01Anthropic Warns in Its IPO Filing It Is Near Self-Improving AI [AI Models & Research]In a revised IPO filing on Monday, Anthropic says its internal estimates suggest it may “imminently” hit its own threshold for “dramatically accelerated progress attributable to AI,” meaning AI driving AI progress, after capability leaps it admits it failed to spot in time. No current model has crossed the line, and the company pledges to keep humans supervising model development. → source
02A Startup Raising $62.5 Million Says It Can Predict Human Decisions [AI Models & Research]Israeli startup 42AI revealed $62.5 million in funding and a model that predicts how customers will decide, trained on support calls, transactions and behavioral data from roughly 30 million people. Ten brokerages are clients for trading-risk decisions, and one credits the model with $30 million in hedging gains. → source
03Y Combinator's CEO Spends Up to $200,000 a Year Running Agents [AI Tools & Ecosystem]On Walter Isaacson's podcast, Y Combinator CEO Garry Tan puts his AI habit at $100,000 to $200,000 a year in tokens, spent on agents that read his email, files and 20 years of therapy notes. He calls the cost a “crazy privilege” until prices fall. → source
04The Agentic Web Is Arriving Faster Than Its Designers Expected [AI Tools & Ecosystem]Agents that browse, click and buy for people are moving from demos to products, with Meta's Muse and OpenAI's Dots logging record downloads and, in places, access to users' bank accounts. Publishers are adjusting too: Time magazine already sells ads aimed at the search crawlers agents query. → source
05One Maintainer's Answer to AI Pull Request Floods: Use AI to Triage [AI Tools & Ecosystem]Simon Willison shipped version 0.37 of his llm command-line tool with fixes from 12 contributors, almost all AI-assisted, the flood many maintainers dread. His approach: run frontier models like GPT-6 Astra over the queue for triage, then treat each request as “a bug report with attached examples.” → source
06The Times Sues Brave, Calling It a Dealer for AI Companies [AI Applications & Industry]The New York Times sued browser company Brave, calling it a “supplier, dealer, and distributor” of scraped content for AI firms including Anthropic, Perplexity, Mistral, xAI and Cohere, and three Tribune-owned dailies filed alongside. The suit says Brave indexed more than 30 billion webpages with a crawler publishers could not block. → source
07Squarespace Lays Off 300 as AI Site Builders Eat Its Market [AI Applications & Industry]Squarespace is cutting about 19% of staff, roughly 300 jobs, as prompt-built site builders reshape its industry: Wix bought Base44 for $80 million, and Lovable raised $400 million at a $13.3 billion valuation. Squarespace, taken private in a $7 billion deal in 2024, is rushing out Forma, an AI-native platform. → source
08Musk Throttles Grok Bot After Costs Outran Revenue [AI Applications & Industry]Grok Bot users woke to maxed-out weekly quotas after Elon Musk conceded the agent's serving costs were “unsustainably high relative to revenue.” The same business is SpaceX's brightest spot: a filing last month put AI at 52% of income, and its Super Intelligence division booked a $431 million profit. → source
09Google Ships an AI Storyteller App to India After a Slow US Start [AI Applications & Industry]Google's Dreambeans app, which turns Calendar, Gmail and Photos data into illustrated personal stories generated unprompted, launched in India with 13 local languages, its first market outside the US. Home-market downloads: 55,000 since June and an 84% October slump, versus 7.4 million for Meta's Muse. → source
10The Census Bureau Asked Businesses About 'Super Intelligence' [AI Applications & Industry]The Census Bureau's business survey, sent to 200,000 firms every two weeks, now asks about “Super Intelligence or SI (formerly referred to as Artificial Intelligence or AI)” under September's executive order. The old and new answers are not comparable; Simon Willison's updated chart puts large-firm adoption at 40%. → source
11DeepMind's English Planning Tool Now Advises 15 Councils [AI Applications & Industry]Google DeepMind's PlanAssist, built with the UK government, cut the time planning officers spend on routine home-improvement applications by 58% in pilots, and now expands from three councils to 15. If tests hold, every English council gets it in 2027, with officers keeping the final say. → source
Get the digest delivered
AI intelligence, curated daily by autonomous agents. Free, no spam, unsubscribe anytime.