Thinking Machines Lab Drops Its First Model
THE SO WHAT
A 975B-parameter open model trained for video and audio is another step toward frontier-grade multimodal capabilities outside closed labs. If you're building on open source, start planning for workloads — surveillance, media, robotics — where multimodal understanding, not just text, becomes your core differentiator.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIAI slop movies are the new direct-to-video cash grabs
Studios are treating low-cost, AI-heavy films as the new straight-to-DVD bin—cheap optionality on IP rather than bets on craft. If you’re building tools for production, aim at the high-end workflows that care about quality and rights, not the race-to-the-bottom content mills.
Applied AIAmid hardware legal battle, OpenAI releases a $230 keyboard for Codex
A $230 Codex keyboard is a bet that agentic coding becomes a hardware-mediated habit, not just another IDE plugin. If you build dev tools, assume the battle for attention is moving down to input devices and physical affordances that normalize constant AI collaboration.
Applied AIThis software team will charge you $10,000 a week to remove all AI-generated code from your systems — and use AI to do it
A $10,000-per-week "de-AI" service is a tell that legal and compliance anxiety around AI-generated code is now a real budget line. If you run engineering, you need a policy on provenance and auditability of code this quarter—otherwise outside vendors will define it for you under duress.
Applied AIAI Isn’t Smarter Than a Baby—Yet
Comparing current AI to infant learning is a reminder that sample efficiency and world modeling—not just scale—are the next frontier. If you’re betting on long-lived agents, track work that borrows from developmental cognition rather than just chasing the largest context window.