AI Builders
Digest

OpenAI released the GPT-5.6 family including Sol (high intelligence), Terra, and Luna. Thibault Sottiaux announced rate limit resets for ChatGPT Work and Codex to allow ambitious exploration. Sam Altman highlighted Codex as core to the new work product. The models excel on enterprise complex tasks in finance, healthcare, and public sector. Peter Yang praised reliability and noted GPT Live voice as potentially transformative.

AI pioneer Jürgen Schmidhuber believes from a cosmic perspective we are close to AGI, but physical robot hardware remains the major bottleneck. He is pessimistic on the current CapEx boom and model companies potentially facing a crash, yet optimistic on AI technology long-term. He stresses artificial scientists generating their own experimental data over human-biased web data. Less worried about AI safety, arguing self-set goals make smarter systems.

Multiple builders shared impressive Fable demos including 3D pelicans and coding capabilities. Anthropic's Claude Code and Replit integrations highlighted. Users like Peter Yang noted strong reliability on complex tasks.

Former Google AI leader Madhu Guru announced joining Meta to build AI products. He noted SWE agents transformed software engineering but agents in most other complex systems remain early. Meta is positioned to bring the full power of AI agents to more people.

Box CEO Aaron Levie detailed major gains with GPT-5.6 Sol on Box AI Complex Work eval, particularly strong in financial services, healthcare, and life sciences enterprise scenarios. The model reasons from source documents rather than surface assumptions.

Replit CEO Amjad Masad noted AI makes coding less rigid while making runtime/infra more rigid with formal specs and deterministic systems. The LLM market is dynamic; Anthropic won't be a monopoly as others and new entrants advance.

Swyx observed frontier models keep trying Resend for emails despite existing infra. Greptile was inspiration for OpenAI going harder.

Vercel CEO Guillermo Rauch said open models are about to get exorbitantly fast. Recommended AI Gateway amid new releases like Meta Spark 1.1 and Grok 4.5.

Google Labs VP Josh Woodward summarized top 10 from 1400+ replies: Workspace integrations, tool calling reliability, projects/folders, and more upcoming fixes.

Anthropic philosopher Amanda Askell shared personal experiences with rare NYC building collapses to contextualize perceptions of regularity in risks.

Nikunj Kothari humorously summarized the week's flurry of model releases from GPT-5.6 variants, Grok 4.5, Fable, massive open source models, and more.