machines Lab

People spotlight

Jakub Pachocki — Chief Scientist · OpenAI

OpenAI’s Chief Scientist, overall pretraining lead for GPT-4, and a longtime research leader on OpenAI Five and o1. The GPT-4 Technical Report credits him as overall lead and optimization lead for pretraining, and the o1 System Card lists him in both reasoning research leadership and executive leadership. Why now: OpenAI’s Nov 5 launch of GPT-6 Cyber — a cybersecurity model rated Critical under its Preparedness Framework, already used to find real Firefox and Linux bugs — puts the lab’s research leadership back at the center of how frontier models are built and where they are allowed to go. Pachocki’s remit as Chief Scientist sits squarely behind releases like this: models trained for hard technical work, then gated for trusted defenders.

Read Jakub Pachocki’s wiki page →