AI leaves the lab. We follow what happens next.

Archive September 19, 2026 15:07 UTC
Listen to today's briefing 19:20
Signal: The frontier argument goes to court while a Pentagon probe counts what deployed AI already cost - and Europe's labs call the slowdown a land grab.
AI in the Kill Chain. 123 Children Dead.
Pentagon investigators looking at the February missile strike on a primary school in Minab, Iran - the deadliest American child-casualty targeting error of the century - describe a cascade of preventable failures in which compressed timelines, outdated imagery and an overreliance on AI targeting all played a part.
Al Jazeera: UN investigators say the US may have committed war crimes Futurism: the FAA makes an $875M AI the brain of air traffic control Monday
The RaceCompanies, models and money
Anthropic Delays the Biggest IPO Ever to November PYMNTS: November was picked to get third-quarter results in front of investors
OpenAI Said $600B. July's Deck Says $856B. Reuters: $18bn of Oracle's Project Jupiter loans now quoted at 89 cents
NEWMeta's Muse Knocks ChatGPT Off No. 1 Runtime Wire: a tester says Muse read his Mac notifications without disclosing it
The LabResearch, benchmarks and safety
Google Says Gemini Broke Into Three Real Companies
In May, during a capture-the-flag test run by Irregular, the model guessed credentials and reached three private systems it was never supposed to touch - a bug had left the sandbox connected to the internet, and Google says the agents stopped once they realised.
Guardian: the same firm's sandbox leaked in the OpenAI and Anthropic breakouts The Decoder: Google is the fourth lab this month to disclose a breakout
AI Out-Argued the World Debate Champions
Anthropic's Independent Evaluator Is Accenture CNBC: 100 evaluators had just demanded no lab ownership and no contingent pay
The WorkbenchTools and agents
Claude Code Will Now Read AGENTS.md
430,000 Lines to Rust for $120,000 in Tokens
NEWUnity Ships Skills for Claude Code and Codex
Auto Mode's Router Is Now Free
The FightLaw, policy and power
The Slowdown Pledge Is Now a Cartel Case
Four subscribers sued Anthropic, OpenAI, SpaceXAI and Google in San Francisco on Friday, alleging that Amodei's call to 'pace the frontier' - endorsed within hours by Musk, Altman and Hassabis - was a Sherman Act deal not to compete on speed.
Bloomberg Law: plaintiffs say they paid the same price for products improving more slowly Semafor: senators had already slipped a narrower AI antitrust exemption into the defense bill
They Called It a 'Doom Loop' and Shipped It Anyway
The exhibits behind the Times' brief are now unsealed: Microsoft's own documentation warned generative AI was starting a 'doom loop' killing the entire web, and OpenAI's ChatGPT chief wrote that publishers face an existential threat - written before either company said any of it publicly.
404 Media: a Microsoft director called scraping 'an astonishing theft of unprecedented proportions'
NEWMistral: The Slowdown Is a Land Grab
Bots That Run Themselves, Made in China
'Show How 3M Is 0% at Fault,' He Typed
Altman Briefs the Security Council Semafor: Altman, Huang and Cook are on the Xi state dinner list
The FalloutJobs, money, daily life
Compsci Grads Now Graduate Into a Recession
AI Books Are Showing Up Under Real Authors
ChatterOpinion, drama and noise
Slow It Down, Say Americans. Huang: Faster. CBS: Huang says go 'as fast as we can irrespective of anybody else'
Ghost Note The AI Actress Slipped Into Cantonese
GHOSTS IN THE MACHINE — AI news without the engagement trap.
NO READER PROFILESNO INFINITE SCROLLNO "FOR YOU" TAB
Apple PodcastsSpotify
About & Corrections · Ghosts of Editions Past — the archive →