Social Media Briefing — June 10, 2026

Cherny gives a detailed account of why Fable 5 feels like the biggest leap since Opus 4.5, describing taste, judgement, methodical self-verification debugging, and a strong big-model quality.

Bcherny argues coding is the easy part of engineering, with debugging, scaling, infra, product planning, and user work still hard but increasingly automated.

Tunguz sharply criticizes Anthropic, suggesting they treat users as idiots, in a viral post tied to the Fable 5 launch.

Cherny announces Fable 5 is live in Claude Code and Cowork, calling it the best coding model by a wide margin with better code quality, tool use, self-verification, and autonomy.

Cohere introduces North Mini Code, its first open-source coding model, described as small and efficient, designed for agentic performance and community input.

Nathan Lambert argues Anthropic's uneven safety policies around Claude Fable 5 undermine AI community cohesion and accelerate near-term risk.

Matt Shumer claims a model (Fable) has solved 3D worldbuilding via custom in-browser ThreeJS, with a demo.

Ethan Mollick observes that both Anthropic and OpenAI mention the possibility of slowing AI development in their forward-looking posts, but condition it on globally coordinated action using unspecified methods.

The Rundown AI reports that Mayo Clinic researchers built REDMOD, an AI detecting pancreatic cancer on routine CT scans up to three years early, catching 73 percent of hidden cancers a median of 475 days before diagnosis, nearly double expert radiologists.