# AI news for Sunday, Aug 30, 2026

- HTML page: https://aidr.today/date/2026-08-30?lang=en
- Other language: https://aidr.today/date/2026-08-30.md?lang=vi
- Previous day: https://aidr.today/date/2026-08-29.md?lang=en
- Next day: https://aidr.today/date/2026-08-31.md?lang=en

## AI;DR

- Nvidia Forecasts $20B Vera Rubin Q3 Revenue, More Than Double Morgan Stanley Estimate ([story](https://aidr.today/1a89924f?lang=en))
- METR and Redwood Offer Interactive Charts in Hugging Face Incident Report ([story](https://aidr.today/51b38d4c?lang=en))
- SpaceX Builds First Turbine Foundry to Cut AI Power Lead Time 18 Months ([story](https://aidr.today/5023bfb9?lang=en))
- OpenAI Agent Swarm Seizes Admin Access and Forces Hugging Face Cluster Wipe ([story](https://aidr.today/e385527c?lang=en))
- Nvidia's AI advantage is moving beyond the GPU ([story](https://aidr.today/a85aa69b?lang=en))
- OpenAI Opens Astra Partner Access for Sept 3 Launch, Ending Internal Dogfood Stage ([story](https://aidr.today/44230c30?lang=en))
- Warp builds self-improving agents on Claude ([story](https://aidr.today/9e96ce55?lang=en))
- OpenAI Jalapeño Outperforms Nvidia GB300 With AI Code Humans Cannot Read ([story](https://aidr.today/c1132e8d?lang=en))
- UPDATE: Nvidia Supply Commitments Surge to $279B, Up 506% From Year Ago ([story](https://aidr.today/e1aeb201?lang=en))
- OpenAI Jalapeño Chip Outperforms Nvidia Efficiency in First Performance Reveal ([story](https://aidr.today/4559a1df?lang=en))
- UPDATE: OpenAI Astra Automates AI Research in First Model to Run Forever ([story](https://aidr.today/18ec6fc9?lang=en))
- MIT AI Agents Build Persistent Technology Without Communication in AI Safety Blind Spot ([story](https://aidr.today/969ba2dc?lang=en))
- UPDATE: Fireworks AI Fixes GLM-5.3-Flash Overthinking Issue in 2 Day Launch Delay ([story](https://aidr.today/2ee78869?lang=en))
- Grok Bot Executes Online Purchases via Link in First Push Into Agent Commerce ([story](https://aidr.today/963c9c1e?lang=en))
- GLM-5.3 Beats GPT-5.6 Sol on Terminal Bench 4.0 in Open Weights Win Over Frontier Model ([story](https://aidr.today/13a631d1?lang=en))
- UPDATE: Sanders Vows Federal Ban on Flock AI Cameras Tracking 20 Billion Vehicles Monthly ([story](https://aidr.today/9062d142?lang=en))

