Artificial Intelligence thread

[meedicx]

Maybe the Chinese are just being pragmatic? Haven’t you thought that it might also be because Claude’s has the most advanced programming and debugging performance? Afterall, nobody will want to rely on something that they don’t think offers them the best value proposition for their work . So I’m not sure it’s western worshipping. lol. It’s just being pragmatic . So I don’t blame them at all.

At least Alibaba banned their workers from using Claude’s code for their work and distillation after Anthropic publicly called them out from doing so, which embarrassed Alibaba. lol Maybe anthropic should publicly called out all the major Chinese tech companies using their code , that might be the only way to stop them from doing so(at least publicly). lol

You guys are late to the party. There was a security warning on Claude Code posted by MIIT on July 9th. All SOE IT departments have banned it already and private enterprises that contract with SOEs probably need to remove it as well. SOEs are also major clients of domestic IT service firms who do custom integrations, so they would naturally need to stop using Claude Code to keep contracts.

Also there is a perception lag. In the past month after the GLM-5.2 release and now especially after Kimi K3, you can feel a huge vibe shift in the CN developer community and dev influencers on using domestic LLMs for coding.

 

[bsdnf]

You guys are late to the party. There was a security warning on Claude Code posted by MIIT on July 9th. All SOE IT departments have banned it already and private enterprises that contract with SOEs probably need to remove it as well. SOEs are also major clients of domestic IT service firms who do custom integrations, so they would naturally need to stop using Claude Code to keep contracts.

Also there is a perception lag. In the past month after the GLM-5.2 release and now especially after Kimi K3, you can feel a huge vibe shift in the CN developer community and dev influencers on using domestic LLMs for coding.

The variance in Chinese developers' understanding of AI is also very large; some are still using Doubao and Gemini, while others already have multiple Claude and Codex 20X accounts.

 

[tokenanalyst]

it's about time Chinese government crack down on Claude code usage. This Western worshipping needs to stop. I was talking to Rui ma recently and she mentioned to me that people in China would rather use Claude Code than any Chinese AI providers. Someone has to put a stop to this nonsense.

Is very difficult to quantify the use of Claude by just anecdotes because Ant. models are geoblocked in China so there is no real metrics of usage, And the situation is even more blurry in the agentic era because you can use Claude as the orchestrator for a project, DeepSeek Pro for the backend and DeepSeek Flash for the frontend. DeepSeek would generate more tokens even if Claude is the orchestrator.

 

[tamsen_ikard]

https://twitter.com/i/web/status/2080819778869219423

This is one point I've been wanting to making for a while and I'm glad David Sacks made it.

If you have the great AI in the history that can apparently do everything, why couldn't you have stopped distillation? Is it because you were too busy getting all that money from users in China?

it's about time Chinese government crack down on Claude code usage. This Western worshipping needs to stop. I was talking to Rui ma recently and she mentioned to me that people in China would rather use Claude Code than any Chinese AI providers. Someone has to put a stop to this nonsense.

Distillation is simply impossible to stop. You have stop serving the product completely if you want to stop distillation.

If they ban all Chinese IP from using claude, they can just use VPNs to mask their access.

If they want to stop any particular account from making too many prompts, they can just mask it by creating thousands of fake accounts.

Distillation is basically fair use, which cannot be stopped with any technical means.

Any public facing AI system can be distilled no matter what measure is taken to stop it. AI by its very nature is "open source", anyone can learn from another AI as a teacher using its output.

 

[Eventine]

You guys are late to the party. There was a security warning on Claude Code posted by MIIT on July 9th. All SOE IT departments have banned it already and private enterprises that contract with SOEs probably need to remove it as well. SOEs are also major clients of domestic IT service firms who do custom integrations, so they would naturally need to stop using Claude Code to keep contracts.

Also there is a perception lag. In the past month after the GLM-5.2 release and now especially after Kimi K3, you can feel a huge vibe shift in the CN developer community and dev influencers on using domestic LLMs for coding.

