Today’s front page moves here as the news changes. Links stay attached to their original reporting so the trail doesn’t disappear when the headline does.
The complaint—not a verdict—draws a line between clearly disclosed AI-assisted music and allegedly deceptive mass uploads. Australia’s copyright proposal, the industry’s unenforced safety consensus, and two big power stories join the desks.
The English beta reaches Apple’s major operating systems with personal context, onscreen awareness, and app actions. Ornith 9B gets a real home-hardware verdict, while Perplexity’s Windows agent misses both reference machines.
The new Streaming Integrity Initiative leads with stronger identity, rights, and content checks—but no blanket ban on AI-assisted music. Frontier slowdown pledges face an enforcement test, and a UK committee proposes a comprehensive AI law.
Dario Amodei’s call to pace capability development leads with a concrete evaluator-access pledge, two harder coordination proposals, and a clear label on his dramatic forecast. AI-assisted ballot matching and the mathematics community’s wider concerns join the desks.
The edition separates mathematical validation from disputed questions about collaboration, credit, and training data. Anthropic’s reports on attempted weapons misuse and boundary-crossing security tests join the three desks.
The distinction between an attempted request and a successful attack leads the edition. OpenAI’s Agents API, California’s child-chatbot law, and Universal’s planned AI remix platform join the three desks.
Formal congressional inquiries led—with every accusation labeled as an allegation. The edition also checked the Navier–Stokes claim, listened to Suno v6 outside its demo, and explained why Windows had a giant patch month.
The potentially historic math result led—with outside validation and the data-use dispute still unresolved. Suno v6 shipped, Meta launched Muse, and GLM-5.3 received a blunt home-hardware verdict.
The distinction between proving and formalizing leads the edition; a new Local Models to Try desk adds practical picks for everyday, creative, and visual work.
A signed warning about recursive self-improvement led the edition; OpenAI’s research-intern claim and a new publisher lawsuit received a reality check.
AT&T’s reported cost savings led a quiet Sunday edition; Perplexity’s hybrid Mac setup received an exact-match verdict for the 32 GB M4 mini.
K2 Horizon and llama.cpp 0.4.0 receive home-hardware verdicts, while Suno v6 remains only a teaser and Lyria 3.5 actually ships.
OpenAI’s AGI claim gets a reality check, PAIR gets a home-hardware verdict, and Microsoft makes a 64 GB bet on local AI.
The first reporting lands on OpenAI’s new frontier model, its cybersecurity restrictions, and its planned rollout to paid ChatGPT users.
Music publishers sue Anthropic over lyrics; artists challenge Suno over names and likenesses.