Gold Blade RPG

[Killian @KilianW146391]

K3 is genuinely a very strong model, and it's incredibly token efficient.

one-shot generated this 3D WebGL landing page with scroll-trigger animations.

every leaf is individually modeled.

Total cost: $2.37.

[HealthRanger @HealthRanger]

So much demand for Kimi-K3 that the company is forced to halt new subscriptions for the moment. That's because everybody is ditching Anthropic and OpenAI and switching to Kimi-K3.

It's just as good in nearly every way, yet a fraction of the cost.

I wonder how long it will take U.S. stock market investors to figure out we've just witnessed the second "DeepSeek moment." Yet in this case, it's Kimi (from Moonshot), not DeepSeek.

[Kevin S. Xu @kevinsxu]

Alibaba is a major investor of Moonshot. Kimi, in turn, uses a big chunk of AliCloud GPUs (oftentimes competing away Qwen team's resources)

10 days ago, Alibaba previewed quarterly earnings saying AliCloud is growing 40+%. This was ofc before K3

Could be a monster quarter for APAC's biggest cloud

[Kimi.ai @Kimi_Moonshot]

Kimi K3 has received far more love than we expected, and our GPUs are feeling it.

Over the past 48 hours, demand has pushed close to the limits of our current capacity. To protect the experience of existing subscribers, we're temporarily pausing new subscriptions and prioritizing compute for current members. Existing subscribed users are not affected.

We're adding capacity as fast as we can and will reopen new subscription spots in batches.

Going forward, we'll also split membership into two more focused plans: Kimi Membership for Kimi Web, App, and Work; and Kimi Code Membership for coding workflows. This will help us match compute more precisely and keep the experience stable.

Thank you for your patience and understanding!

[Max For AI @MaxForAI]

K3发布后，很多美国科技圈的人都在问，为什么 @Kimi_Moonshot 创始人杨植麟在美国完成学业后却没有留在美国？

答案可能有些反常识：不是他不想在美国，而是美国留不住他。

以杨植麟当时的履历，如果留在美国，他大概率会进入苹果、Google、Meta这样的公司，成为一名顶级科学家。

拿高薪、发论文、训练模型，在一个成熟的研究体系里不断向上走。

但他很难做出Kimi这样的模型。

他自己也在早期的访谈里表示过，在大厂里，科学家可以决定一项技术怎么做，却很难决定公司把多少算力、资金和人才押在这项技术上。

研究、产品和商业被切割在不同部门，个人能力再强，也只是庞大机器中的一个环节。

另一个重要的原因是如果当时留在美国创业，他同样未必能融到后来月之暗面获得的资金。

一个刚毕业、没有美国本土创业网络的中国研究者，很难让资本迅速把巨额资金和算力资源押在自己身上。

美国的钱更多，但有资格拿到这些钱的人，并没有想象中那么多。

回到中国后，情况反而不同。

杨植麟既有CMU和顶级AI研究的背景，又熟悉中国的人才、资本和市场。

他在这里不是大厂体系里的一名科学家，而是少数能够同时连接技术、融资、算力、团队和产品的人。

所以他不是因为H-1B、移民政策或者找不到工作才离开美国。

他完全可以留下，只是不愿意把自己最好的十年，变成一家大厂履历中的几行字。

如果杨植麟留在美国，世界上可能会多一位在硅谷大厂的AI科学家。

他选择回到中国，才有了月之暗面和Kimi。

当然，如果他留在美国很大可能会成为ICE（现在叫NICE）的目标，或者受到一些不该有的歧视。

[Polymarket @Polymarket]

JUST IN: Kimi AI temporarily pauses new subscriptions after demand for Kimi K3 pushes its computing capacity near its limits.

[AICodeKing @aicodeking]

Qwen 3.8 Max is actually a very good model. Outperforms all Opus models in my benchmarks. There are some issues in tool calling with some harnesses. But, the raw intelligence is just crazy.

[Young @Young_AGI]

Impressive！

[Emad @EMostaque]

Was thinking of trying different agents & harnesses like @cognition & @opencode with Kimi K3

What are people’s favourites these days

[Qwen @Alibaba_Qwen]

Qwen3.8 is launching and going open-weight soon!🌐

With a massive 2.4T parameters, this model is continuously evolving. We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5.

You don't have to wait to test it. Just now, the Qwen3.8-Max-Preview made its debut on Alibaba’s Token Plan, Qoder, and QoderWork. Be among the very first to try it out.

Can't wait to hear what you build. Stay tuned! 🚀 

Token Plan
international：https://t.co/YRvcGdB9Bv
China：https://t.co/PKMUNwUuRp

[Arnaud Bertrand @RnaudBertrand]

Funny how now that Chinese open weight models have closed the gap, Fable is suddenly not "too powerful and dangerous for humanity" anymore.

[tldraw @tldraw]

We're training neural nets in tldraw offline

[dex @dexhorthy]

today's fun ai coding tricking - when doing the high-level part of your plan, do an amazon-style "working backwards" approach - write the customer-facing blog post about the feature BEFORE you start building

this can surface so many edge cases and help you dial in on prioritizing the important parts, and just is much nicer to read (plus you can post it after you ship 🙂)

[Justin Schroeder @jpschroeder]

REALLY disappointed to see prominent @OpenAI folks lobbying like Anthropic. I told you:

👏 they 👏 are 👏 going 👏 to 👏 ban 👏 open source 👏 models.

[Rohan Paul @rohanpaul_ai]

For every $100 spent, Kimi completes 14.7 tasks versus Fable’s 5.3, giving you 2.8x more work per dollar.

on software engineering/DeepSWE tasks.

[Command Code @CommandCodeAI]

We tested Kimi K3 to build liquid metal text that wraps with your cursor in Three[.]js

- Used the /design command
- 3 prompts to get this effect
- Full session cost: $0.77

Open models deliver their best when used with the right harness.

[Kimi Developers @KimiDevs]

Dynamically Loaded Tools let you inject tools on demand during a conversation: start with only a few core tools, and insert additional tools into messages when the conversation actually needs them — reducing token usage and improving tool-selection accuracy at the same time. For the reasoning behind this design (lazy loading, tool registry) and combined practices.
👉 https://t.co/EcPhEljrBP

[Kimi Developers @KimiDevs]

When your application needs a large number of tools, declaring all of them up front in the top-level tools field of every request leads to Tool Definition Bloat: every request carries the descriptions and parameter schemas of all tools, driving up token usage, and the more candidate tools there are, the more likely the model picks the wrong tool or constructs invalid call arguments.

[Dillon Mulroy @dillon_mulroy]

what’s the best tooling to let agents use a browser rn? agents always seem to burn a ton of tokens using things like agent-browser esp using helium (@uwukko maybe you have some good guidance here?)

[dax @thdxr]

@dillon_mulroy @uwukko we use this: https://t.co/5LONchzLTE

will release properly soon