The historical issue with cracking down on Claude usage is that it will put you behind from a productivity perspective, since Claude was superior to any Chinese solution. At a national level, de-risking American AI suppliers has to be balanced against the competitiveness of your enterprises globally, since if countries that compete vs. China like South Korea, Japan, US, and Germany have superior AI, that could spell trouble for Chinese industries. Hence, for a while, China was tolerant of these "under ground" token sellers.

Now that the gap has been further closed and the compute capacity bottle neck is getting resolved, the government has more of a free hand and things will right themselves over time, presuming Chinese labs continue to keep up. The only way this changes is Open AI and/or Anthropic manages to widen the gap with their next releases (not likely, but we'll see).

 

[AI Scholar]

The historical issue with cracking down on Claude usage is that it will put you behind from a productivity perspective, since Claude was superior to any Chinese solution. At a national level, de-risking American AI suppliers has to be balanced against the competitiveness of your enterprises globally, since if countries that compete vs. China like South Korea, Japan, US, and Germany have superior AI, that could spell trouble for Chinese industries. Hence, for a while, China was tolerant of these "under ground" token sellers.

Now that the gap has been further closed and the compute capacity bottle neck is getting resolved, the government has more of a free hand and things will right themselves over time, presuming Chinese labs continue to keep up. The only way this changes is Open AI and/or Anthropic manages to widen the gap with their next releases (not likely, but we'll see).

Being a few weeks behind on an incremental performance bump (at a significantly higher price) is a reasonable tradeoff when the alternative is fighting Anthropic's censorship and silent nerfing that will quietly degrade our work without warning. Given the release cadence of Qwen, Kimi, and Zhipu, the gap to SOTA is a matter of weeks. Fable 5's final release was July 1, and Kimi/Qwen matched it within two weeks, and I don't expect that to widen, no matter what GPT-6 or Fable 5.1 bring. The cost advantage is substantial, especially now that Qwen 3.8 and Kimi K3 already cover the vast majority of use cases.

 

[henrik]

Distillation is simply impossible to stop. You have stop serving the product completely if you want to stop distillation.

If they ban all Chinese IP from using claude, they can just use VPNs to mask their access.

If they want to stop any particular account from making too many prompts, they can just mask it by creating thousands of fake accounts.

Distillation is basically fair use, which cannot be stopped with any technical means.

Any public facing AI system can be distilled no matter what measure is taken to stop it. AI by its very nature is "open source", anyone can learn from another AI as a teacher using its output.

It is the US models that are distilling superior Chinese AI models and copying new concepts from Chinese AI research papers.

 

[bsdnf]

Qwen 3.8 Max Preview is terrible.

The inference level wasn't set properly on the first day, resulting in an extremely clumsy model. The second day was way way better, showing potential to challenge Kimi K3

But its performance has been gradually declining each day since, completely contrary to the advertised daily evolution. It's now showing signs of being worse than Qwen 3.7 Max.

What is Qwen labs doing?

 

[Kalum Pupeter]

https://twitter.com/i/web/status/2080819778869219423

This is one point I've been wanting to making for a while and I'm glad David Sacks made it.

If you have the great AI in the history that can apparently do everything, why couldn't you have stopped distillation? Is it because you were too busy getting all that money from users in China?

it's about time Chinese government crack down on Claude code usage. This Western worshipping needs to stop. I was talking to Rui ma recently and she mentioned to me that people in China would rather use Claude Code than any Chinese AI providers. Someone has to put a stop to this nonsense.

White worship has been a problem in China since the early 20th century. It appears to have grown by leaps and bounds between the 1900s and the 1930s, and since then it has remained widespread. It was never meaningfully reversed or even substantially diminished by the country's leaders. They have had every means at their disposal to reverse or at least reduce it, given their extensive control over education, the media, and other forms of cultural propaganda and social messaging.

 

[tamsen_ikard]

It is the US models that are distilling superior Chinese AI models and copying new concepts from Chinese AI research papers.

Well its free training data, why won't they use it.

Anyone who is not distilling from others is foolish. If you wanna compete, you have to use every advantage you can get, which means you must distill.