


Field CTO @sixfeetup | Cloud Therapist | AWS Hero | Co-Host of @vBrownBag | Fearlessly Stupid™ 🤣 linktr.ee/mistwire


A little something for all the "Reinstate @JensenHuang!" apologists out there:
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.

Why pay for more AI than you need? 💸 Smaller models can deliver results at lower costs with more control. 🎙️ Six Feet Up Field CTO and AWS Hero @mistwire discusses local AI and smarter model choices on @codetocloud_io. 👉 sixfeetup.com/company/news/b…


PSA: You don’t get to refuse to support @NYCMayor, actively oppose @AbdulElSayed , take AIPAC $, oppose Medicare for All and be chicken shit about abolishing ICE… then turn around and act shocked that people don’t see you as leadership material. @SenSchumer @RepJeffries

AI is going to solve everything! Right? 🤣 Six Feet Up Field CTO @mistwire has seen a few tech revolutions come and go. 🔄 At @TechFieldDay, he tells @DemitasseNZ why hands-on expertise & separating hype from reality matter more. 🎥 His #CFD26 take 👇 youtube.com/watch?v=nYs5GfFn…


What happens when an AI pipeline has to process 300+ TB of audio across 500 GPUs? ⚡ @mistwire, AWS Hero & Six Feet Up Field CTO, will share what worked, what broke, and lessons learned at @AllThingsOpen (Oct. 19–20 in Raleigh, NC). #AI #OpenSource #AllThingsOpen
Chris Williams (@mistwire), Field CTO and AWS Hero for @sixfeetup, joins us at #AllThingsOpen to present "Scaling AI: Operational Lessons from 500 GPUs"! 2026.allthingsopen.org/sessi…

Hard problems have a way of finding the right people. 🫶 We’re thrilled to welcome @mistwire to Six Feet Up as our first Field CTO! 🚀 In @codetocloud_io, Chris talks about his career journey & why Six Feet Up was the right place for his next chapter.👇 youtu.be/D1u-LVIG6Z8?si=Y7BK…

Hear @mistwire, Global Developer Relations Manager at @HashiCorp, chat with me about everything from consulting to tech certifications to corporate espionage (sort of) on the latest @redmonk MonkCast. redmonk.com/videos/from-data…

PyPI (the Python Package Index) is the official repo for Python packages, which means they deal with a LOT supply chain security issues from malicious actors 🙀 Join us and PyPI Safety & Security Engineer & AWS Hero @mikefiedler Wed, May 7th @ 730pm CDT youtube.com/live/UYqlHuznLCI

Okta, Identity, & AI Agents: Why Vibe Coding is the Worst Thing Ever™ 😅 linkedin.com/pulse/okta-iden…

Boost your Azure automation skills! Join @MarkTinderholt to learn how to avoid the top 5 mistakes, and discover a free tool to simplify your workflow - building a solid foundation for success! #Azure #Automation #Terraform Wed, March 26 730pm Central: youtube.com/live/2pcRSKKpPP8…

JAWS-UG東京 IaC Night 〜入門から上級まで!AWSをコードで構築しよう〜 jawsug.connpass.com/event/34… #jawsug_tokyo 来週こちらのお手伝いを少しばかり。 @mistwire is back in Japan again!

イベントを公開しました! #jawsug_tokyo JAWS-UG東京 IaC Night 〜入門から上級まで!AWSをコードで構築しよう〜 jawsug.connpass.com/event/34…

At #reinvent this year I stopped by the incident.io booth & realized they were doing ALL of my incident response recommendations AND MORE 🤣. Interview with CEO Stephen Whitworth ⤵️ Incident playbook: bit.ly/3ZQ2HsY Demo: bit.ly/4gO4d4T