AI Weekly Roundup — Oct 24, 2026

Model releases

Recursive Argues AI-Building-AI Is Worth the Risk

Startup Recursive makes the case that the benefits of using AI to build AI outweigh the risks. Whether the industry agrees is a different question.

Source: Wired AI

Kimi K3.1 Pro Lands on API, Weights Promised for November

Moonshot's Kimi K3.1 Pro is live on the Kimi API, not yet open weights, with the company promising full weights by November 4, 2026. A "Highspeed" variant runs 2-3x faster at twice the price, except for input token cache hits.

Source: Simon Willison

OpenAI Ships a Jev Competitor

OpenAI is launching its decision model broadly, taking direct aim at the original Jev. Whether it can actually compete is the open question.

Source: TechCrunch AI

Research

Microsoft's Diagnostic AI Beats Generalist Doctors on Hard Cases

A Science study found Microsoft's new AI diagnostic system outperformed generalist doctors on difficult cases. The catch, per the reporting, is that most of the gain appears to come from the underlying models rather than Microsoft's own system.

Source: TechCrunch AI

Industry & funding

OpenAI Cuts Frontier Models Off From the Live Internet

OpenAI will stop training frontier models on the live internet, citing significant risks. The move lands just weeks after one of its own engineers argued that approach wouldn't work.

Source: TechCrunch AI

Tools

Anthropic Opens a Waitlist for Claude Duet

Claude Duet, Anthropic's agent-focused app, is available via waitlist for Pro and Max plan subscribers. It's the company's latest signal that agents are where it's placing its chips.

Source: TechCrunch AI

Microsoft Says Its AI Is 2x Better Than Doctors at Diagnosis

Microsoft claims its new AI is more than twice as good as doctors at diagnosing diseases, a figure now being picked apart on Hacker News. Treat the headline number as the vendor's own until independent review catches up.

Source: Hacker News (AI)