Waitlisted @cosign users can now browse existing profiles, Wikipedia rabbit hole as far as they want, and suggest edits where they see mistakes. You can check out https://t.co/WIq4Qa8xIL and hop around a bit!
Waitlisted users can also contribute to community cosign lists.
Users with access should now see a new, simpler home feed excperience geared around cosigns and forming new connections.
Our private beta is a way to allow us to iterate and really get the product right before allowing for mass consumership, we are working hard every day at nailing this for all of you!
The only mentions of China from NVIDIA's most recent CFO commentary (https://t.co/QHy6kuGOFo)
> Shipments of Data Center Hopper products to China during the quarter were less than 1% of Data Center revenue.
> Revenue is expected to be $108.0 billion, plus or minus 2%. We are not assuming any Data Center compute revenue from China in our outlook
How much did the CFO who's now the interim CEO know about their dealings with China?
- The latest Chinese models used prohibited chips (see quote tweet)
- Supermicro and Megaspeed together account for $10 billion, a non-negligible amount of money ($2.5 billion with Supermicro and $7.5 billion with Megaspeed)
- The official statements are usually worded as NVIDIA not shipping _to_ China or NVIDIA not doing business with companies that China has a stake in (Megaspeed had a shell company with a liability, not equity); arguably legalese gymnastics
Was everyone involved blissfully optimistic about a world built on one company's tech stack? Or, were few "in the know" with a lot of people blindsided by recent announcements? Or, perhaps, there are more people than just Jensen who supported doing business with China 'as usual'
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.
The only mentions of China in their latest CFO commentary (https://t.co/QHy6kuGOFo)
"Shipments of Data Center Hopper products to China during the quarter were less than 1% of Data Center revenue."
"Revenue is expected to be $108.0 billion, plus or minus 2%. We are not assuming any Data Center compute revenue from China in our outlook."
@lulumeservey There is another possibility, Lulu, one that dominates my group chats.
Jensen is sleeping with all his customers. When they say the interactions were not unusual or unexpected, they mean he is a sex machine on a rampage, seducing thousands of AI hardware execs with a glance.
Monorepos eventually turn into a salad of services in the codebase. Microservices/serverless eventually accumulate into an effective monorepo
Pick your poison
@lulumeservey There is another possibility, Lulu, one that dominates my group chats.
Jensen is sleeping with all his customers. When they say the interactions were not unusual or unexpected, they mean he is a sex machine on a rampage, seducing thousands of AI hardware execs with a glance.
Over the course of free software (free like speech not beer) to open source (Open Source Initiative) to the state it's in today, there was always a convenience to stuff that's available online. However, I suspect the real value is the community/network
While this may be a Tower of Babel story (we got AI that can code everyone's ideas but every dependency gets rewritten), we have specs like SAML or YAML w/ implementations in different languages.
Development is cheaper but developed stuff is not; I'd predict more OSS to exist
What’s going to happen to open source?
A prediction:
Dependencies are over.
The same Brotli vulnerability impacted .NET, Node, Bun, Envoy and more. Dependencies are too big of a target.
When rewriting a dependency is one prompt away, why take the risk?