📴 The Offline World: OpenAI Now Trains Its Frontier Models Inside a Copy of the Web

In Today's Issue: What OpenAI announced and left out, what earlier environments rewarded, the cached web as infrastructure and its weak points, the September 20 timeline against OpenAI's own ts, what a copy of the web can teach, how other labs build, and who checks the builder. Records as of October 24, 2026.

Wanna learn more? Link down below!

Here's the link: getsuperintel.com

AWS capacity for 2026 is sold out and 2027 is "largely sold out," Andy Jassy said on an a16z podcast released Oct 23. "A good chunk of '28 is sold out," he added. When a cloud has most of its capacity booked through 2027, AI demand is still running ahead of supply.

Rippling's engineers ran 57% of their coding work on Anthropic models in May and 16% on OpenAI. Now it's 52% OpenAI and 18% Anthropic, CEO Parker Conrad told The Information, because OpenAI came in cheaper. Replit moved most of its spend after a discount of more than 30%, per a itched in hours. When switching is that cheap, a discount holds a customer until the next model release.

Up to 70% of websites could be taken over through bugs that agentic attackers are already probing, CrowdStrike told the WSJ. Amazon says exploitation is underway. NSA and FBI told some firms that could not patch to go offline. Defenders are calling it Black October; the tempo looks structural.

400+ Latin American banks and finance firms got hit by CyberCartel running AI agents on Claude, Grok, and Moonshot, Straiker told The Information. Anthropic cut their Claude access after the report. Offense is shipping agent tooling faster than most banks can watch.

Sam Altman told OpenAI staff last week about three proposals for developing AI safely, per The Information: an IAEA-style body, merging all labs into one, or one firm starving everyone else of compute. He called that last option "terrible".

Still super confused how its possible that Grokbot got approval for EU but OpenAIs dot, Anthropics duet and Metas Muse not.

I mean, I take it. Grok bot is just insanely good. My daily always on agent(s) for almost every task. Its just so good and it gets better almost every day. Kudos spaceXSI.

🏁 Who ships the next frontier model first?

Hackers reach patient data at Quest Diagnostics, OpenAI nears a $1.25 trillion round led by UAE money, and Washington orders DeepSeek off federal computers

In Today’s Issue:
🧪 Hackers reach patient data at Quest Diagnostics
💰 OpenAI nears a $1.25 trillion round led by UAE p its M3.1 Flash weights
📱 Grok Bot learns to use your phone's apps
🐋 Washington orders DeepSeek off federal computers
✨ And more AI goodness…

Wanna read more? Link down below!

Here's the link: getsuperintel.com

Met Jensen again today. Seriously one of the nicest person there is. Editings of the Interview with him almost finished. Will be released in the next few days.

🏁 Who ships the next frontier model first?

Now I can finally share it: I’ve spent the past week traveling across Europe with NVIDIA, covering all their Build a Claw events on the ground.

The highlight was something I could barely have imagined: spending almost an hour talking with Jensen Huang himself about everything quantum computing.

After the birth of my son and marrying my wife, this was probably the best day of my life.

The full video and some teaser
is coming soon: Jensen Huang x Chubby in Stockholm.

(Ps: yes, it’s real, not AI generated. @nvidia)

AI uncovered decade-old holes in the software behind most websites, AI agents went after 2,250+ corporate cloud accounts, and governments from Tokyo to Helsinki pulled services offline to patch.

In Today’s Issue:
🚨 The night hacking hit machine speed
🧮 A mathematician hands AI laude Duet arrives, OpenAI moves onto Google Cloud
🧬 Turning an 80-year-old's blood into young brain cells
✨ And more AI goodness…

Read more, link down below!

Here's the link: getsuperintel.com