alexusadays

QA Engineer that talks about testing, AI and tech news

[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.

[alexusadays @alexusadays]

@elonmusk @ChrisJBakke You are Elon Musk you really don’t need to work on it 😂 your name is your pitch

[FAKEY FAKEY ✟ @4winds_]

New job interview rule:
If the interviewer is AI, you are legally required to bring an AI interpreter.

Do not go in raw. You will get cooked.

Watch what happens when the candidate has backup.

#AI #JobInterview #Hiring

[alexusadays @alexusadays]

@elonmusk Haha yes, just recently made video with this quote https://t.co/eWUGzcrPA3

[Claude @claudeai]

Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges.

They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.

[alexusadays @alexusadays]

@MilanGlow Dots

[alexusadays @alexusadays]

@OverlyTrev Well if you have proper planning and testing you don’t need a meeting on Sunday 🤷‍♂️

[alexusadays @alexusadays]

@melvindvivas Some actual numbers?

[alexusadays @alexusadays]

@mdamore9 @poteto @bot People have been complaining for six months or even longer. Where have you been?

[alexusadays @alexusadays]

@elonmusk @paranoidream Dude search grok bot super heavy people complained like for half a year now

[alexusadays @alexusadays]

@elonmusk @santiagomed @bot Yeah I don’t, stopped using grok bot since I was paying $300 a months and could not drive any video production with grok bot + grok imagine.

[Anthropic @AnthropicAI]

Memory is what allows Claude to learn from experience. In a new Engineering Blog post, we share how “dreaming” helps Claude organize what it’s learned, correct its mistakes, and find what it needs more easily the next time. https://t.co/uUEN4bNkr9

[alexusadays @alexusadays]

@NVIDIAGeForce #RTXON I want it!

[alexusadays @alexusadays]

@elonmusk until the next time than

[alexusadays @alexusadays]

🚨 BREAKING: Massive amounts of water rushing through the Tempe Town Lake dam in Arizona right now! 🌊

Following heavy storms and reservoir releases, the Salt River is surging, with a flood warning in effect downstream.

📍 Tempe, Arizona — November 2, 2026

#Tempe #Arizona #Flooding #SaltRiver

[Brandon Wurl @bdubs_14]

Landing in Phoenix and I’ve never seen the salt river in the city so full from bank to bank 😱

[Jeremy Ross @JeremyRoss77]

It's starting to happen, the end of the drought in the American SW and the partial refilling of the lakes, primarily Lake Powell. The super El Nino is putting up numbers here.... can we see Lake Powell at 3580' by next spring? 40 feet is a very doable goal. Everything in the blue circle will be drought free by spring of next year. This map is 4 days old, and will have even less drought colors after this last batch of rain.
#ElNino

[alexusadays @alexusadays]

I wanted to introduce my seven-year-old daughter to Greek mythology. We looked at different books, but couldn't find one that felt right for us. So I decided to create one myself, with AI helping along the way 😂

What started as a book for my daughter became a project I really enjoyed building.
I used AI to help draft the stories, create illustrations, and review the pages. I also created a harness for the book: a workflow to keep the stories, artwork, revisions, and reviews organized in one place.

The Minotaur page attached is a little peek inside.

For me, this is a positive, personal use of AI: helping turn an idea for my daughter into a book I can share with other families.

The result is Myths of Ancient Greece for Kids: Gods, Heroes, Monsters, and Legendary Adventures.

If you'd like to take a look:
https://t.co/aTHY9jWMP1

#AI #GreekMythology #ChildrensBooks

[alexusadays @alexusadays]

@tetsuoai https://t.co/2FO4zms4Wl I made xSI future vision :) not getting much interaction or visibility unfortunately

[alexusadays @alexusadays]

The most entertaining outcome, especially if ironic, is most likely.

What if this is everyday life in 2042?

I made an anime short inspired by Elon Musk's vision of the future: robot cafés, Starbase, a Moon base and life on Mars.

Animation: Grok Imagine. Music: ElevenLabs.

[Mike Nichols @MistrDeplorable]

@alexandr_wang said @Muse can make you $1,000. My receipt: Muse switched me from a $2,123 GEICO renewal to $1,536 at Progressive with BETTER COVERAGE ($1,174/yr saved), then Muse CALLED GEICO to cancel it for me. Then when GEICO asked for me, Muse called me, merged the call, and I was on the phone ~60 seconds to verify my identity and give my reason for cancellation. #MuseMoneyChallenge