Thinking Machines’ Inkling 2 Small: a 594B AI That Works Alone for 4 Hours — but You Can’t Download It Yet
Mira Murati’s Thinking Machines Lab launched Inkling 2 Small on 5 November 2026 — a 594-billion-parameter model built for long, unsupervised work. The weights are still closed.
Google’s New Engineering Exam for AI Agents: 85 Real Tasks, and Every Agent Scores Under 50%
Google’s Physical-Engineering’s Last Exam gives AI agents 85 real engineering jobs — and every model scores under 50%. The best, Claude Opus 5.5, reaches 48.3%.
OpenAI GPT-6 Cyber Gets ‘Critical’ Rating
OpenAI’s GPT-6 Cyber is the first of its security models to earn a ‘Critical’ cybersecurity rating — and almost nobody can use it yet.
Anthropic’s FlashProve: The Cryptographic Receipt That Proves Which AI Really Answered You
Anthropic’s FlashProve is a cryptographic receipt for AI replies — proof of which model answered, without exposing its weights. Here’s what it can and can’t do.
Claude Mapped a 40-Year-Old Protein Vault Mystery
Anthropic’s biology team used Claude to solve the first 3D structure of a vault particle from bacterial life — a 96-copy protein barrel whose function remains unknown 40 years after vaults were discovered.
India’s EKA-1 and Dhīr AI Models Explained
Three homegrown foundation models launched under the IndiaAI Mission on 3 November 2026 — what EKA-1 and the offline Dhīr pair are actually for, and what is still unproven.
Muse AI Is Getting Its Own Shopping Carts — Instacart First
Muse is getting shopping carts it draws itself, Instacart first: the connector supplies cart data, Muse renders the cart, and you pay in the Instacart app.
Sarvam Prism: India’s 44-Millisecond Decision AI Explained
Sarvam Prism scores 71.8 in 44 milliseconds by deciding, not chatting. What the open-weight model does, where it beats rivals, and what the numbers do not prove.
The Best AI in the World Scores Just 12% on Real Work
SemiAnalysis GDPx puts the best AI at just 12% on 206 real professional tasks, with finance and legal at zero. What the scores mean — and what they do not.
Anthropic’s $17B Quarter and $895B Compute Bill
Anthropic’s 2 Nov 2026 S-1 amendment: $17.0B Q3 revenue, a $567M operating loss, $895.4B in compute commitments — and still no IPO price, share count or date.