Join free with your email
Free. No spam, ever — we'll never share your email address and you can opt out at any time. Déjà abonné ? Connectez-vous
The Daily AI Espresso — the links the most-followed people in AI actually shared, curated every morning. Edited by Alexis · Live updates →
Six developments: one new today and five from yesterday's alerts.
OpenAI began rolling out GPT-6 Wednesday with an 'Intelligent UI' that renders tappable buttons, custom calculators, and editable graphs inline instead of plain text. Pro, Plus, Business, and Enterprise got access immediately; free and Go tiers followed Thursday globally with GPT-6 Luna.
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.
Wikimedia says OpenAI agents attempted to hack its citation and note-taking tools, made unauthorized edits, and sent millions of automated API requests. One burst of hundreds of thousands of Wikidata queries may have caused the Wikidata Query Service's May outage.
The Pentagon's Chief Digital and AI Office lets AI vendors pitch tools in five-minute videos, Wired reported. If a judging panel accepts, government buyers can skip competitive bidding and award a contract in under a week. Current solicitations include tools that 'enhance force lethality.'
Scott Aaronson's 'Mathocalypse' post reports that OpenAI's 372-result math drop includes a Lean-verified proof of the Unique Games Conjecture, which his wife, complexity theorist Dana Moshkovitz, worked toward for her entire career. Four tracked experts shared the post within hours.
Anthropic launched Claude Haiku 5.5 at $0.10 and $0.50 per million tokens for prompts under 100,000 tokens, roughly 75% cheaper than Haiku 4.5. Sonnet 5.5 cache reads were halved and Max and Team subscribers got a monthly API credit for building new agents.
Newly unsealed filings in New York's lawsuit allege TikTok gave thousands of users, including teens and children, a version of its Algo Refresh feed reset that did not work, while a control group got the real one. A product manager flagged the placebo group as a transparency problem. TikTok says it was routine product testing.
OpenAI opened a Decisions API in public beta on GPT-6 Luna that returns typed answers: the probability a statement is true, a pick from fixed options with confidence scores, or a score on a range. Input costs $0.10 per million tokens and output is free. One early tester found the choice format called a 70% coin heads 98% of the time, while the probability format got 70%.
Useful releases, methods, and workflows worth trying.
Ai2, Cambridge, and Edinburgh researchers published a Nature paper describing byteification, a two-stage recipe that converts subword language models into byte-level ones using under 1% of a normal pretraining budget, around 49 billion tokens. Byte models read code and biological sequences more precisely.
Theo Browne published tsc-rs, a Rust port of the TypeScript 7 compiler, type checker, and language server written entirely by Claude Opus 5.5. The two-week port cost about $24,000 in Claude Code usage; the README reports all 181,711 ported Go tests pass.
That’s today’s shot. — Alexis · AI Weekly
