The latest AI news, updated throughout the day — model releases, policy shifts, product launches, and research worth your time.
A stealth reasoning model appeared free on OpenRouter this week, described only as built for coding and sustained agentic work. Stripe's CEO called it "very impressive"; industry watchers are busy guessing who's behind it.
The company now backs amendments including training-time monitoring of frontier models and stronger cybersecurity requirements throughout development — a notable shift after earlier opposition.
Developers can now manage agents running across laptops and servers from any browser — multi-instance sessions with full file and credential access, plus push notifications when an agent needs input.
Dozens of fixes across folders, upscaling, gallery performance, and settings persistence land just two weeks after the redesign draft. The team warns new bugs may have been introduced.
Since its July launch, users have flagged over a million posts; the platform says views of classified slop content are down 40%. Authors now get notified when members flag their content as AI-generated.
A limited-time price drop on its flagship model's API and credit pricing aims to pull developers back into the ecosystem amid intensifying competition.
Convert standard-dynamic-range footage to 16-bit HDR in ProRes and EXR formats — works with any existing clip or prior Runway output up to 30 seconds.
xAI broadened availability of its autonomous agent platform across several partner plans, positioning it as a digital teammate that works across apps and inboxes.
The Defender Advantage Fund will provide credits to organizations securing open-source software — patching live vulnerabilities and automating scans top the priority list.
Our video team compresses the week's biggest stories into one watchable briefing — releases, funding rounds, and the demos everyone is arguing about.