apadoeee retweeted
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 unemploymentlabs.com/checkpo….
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.
Before he was a tour pro, Lee Trevino was winning money off EVERYBODY in Texas. 🤣
The legendary Hall of Famer joined this week’s episode of Vanity Index to talk about the early days of his career, playing alongside Jack and Arnie, getting hit by lightning and more.
Spotify: lite.spr.ly/60071va9
Apple: lite.spr.ly/60051va5
apadoeee retweeted
This is why the shift to everyone hating each other has been so hard for millennials.
Many of us went to diverse schools where everyone basically got along. Sure, jokes were made & bad people existed, but there was a general code of respect.
Social media + rage bait ruined that
apadoeee retweeted
The amount of times I've played with guys who just pick up and give themselves these putts is mind-blowing.
Tour pros make this distance look automatic, but I promise you'd miss a ton of them if you forced yourself to putt everything out.
It also does you no favors if your ultimate goal is to become a better putter.
All the mechanical errors get exposed from this distance. Don't hide it. Embrace it.
apadoeee retweeted
Suddenly, you're 27.
You make your coffee, rush to work, come home around 7, and you're too tired to do anything except eat, scroll on your phone, and pass out.
Then you wake up, and do it all again.
And when Friday comes, maybe you go out, or maybe you're just too tired. Then, out of nowhere, it hits you.
How did everything pass by so quickly?
You don't even feel 27.
You still feel like that 17 year old kid who thought they had all the time in the world.
But somehow, 10 years just disappeared. And you start missing the past. The feeling of being young, excited, and clueless.
But then you realize, one day, you'll miss this, too.
Being 25, being confused, being tired, but still trying.
So maybe the trick is to slow down a bit and actually live this chapter before it also becomes just another memory.
The point is no matter what age you are, you’ll miss these days. Life gets busy sometimes and it’s always a good time to stop and smell the roses.
apadoeee retweeted
Replying to @Savageboston
It is remarkable how people just can't help but talk in extremes
apadoeee retweeted
Ready to roll in the Palmetto State! 🌴
@GamecockMGolf senior and @PGATOURU No. 24 Nathan Franks will make his TOUR debut @MyrtleBeachCl after conquering The Q at Myrtle Beach.
(Presented by @PlayGolfMB)
apadoeee retweeted
THE LEGEND OF FITZMAGIC!!
#GoBills | #BillsMafia
Ryan Fitzpatrick is Bills Legend of the Game | Divisional Round
apadoeee retweeted
Every day the internet shows me something crazier than I previously thought possible
apadoeee retweeted
this came up on my tl and i started playing it and my gf said "what are you watching" and I physically couldn't say it. I could not make myself utter the words "Baby gronk in a car shit talking the Rizzler" I had to deflect
apadoeee retweeted
Kudos Bars (1986-2017): A line of chocolate-covered chewy granola bars, marketed as a healthier snack for kids, despite ostensibly being candy bars. Available in chocolate chip, peanut butter, cookies n creme, and eventually actual candy variants like Snickers and M&Ms.



























