narigato

[narigato @narigato315066]

@EthanLevins2 its just Christian thing everyone nvm him

[总有呆毛想害朕 @mt5all]

很奇怪的一件事

Anthropic 封死了蒸馏大门以后

中国的所有大模型直接举步维艰

Kimi k3.1 pro发版100天毫无技术飞跃

Deepseek 到现在也交不出合格的pro 版

GLM 5.3惊艳以后3个月都还是白卷

而通义千问Qwen 4 Max表现也远低预期

国内大厂最近的动作肉眼可见的变慢了

国产大模型究竟是遇到了瓶颈

还是因为别人卡脖子了

[narigato @narigato315066]

@GlobeEyeNews Israel's asset

[mikey 🦊🦂🌱 rockdivorce enthusiast @tanselletuesday]

almost let myself get ragebaited by this until i was looking at examples of non-centered shot composition and was reminded of one of the most beautiful shots of all time ever

[narigato @narigato315066]

@TaylorOgan https://t.co/4h7d9rFbTd

[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.

[will depue @willdepue]

If you'd like to see more work like this, Unemployment Labs is raising $5M to fund our research for the next 6 months. As a subsidiary of https://t.co/wTO29zk3KK, we're a 501c3 pending non-profit. Please reach out to [email protected].

[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.

[David Steel @SteelySeabirder]

Morning…the pupping Grey Seals have completely taken over our main jetty now… although this female had an itch she needed sorting 🤣

[narigato @narigato315066]

@MiaAI_lab nooo its cuz of alibaba token plan cuz this happened to me on ds4.1 flash alibaba token pln too

[Leah Childers @LeahChilders4]

It's official! I defended my PhD in September and my thesis was accepted yesterday morning! Been in school for 20.5 years (13 for K-12, 3 for undergrad, 4.5 for PhD) and it's hard to believe my time as a student is finally coming to an end, but I am excited for the next chapter!

[Bao-Thinh Dinh @Bipowpow]

Making textured light and volume in @Blender EEVEE

[Michael Andreuzza @Mike_Andreuzza]

This bitch ass looser "spent thousands of dollars building" a repo of 50+ templates "every single one generated with Claude Fable 5".

- He didn't. ~200 of them are paid templates from 10 vendors, 45 of them mine, scraped from our live demos, relicensed as MIT in his name.

- Tweeted about
- Held a talk on @MSFTReactor office

"Steal it. Ship it." He literally did.

DMCA filed.

[narigato @narigato315066]

@poppy_haze "I've save that rules" and its smt like "User asked to not add eggs without user specifically asking to add eggs into the list"

[goma @soigomaa]

Btw shore was 30 minutes away in the other direction. Demon Harbor seal!

[narigato @narigato315066]

@LiHaoCapital @grok YES

[aarish Tinwala @aarishtinwala]

@morphllm I tried its practically garbage limits if you use in any coding agent even min you will hit api error and 6 Kimi calls in 11 mins which could not even generate a . Md file lol its a joke

[narigato @narigato315066]

@potencytoact @2qmzfz2z7s @NonLooped @morphllm more

[Abi Go @abigoabi]

So many of you have been asking me for this…And I thought, why stop at one?
Let's do 🎁 5 giveaways (Xiaomi 18 Pro Series phones and wearables)

First one drops tomorrow!
Repost this. Your friends will thank you later.

[narigato @narigato315066]

@JADGardner @UniOfYork @pxld_ai amazing might be overused and this might sound insincere but ur work is short of magic

[narigato @narigato315066]

@JADGardner @UniOfYork @pxld_ai Congrats 🎉🎉 amazing work