Anthropic defends its IPO math and its bug-hunting model while AI executives face a New York City Council under oath. Plus Stripe wires agentic checkout into Constructor, OpenAI puts ads next to ChatGPT images, Gergely Orosz surveys a tech industry rebuilt around coding agents, Noah Smith counters that the AI boom is paying real wages, Simon Willison stress-tests a small model's arithmetic, and a wildcard: Stanford finds a way to regrow knee cartilage.

CNN · Oct 5

Reports Anthropic is proceeding toward a roughly $2 trillion IPO despite market jitters and its own safety warnings.

Cris Tolomia, Quartz · Oct 5

Anthropic, OpenAI, Google, and Meta leaders testified under oath to the full NYC Council on catastrophic AI risk.

Jessica Lyons, The Register · Oct 3

Anthropic's Mythos model chained a cryptographic flaw in Rejetto HTTP File Server that attackers exploited within a day.

PYMNTS · Oct 5

Constructor added Stripe-powered checkout inside its AI shopping agents, letting purchases finish within one conversation.

Sarah Perez, TechCrunch · Oct 5

OpenAI will test labeled visual ads next to ChatGPT image-generation results for an initial US advertiser group.

Gergely Orosz, The Pragmatic Engineer · Oct 6

Surveys how coding agents, parallel-agent workflows, and agent-authored PRs have reshaped software engineering this year.

Russell Brandom, TechCrunch · Oct 6

Anthropic expanded its startup program with free Claude Team seats, API credits, and Claude Marketplace access.

Noah Smith, Noahpinion · Oct 5

Argues data-center investment is lifting construction wages, local tax revenue, and GDP growth for ordinary Americans.

Simon Willison · Oct 4

Benchmarks a local Qwen3.8 model adding numbers and writing results in English words, with accuracy dropping on larger digits.