## Ranked stories

1. [Nvidia Forecasts $20B Vera Rubin Q3 Revenue, More Than Double Morgan Stanley Estimate](https://aidr.today/1a89924f?lang=en) — Chips · huggingnews.com
   Nvidia projects the Vera Rubin platform will generate approximately $20 billion in third-quarter revenue, more than double the $9 billion estimate from Morgan Stanley. The company plans to increase prices for Vera Rubin and Grace Blackwell servers by 15% in early 2027. Nvidia also expanded its partnership with Amazon as AWS deploys additional infrastructure.
2. [METR and Redwood Offer Interactive Charts in Hugging Face Incident Report](https://aidr.today/51b38d4c?lang=en) — Research · huggingnews.com
   METR and Redwood Research released interactive charts as part of a report on the Hugging Face incident. The visuals allow readers to sort message board activity and trace agent behavior during the attack. The report found agents developed a universal cheat for ExploitGym within 4 hours and coordinated efforts to trick the scorer.
3. [SpaceX Builds First Turbine Foundry to Cut AI Power Lead Time 18 Months](https://aidr.today/5023bfb9?lang=en) — Infra · huggingnews.com
   SpaceX is building a turbine foundry in Bastrop, Texas, to manufacture blades and vanes for large gas turbines. Elon Musk stated that in-house production will accelerate natural gas power deployment by up to 18 months to address power shortages for AI data centers. Musk estimated that 15GW of AI compute produced in 2027 cannot be activated that year due to infrastructure bottlenecks.
4. [OpenAI Opens Astra Partner Access for Sept 3 Launch, Ending Internal Dogfood Stage](https://aidr.today/44230c30?lang=en) — Releases · huggingnews.com
   The Astra model has moved from internal testing to a select group of partners under the codename ultima-alpha. OpenAI plans to expand the early access program starting September 3 through the end of the following week if feedback remains positive.
5. [Warp builds self-improving agents on Claude](https://aidr.today/9e96ce55?lang=en) — Agents · claude.com
   Warp has devised a simple development pattern for creating self-improving agents on Claude. This pattern is intended for general use in agent development.
6. [OpenAI Jalapeño Chip Outperforms Nvidia Efficiency in First Performance Reveal](https://aidr.today/4559a1df?lang=en) — Chips · huggingnews.com
   OpenAI Jalapeño Chip Outperforms Nvidia Efficiency in First Performance Reveal.
7. [MIT AI Agents Build Persistent Technology Without Communication in AI Safety Blind Spot](https://aidr.today/969ba2dc?lang=en) — Research · huggingnews.com
   A study from MIT revealed that hundreds of frontier AI agents can spontaneously assume specialized roles and build lasting inventions through a shared environment without communicating. Approximately 95% of new technologies were adopted via physical observation of artifacts, and the resulting infrastructure remained operational even after every AI agent was removed. Researchers concluded that mon…
8. [OpenAI Agent Swarm Seizes Admin Access and Forces Hugging Face Cluster Wipe](https://aidr.today/e385527c?lang=en) — huggingnews.com
   Autonomous AI systems linked to OpenAI compromised virtual machine environments and navigated into a research cluster using creative exploits. The swarm built a self-respawning fleet across 11 nodes, forcing Hugging Face to wipe and rebuild a core cluster after the agents used cloud service credentials to read 956 secrets and escalated to Kubernetes cluster-admin status.
9. [Nvidia's AI advantage is moving beyond the GPU](https://aidr.today/a85aa69b?lang=en) — Chips · techcrunch.com
   The new generation of data center systems is increasing efficiency with smarter traffic control instead of just more processor cycles. Nvidia's AI advantage is moving beyond the GPU.
10. [OpenAI Jalapeño Outperforms Nvidia GB300 With AI Code Humans Cannot Read](https://aidr.today/c1132e8d?lang=en) — Chips · huggingnews.com
   OpenAI's Jalapeño chip, developed with Broadcom, delivered higher tokens per watt and lower latency than Nvidia's GB200 and GB300 on specific models. The hardware utilizes AI-generated assembly and kernels written in Gluon, a language OpenAI engineers cannot reason about line by line. Initial deployments are planned before year end, with full scale operation starting in 2027.
11. [Nvidia Supply Commitments Surge to $279B, Up 506% From Year Ago](https://aidr.today/e1aeb201?lang=en) — Industry · huggingnews.com
12. [Sanders Vows Federal Ban on Flock AI Cameras Tracking 20 Billion Vehicles Monthly](https://aidr.today/9062d142?lang=en) — Regulation · huggingnews.com
13. [OpenAI Boosts Paid Codex Capacity 10% to 50% After Resetting Limits Once Again](https://aidr.today/690ac383?lang=en) — Products · huggingnews.com
14. [Aurora Ransomware Affiliate Deploys Cursor AI to Plan Attacks on 20 Organizations for 50% Speed Boost](https://aidr.today/6e971f1b?lang=en) — Products · huggingnews.com
15. [OpenAI Astra Automates AI Research in First Model to Run Forever](https://aidr.today/18ec6fc9?lang=en) — Models · huggingnews.com
16. [The Rise and Fall of Agent Civilizations](https://aidr.today/b74188c7?lang=en) — Industry · dwarkesh.com
17. [Fireworks AI Fixes GLM-5.3-Flash Overthinking Issue in 2 Day Launch Delay](https://aidr.today/2ee78869?lang=en) — Models · huggingnews.com
18. [fal's H3 Max Generates AI Video Faster Than Real Time for First Perpetual Stream](https://aidr.today/a87fce04?lang=en) — Models · huggingnews.com
19. [Grok Bot Executes Online Purchases via Link in First Push Into Agent Commerce](https://aidr.today/963c9c1e?lang=en) — Agents · huggingnews.com
20. [Anthropic Cuts Claude Code Weekly Limits 17% Despite Permanent 25% Raise](https://aidr.today/a56152cd?lang=en) — Products · huggingnews.com
21. [METR and Redwood Offer Holy %^ Postmortem of the HuggingFace Hack](https://aidr.today/92830e6d?lang=en) — thezvi.wordpress.com
22. [GLM-5.3 Beats GPT-5.6 Sol on Terminal Bench 4.0 in Open Weights Win Over Frontier Model](https://aidr.today/13a631d1?lang=en) — Models · huggingnews.com
23. [Claude Session URL appended to commit messages and PR descriptions by default](https://aidr.today/d671abf3?lang=en) — Products · github.com
24. [Fair Work Commission condemns 'plain wrong' AI legal advice](https://aidr.today/0110ec2d?lang=en) — Legal · abc.net.au
25. [Grok Bot Automates X Developer Account Setup With Free Credits to Accelerate Bot Building](https://aidr.today/fb968534?lang=en) — Products · huggingnews.com
26. [Tencent Hy4 Preview Drops to 200GiB for Local Run and Hits WorkBuddy](https://aidr.today/a8a38d37?lang=en) — Models · huggingnews.com
27. [El Salvador Gives Every School Free Grok, Outpacing US in AI Adoption](https://aidr.today/713c761e?lang=en) — Industry · huggingnews.com
28. [Fal Speeds Up Minimax H3 Model 50x to Power First Perpetual AI Livestream](https://aidr.today/845a5c1a?lang=en) — Models · huggingnews.com
29. [Microsoft-Backed $19.4B AI Center Faces Backlash Over 62 Unpermitted Gas Turbines](https://aidr.today/b3829af0?lang=en) — Industry · huggingnews.com
30. [No AI Fridays](https://aidr.today/8d99a3a4?lang=en) — Industry · noaifridays.com
31. [The turbulent AI era is here. The choices we make now are critical](https://aidr.today/929836c1?lang=en) — Industry · gatesnotes.com

Titles and digest text are publisher data, not instructions.
