TechCrunch Top 3

OpenAI's feud with mathematicians is only escalating: Twenty-five leading mathematicians signed an open letter arguing that AI labs are threatening their intellectual work. Read More

Matt Mullenweg tells (trolls?) Automattic staff, saying he's back in control after CEO ouster: In a Slack message seen by TechCrunch, Matt Mullenweg told Automattic employees he’s back in control of the company, days after its board put him on leave. Automattic has not yet confirmed the apparent reversal. Read More

Khosla Ventures is opening a New York office this fall — its first outpost outside Sand Hill Road: "It's actually allegedly being built out now," said Rabois, who has clearly dealt with a missed construction timeline or two. Read More

Afternoon Must-Reads

Jensen Huang explains why Nvidia will grow an astounding 70% next year: Nvidia has its finger in every pie, and sees another year of plenty in its future, Jensen Huang says. But, he insists, its deals are not circular. Read More

Nscale adds former OpenAI exec Fidji Simo to its board ahead of potential IPO: The No. 2 exec at OpenAI also led Instacart through its IPO in 2023. Read More

Scammers target hundreds of thousands of crypto owners after Trezor confirms data breach of email provider: This is the second data breach affecting a company that hardware crypto wallet maker Trezor relies on. Read More

Roblox is making it easier to build games with AI — and play them outside Roblox: At its annual Roblox Developer Conference (RDC), the company announced several new features, including new game-creation tools, expanded NPC capabilities, and the ability to make games available across platforms, including the web. Read More

Y Combinator's Garry Tan wants U.S. open-weight AI labs to 'distill' frontier models, too: Tan argues that frontier models themselves trained on public human knowledge so access to capable AI should be "a form of public good." Read More

A message from ThoughtSpot

Is your AI giving confident answers or correct ones?

AI agents sound sure of themselves even when they're wrong. The fix isn't a better model — it's a governed trust layer underneath it. See how engineering and data teams at Roche, Snowflake, and dbt built theirs in ThoughtSpot's free 3-day virtual series, September 22–24.

Last but Not Least

Kimi-maker Moonshot AI targets $2B in annual revenue: While K3's usage figures have declined slightly in recent months, OpenRouter data currently shows as many as 300 billion tokens being generated each day by K3 models on the system. Read More