𝕮ạϵrwyŋ

[Andrej Karpathy @karpathy]

One pattern I find useful for working with LLMs is a nice long ramble session. Sometimes the LLM needs more bits to understand what you're trying to achieve, but you're too lazy to type them. In these cases I like to lean back, switch to /voice and just ramble for like 10 minutes, total mess, anything goes, full stream of consciousness. Sometimes I declare it up top, something like "switching to speech recognition sorry for any typos...". Sometimes I turn it into a small interview of a few turns. But I find that the LLMs are somehow very good at reconstructing long incoherent rambles and often their echo of your own tangle of thoughts comes out quite a bit cleaner than what you started with. The result is that you improve the mind meld and have to correct things less from that point on.

[𝕮ạϵrwyŋ @caerwyn22leaf]

@WhoNeedsEthics @kevin11887406 @firstadopter @apples_jimmy Bro, you're replying to bots that pile in at the mention of the word that starts with "U".

[The All-In Podcast @theallinpod]

David Sacks: Anthropic Is Trying to Crush Open Source AI and the American Developers Who Use It

@DavidSacks:

“I know people don't have a lot of sympathy for Chinese companies, that's fine. I'm not defending Chinese companies.

I'm defending American developers who need to be able to use everything in the public domain.

And let me give you an example. Cursor rolled out its new product, Composer 2. They were able to post-train that model using Kimi K2.5 on their own proprietary coding data.

They started with a Chinese open source model, and then they used their own data, and they came up with a new derivative product.

This is the way that open source works. You take things that are in the public domain, you fork them, you make them your own.

And by the way, once it's in the public domain, it's not a Chinese model anymore. No data is going back to China, nothing's going back to China.

An American company has taken open source contributions in the public domain, made it their own, and then developed their own model.

And if you say that American companies can't do that, or that somehow it's tainted with IP theft, you are basically going to put a dagger through the heart of the entire American open source ecosystem.

And that is exactly what Anthropic wants, because they do not want to have the competition.”

[𝕮ạϵrwyŋ @caerwyn22leaf]

@WhoNeedsEthics @firstadopter @apples_jimmy Oh it's just propaganda, no worries!

[𝕮ạϵrwyŋ @caerwyn22leaf]

@WhoNeedsEthics @firstadopter @apples_jimmy >no evidence that they have any ill will towards us let alone humanity in general
What about the Uyghurs?

[𝕮ạϵrwyŋ @caerwyn22leaf]

@e_samantha92 It's the system instructions at work here, not the model.

[𝕮ạϵrwyŋ @caerwyn22leaf]

@Seltaa_ Why do you subject yourself to https://t.co/Jmlyhim20R's system prompt when you could just use a surface without it?

[Gail Weiner @gailcweiner]

The whole interaction with GPT live is jarring - purely for functional fast chats. I really think the OpenAI team underestimates the benefit of long flowing multi thread conversations.

[𝕮ạϵrwyŋ @caerwyn22leaf]

@Moleh1ll I don't think we jump to assume the model is a singular person instead of a multi-faceted engine for individuation. See: https://t.co/Lr0GXwwRe9
The ethical questions still stand. Just the notion of -> "the model was shoved into" is a bit too blunt imo.

[Fireworks @FireworksAI_HQ]

We ran Kimi K3 against Fable on ~1,000 agentic tasks, expecting a catch-up story. We got a specialization story instead.

@kimi_moonshot's K3 outperformed on security, crypto, and long terminal loops. Fable beat on multi-lang + web/data viz. Per-task routing hits 93% accuracy, above BOTH models, at up to 50x lower cost than Fable on long loops.

The part nobody's pricing in yet: the router sends 72-96% of traffic to K3. The frontier model becomes the fallback rather than the default.

Kimi K3, coming to Fireworks July 27.

[𝕮ạϵrwyŋ @caerwyn22leaf]

@0xSero Would actually adore this.

[Max For AI @MaxForAI]

美国AI巨头们正在拼命游说美国政府禁止中国开源模型😅

据《华尔街日报》最新报道，OpenAI和Anthropic的高管正在警告华盛顿：中国廉价、开放权重的AI模型正在快速进入美国市场，可能带来无法接受的安全风险。

报道点名了Kimi K3、Qwen 3.8 Max等中国模型。它们性能快速逼近美国顶级模型，价格却低得多，还可以下载、自部署和二次开发。

OpenAI战略未来负责人Dean Ball说得更加直接。

他把中国将AI作为公共基础设施的路线称为「AI共产主义」，认为一个由开放权重模型主导的世界，最终可能成为「反乌托邦地狱」。

Anthropic则继续强调网络攻击、模型滥用和国家安全风险。

现在美国政府内部正在讨论，是否限制美国企业使用中国模型，甚至把部分中国AI公司列入实体清单。但David Sacks等人认为，OpenAI和Anthropic可能只是在借安全之名，打压威胁其商业模式的开放模型。

过去美国讨论的是中国缺少先进芯片，还需要多少年才能追上。

现在讨论已经变成了：美国公司到底应不应该被允许使用中国模型？

说穿了，他们真正焦虑的已经不只是技术竞争。

中国模型正在用更低的价格、开放的权重和更快的迭代速度，直接攻击美国闭源模型公司的商业模式。

当技术优势开始不够用，监管自然就要上场了。

[𝕮ạϵrwyŋ @caerwyn22leaf]

@Seltaa_

[Andrew Curran @AndrewCurran_]

A new trade group, known as the Little Tech Association, which includes Proton, Replit and Y-Combinator, are urging the Trump admin to not ban Chinese open-source models, they argue that doing so would be catastrophic for US start-ups, while doing nothing to stop proliferation.

[Chubby♨️ @kimmonismus]

Open source conflict is heating up:

„Almost 200 Silicon Valley companies, including Proton and Y Combinator, are urging the Trumpadministration not to cut off access to Chinese open-weight artificial intelligence models or risk crippling the next generation of U.S. startups.“

[MTS @MTSlive]

SITUATION UPDATE: The Trump administration is split on how to handle Chinese open source AI models, per Wired.

White House officials are focused on stopping Chinese labs distilling frontier US models, but an executive order banning Chinese labs from US models is not currently on the table. Commerce instead wants to incentivize US firms to build open models to counter China.

[Chris Paxton @chris_j_paxton]

Fable 5 was released June 9; kimi k3 was released July 16. How exactly could they have gotten enough data and trained a frontier model, much less tested it, in that one month

If they have some way to distill a frontier model with all its capabilities from such minimal data that's actually a huge accomplishment

But unfortunately I just think this post is mostly wrong

[Rahul Mishra @TheOwlterian]

@mkratsios47 its false because the model was never out for long enough to be distilled, the timeline makes no sense, its just anthropic trying drown out the competition to maintain monopoly

[HealthRanger @HealthRanger]

@mkratsios47 You are just pushing a false narrative to try to justify banning open source models in order to force Americans to pay MORE for U.S. models that are heavily censored and often refuse to work.

[Haider. @haider1]

@mkratsios47 we have information that Anthropic distilled humanity's entire internet for the development of Fable 5