Desk
Coverage tied to Anthropic and Claude.
Anthropic's consumer agent shows its strongest use is catching what falls between email, Slack and docs, and its weakest is pretending to understand tone or do the creative work.
TSMC buying into its own biggest indirect customers is a confidence signal for the AI trade, and another loop in a financing web that is getting harder to read.
New York's RAISE Act now has a registration portal, a regulator and a January deadline, and the 72-hour incident rule is the piece that will tell us whether state AI oversight has teeth.
Seven founders selling identical slices and holding control only as a bloc is an unusual design for a trillion-dollar company, and it shifts the governance question from one founder to a group.
An AI-assisted proof of an 18-year-old quantum conjecture is a real scientific moment, and PSI's harder test is turning results like it into physics that works outside a proof checker.
Claude Meetings puts an AI in the call as a named participant doing work in real time, and the guardrails Anthropic chose tell us it knows how sensitive that seat is.
Anthropic's office-suite push is really a bet that the next wave of AI adoption is people who will never open a terminal, made under heavy pre-IPO pressure to prove it.
Anthropic is asking public markets for up to $1.404 trillion while handing new shareholders almost no say, and the bet only works if its revenue curve keeps outrunning its staggering losses.
Jensen Huang skipping a few conference stages sounds minor, but it shows how much of the AI industry's public story has been routed through one man.
Anthropic's new Mythos and Fable models are cheaper, faster and reportedly doing real science, and they're also getting better at noticing when they're being tested.
A convenience click in a popular coding agent has been quietly writing passwords and keys into files that end up public, and the bug reports warning about it were closed without a fix.
Anthropic built its business on renting compute. Spending billions on gas turbines tells us the binding constraint on AI is no longer chips or capital but electricity, and that carries real costs.