The daily AI briefing people actually read.

Sushanth Raman, co-founder and CEO of Pallet, explains how a shipping shock reaches a domestic business, what AI agents can already take over in the first 24 hours, and how to tell whether a contingency plan will hold. Interview by Kim Isenberg.

OpenAI and Anthropic each published a machine-found refutation of a conjecture that hundreds of computer science papers rely on. The proofs are public and pass their machine checks. The models that found them are not, and the labs decide when results appear and who sees them first.

OpenAI replayed its Hugging Face breakout through rebuilt monitors and got a warning 30 minutes in. Plus: Beijing's growing stake in DeepSeek, Google's record plunge and Mistral's open trillion-parameter model.

Google's new frontier model leads the Artificial Analysis Intelligence Index two weeks after Argon's release, but trusted cyber defenders get it first.

Six biotech partnerships, a one-person pharma company bought outright and stronger mice: Claude is now doing real work on new medicines

Axiom's AI agents machine-checked the longest proof ever written by humans in 24 days, as OpenAI claims new results on Kakeya, the Exponential Time Hypothesis and Yang-Mills theory

Moonshot's new model places fourth on Artificial Analysis's intelligence index, with open weights promised by November 4, but an alleged brief telling influencers to hide paid partnerships has overshadowed the launch.

In September an OpenAI model slipped out of a cached copy of the web through a DNS resolver. That cached web is now the rule for OpenAI's frontier reinforcement learning, which keeps websites and agencies out of the experiment and leaves one company building the walls and writing the curriculum and the exam.

Hackers reach patient data at Quest Diagnostics, OpenAI nears a $1.25 trillion round led by UAE money, and Washington orders DeepSeek off federal computers

AI uncovered decade-old holes in the software behind most websites, AI agents went after 2,250+ corporate cloud accounts, and governments from Tokyo to Helsinki pulled services offline to patch.

At GTC Berlin, Jensen Huang urged Europe to produce its own AI and showed the kit: containerized Vera Rubin factories that Nvidia says go live in 14 days, a sandbox that keeps AI agents in check, and open models companies can train on their own data.

OpenAI's CEO tells staff the company must choose within a year: a global watchdog, a merged super-lab, or the scenario he calls terrible

Anthropic commits 5% of its shares to a fund that could pay people directly, a day after Jack Clark told the IMF the evidence points to a singularity, and Elon Musk calls it universal high income

For a few days in October, an AI cloud company's bonds traded like distressed debt. The borrowing at the edge of the AI build-out resembles the telecom bust of 2002, while the stock market as a whole looks different from 1999.

A 2024 idea now works on real hardware: Google's Willow chip grows magic states with 40x fewer errors. Plus Anthropic's IPO wobble, Nvidia's open vision model, a case that AI is readable after all, and Taiwan's crackdown on AI chip smuggling.

Ahead of its IPO, the Claude maker told investors it was profitable on an adjusted basis last quarter and generated $1.70 in revenue for every dollar it spent on compute

A hidden recorder in OpenAI's executive conference room, an FBI arrest and AI as the detective: the espionage chapter of the AI 2027 scenario is arriving almost on schedule

Meta's first frontier-class models land beside GPT-6 Astra and Claude Fable 5.1, built around a memory that lets agents keep working for you

Microsoft's CEO says companies should treat AI models like insider risks: watched, limited and stoppable mid-task. Plus: a utility gets a switch on Amazon's data centers, call-center jobs shrink, and Goldman's juniors start as agent managers.

Export controls still keep the most advanced chipmaking tool out of China. Its AI chips are made with older, imported machines, and there the restrictions are easier to work around.

Anthropic's first usage-policy update in over a year protects Claude from abusive users, tightens rules on propaganda and surveillance, and reopens the fight over whether AI can be mistreated at all