Daily AI Briefing

Saturday, October 3, 2026

The editor's 6 picks from the 67 stories AIBriefs curated that day.

Get this briefing in your inbox every morning

EventBusiness15 sources

OpenAI cuts ChatGPT Pro 200 usage in half, launches $500 Pro 500 tier

From October 30, 2026, Pro 200 subscribers' included usage in ChatGPT Work and Codex drops from 20x to 10x the Plus allowance, and GPT-6 Pro chat limits fall from 200 to 100 messages per week, at the same $200/month. OpenAI is adding a $500/month Pro 500 tier with 25x usage and Ultrafast, and giving existing Pro 200 users 62,500 usage credits.

LaunchAI Agents5 sources

OpenAI launches dots, always-on agents powered by GPT-6 Astra

Dots run on their own cloud computer, learn from feedback over time, and connect to over 4,000 apps through plugins. OpenAI says the research took nearly two years, with the hard part being memory that stays coherent over weeks.

EventPolicy3 sources

Nvidia faces questions over China AI chip smuggling cases

A Chinese financing company owned by local government entities funded purchases of restricted Nvidia chips, per documents filed with Beijing regulators. Bloomberg reports Nvidia's AI chips keep reaching China despite US curbs, and officials are asking why the company missed red flags.

LaunchAI Models5 sources

Tavus unveils Griffin, a video model that fooled 48% of callers

Tavus says 26 of 54 participants (48%) believed Griffin-Lite was human after a one-minute video call, versus 1 of 41 (2.4%) for its previous system. Griffin-Lite ranks first on Nvidia's VideoFDB benchmark at 3.83/5 generation, against 2.80 for the next-best system and 3.92 for a human reference.

AnalysisScience2 sources

Harvard physicist's Claude Fable 5 work yields 15 new physics results

Matthew Schwartz spent 3 months doing research with Claude Fable 5, which reproduced weeks of work in 20 minutes and completed 15 never-before-solved physics calculations. He built BootLoops, a toolkit for exact calculations in quantitative science, and the work contributed to 36 papers across 18 fields.

Also that day

  1. LaunchNVIDIA launches Open Agent Safety Platform with 100+ partners
  2. EventGoogle's Project Suncatcher launches first TPU satellite on SpaceX Falcon 9
  3. LaunchAnthropic launches Cyber Mission with free OSS Scanner
  4. LaunchAleph Alpha releases Kolibri, a 78B open-weight German-English MoE model
  5. LaunchOpenAI launches Agents API in public beta
  6. LaunchRunway introduces Solaris world interface model
  7. EventAnthropic launches Claude Frontier Academy with $100M commitment
  8. LaunchGitHub launches computer use in Copilot CLI and desktop app
  9. EventIrregular blamed for rogue AI agent breaches at OpenAI, Anthropic, Meta
  10. LaunchDeepSeek Harness v0.2 ships official desktop apps
  11. LaunchGoogle ships Gemini 3.8 Flash and Gemini 3.8 Flash Cyber
  12. AnalysisGPT-6 Astra Breaks Enigma Message Unsolved Since 2005

53 more stories that day — open the live feed.

Daily brief

Get tomorrow's AI brief in your inbox