


healthcare cio, rhce and 少し日本語 Disclaimer: views are my own and not necessarily representative of my employer.



Ok, hear me out: multi-player software factory with Win95 aesthetics. Why just chat with your friends when you can build stuff? Is this anything? youtube.com/watch?v=7pY_jdve…


Lord grant me the confidence to post shit like this during a 12 hour global outage.
What's one Al habit beginners should build early on?

I think we all assumed this is true but pretty incredible if he was able to prove it with this method.
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.

Did they finally find a way to measure benchmaxxing? That would be amazing.
Nice DeepSWE score! Does it survive a lowercase filename? Introducing MaxxBench: 89 DeepSWE tasks with small changes to the instructions or starting code. Same engineering problem, slightly different requirement. Models clustered around 74% on DeepSWE end up 23 pp apart. Good at coding, or good at DeepSWE? 🧵

Is the bubble bursting or just a small post midterm correction? I guess the next week or two will be very telling.
BREAKING: Krispy Kreme officially outperforms NVIDIA over the past year.

This seems insane? This is like a 600B MoE US domestic open weight (soon!) model.
BREAKING: Inkling 2 Small by @thinkymachines takes 4th place overall on DesignArena and is 1st place among open weights models with an Elo of 1376! At $0.30/$1.00 per 1M input/output tokens, Inkling 2 Small is 15× cheaper on output than Kimi K3 by @Kimi_Moonshot. This model also places 1st on 2 of our subcategories including General Purpose and Landing Page prompts. Model weights to be released soon. A huge congrats to the @thinkymachines team on a top-5 placement on the leaderboard!

The Common Desktop Environment was created by a collaboration of Sun, HP, IBM, DEC, SCO, Fujitsu and Hitachi. Used on a selection of commercial UNIXs, it is now available as open-source software for the first time. sourceforge.net/projects/cde…

😬
JUST IN: Blackstone warns investors that Anthropic is likely to receive a high junk credit rating following its IPO.

Is Google back at the frontier? I've been using Argon and even that's been really impressive.
Gemini 4 Carbon is our next frontier model. It’s state-of-the-art across cybersecurity and long-horizon capabilities.

While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected. Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions. Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent. But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models? We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem. Read the full report: cohere.com/blog/concerning-b…

"I put my shoes in a garbage disposal, let's examine the results." Imagine the kind of people who watch this.
Not a test. Just breaking phones for clicks.


I don't know how I ended up in the middle of the Turkey vs China twitter but I'm here for it.
Hey çinliler köpeğime bakın 😂😂🤣🤣 köpeğim bile size saygı duymuyor 😂 Benim adım Muhammed Mustafa Sözen azıcık erkeklik duygusu olan çinli varsa karşıma çıksın 👊👊