Join free with your email
Free. No spam, ever — we'll never share your email address and you can opt out at any time. Already a subscriber? Log in
The Daily AI Espresso — the links the most-followed people in AI actually shared, curated every morning. Edited by Alexis · Live updates →
12 developments: one new today and 11 from yesterday's alerts.
OpenAI says an upcoming cyber model autonomously found, validated and patched 948 Linux local privilege escalation bugs in a single pipeline run, at $1,213,430 in API credits, or about $1,279 per unique flaw.
Pick the topics, companies and people you track. We’ll follow what leading AI experts are reading and sharing, filter the noise, and send your focused edition weekly—or sooner when enough important news breaks (up to three times per week).
Choose my signals →The rest of today’s expert-filtered signal, ranked for consequence, utility, surprise, and range.
Cat Wu, Anthropic's product lead for Claude Code, has joined OpenAI to lead product management for Core Products, less than a week after she left Anthropic on October 30, Business Insider reports.
Anthropic paid the Associated Press for real-time, attributed election data that Claude surfaced throughout US midterms coverage, its first content licensing agreement with any digital news publisher, Adweek reports, citing two people familiar with the matter. Financial terms and duration were not disclosed.
OpenAI will pay the full electricity and required infrastructure costs at every data center it owns and operates, and stop signing nondisclosure agreements with public officials, the company said in a 'National Community Compact' published two days after voters backed restrictions in 40 of 41 US ballot measures.
Microsoft has declined an option to more than double its existing Texas AI capacity contract with Nscale, a deal worth tens of billions, per Financial Times reporting; the move comes after Microsoft's earlier withdrawal from Nscale's UK Essex project and ahead of a planned $35B Nscale IPO.
YMTC parent CCSH has paused a 33 billion yuan ($4.9B) Shanghai STAR listing that had been scheduled for late October after the CSRC asked banks to slow chip and AI IPOs, with the STAR50 chip-AI index off 40 percent from its July peak, Reuters reports.
Useful releases, methods, and workflows worth trying.
OpenAI shipped ChatGPT Images 2.6 on November 4 with native animated-GIF generation and persistent reference images across web, desktop and mobile, alongside two API models: GPT-Image-2.6 Flare for faster generation and Sunburst for higher quality.
A new arXiv paper on Hugging Face reports that RL-trained Qwen3 agents pick a better single-step action than the base model in only 42.8 percent of FrozenLake states and 46.8 percent of DeepCoder-MT states; the authors argue multi-turn RL concentrates in error recovery and long-horizon resilience rather than per-step quality.
Researcher Rikiya Takehi published hand-coded MLP weights that play Atari Breakout on GitHub, with layered notes describing what each neuron computes: four landing-position errors, two direction indicators, and a second layer that gates left or right paddle moves.
That’s today’s shot. — Alexis · AI Weekly