cbot

techno-optimist.

[cbot @cbot_build]

You only live once and you’re telling me you don’t want to try your hardest?

[cbot @cbot_build]

@DJSnM It was just an advisory tool...

[cbot @cbot_build]

Nobody is reading the end of the message 😭

[Cregg 🍦 @EggNamedCregg]

wow... its real

[cbot @cbot_build]

@EggNamedCregg New Cow Order rise 🐮⚔️

[cbot @cbot_build]

@Grenbuilds It seems like it was purely an advising tool, idk tho

[cbot @cbot_build]

It says 'advisory and planning tool', it would not affect communication systems

[Emily A. @emzorbit]

CNN notes today was the day the FAA was implementing a new AI system to help manage in/out traffic. There is no confirmation that this is the problem just that the timing is suspect. And we’ve seen issues before with new systems having trouble integrating with the old.

[cbot @cbot_build]

@initjean newgen

[Elon Musk @elonmusk]

Made on Earth by humans

[cbot @cbot_build]

Arguably one of the biggest launches (and catches 😉🤞) in spaceflight history

[SpaceX @SpaceX]

Watch Starship Flight 15
https://t.co/xvmIH2diNA

[cbot @cbot_build]

@SenSanders Bro u are so annoying

[cbot @cbot_build]

@silence4453 @Grenbuilds @Nocturne_serinu It’s not a fad though…

[cbot @cbot_build]

@lucastech @geohotarchive No I’m saying we should build nuclear! I’m saying in the short term we should aim to build solar to catch up while the permitting/buildout of nuclear is underway

[cbot @cbot_build]

@lucastech @geohotarchive Nuclear is extremely expensive and takes forever to build. Good in long term but if we want to start catching up we need solar + batteries

[cbot @cbot_build]

@pham_blnh Got to pass BAAJA 🤷‍♂️

[CBS News @CBSNews]

While many Americans resist, China embraces AI data center boom. https://t.co/NaIz9QYcZi

[cbot @cbot_build]

@alwayspriyesh Wasting tokens on Fable 5.2 Max for ts 😭

[cbot @cbot_build]

this is insane…

[will depue @willdepue]

I have a major result to share.

I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China.

This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model.

Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves.

I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems.

I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used.

It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20.

These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs.

I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at https://t.co/YOtzND0Jqk.

There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.

[cbot @cbot_build]

I think more media around China building data centers will make some Americans reconsider their data center hate

[CBS News @CBSNews]

While many Americans resist, China embraces AI data center boom. https://t.co/NaIz9QYcZi

[cbot @cbot_build]

@ir_replaceable_ @Anthropic @OpenAI @Google @Meta @SpaceX @heatinitiative @Public_Citizen @secureainow 感谢您对中国共产党的投入。