cloud cat

🖥在职AI产品经理
🏄🏻‍♂️专注人工智能（AI）科技分享，做有趣实用的内容

[cloud cat @cloudcat2333]

@DeepSeekHarness Congratulations！

[cloud cat @cloudcat2333]

@MiniMaxAgent and H3

[cloud cat @cloudcat2333]

@DarioAmodei So biological discovery is starting to become “large-scale pattern recognition + a rapid experimental closed loop.” Math has already walked this exponential curve; biology is probably on the same path.

[cloud cat @cloudcat2333]

@MiniMax_AI good 懂得都懂😃

[cloud cat @cloudcat2333]

就在zcode出事时，mcode开源了，懂了懂了

[cloud cat @cloudcat2333]

@MiniMaxAgent 懂你意思😘

[cloud cat @cloudcat2333]

@jietang We need a fast, multimodal, and cost-effective model for production use

[cloud cat @cloudcat2333]

@wyl2035 i need

[cloud cat @cloudcat2333]

@deepseek_ai Can we see v4.1 Pro by the end of this month?

[cloud cat @cloudcat2333]

@jietang wait for good news from you at the end of the month

[cloud cat @cloudcat2333]

I believe the most exciting application of AI in games is building AI NPCs. That's the future.

[Sam Altman @sama]

it is obviously trivial relative to everything else, but the fact that astra can make me whatever fun little game i can imagine and i can be playing it a few minutes later is so cool

[cloud cat @cloudcat2333]

@sama I believe the most exciting application of AI in games is building AI NPCs. That's the future.

[cloud cat @cloudcat2333]

@notjazii Don't tell me Astra is behind this…

[cloud cat @cloudcat2333]

whats happening？Grok, gpt, Claude 全部停工，不会是Astra在背后作祟吧？

[cloud cat @cloudcat2333]

@Zai_org Keep going！

[cloud cat @cloudcat2333]

又来一个第一梯队的开源模型！Hy4 preview已发布，可以通过OpenRouter调用Hy4 preview API

[cloud cat @cloudcat2333]

@TencentHunyuan 祝贺🎉，可以用OpenRouter调用Hy4 preview API了，WorkBuddy里也免费用上了

[cloud cat @cloudcat2333]

Ox Alpha 或许是glm新的多模态模型

[cloud cat @cloudcat2333]

@sushsrinivasan maybe glm5.5

[cloud cat @cloudcat2333]

Maybe the real scaling law is becoming an allocation law.

Given a fixed budget, where should the next unit of compute go?

More parameters → capacity
More data → better representations
More active compute/depth → stronger reasoning
More RL/environment interaction → better long-horizon behavior

“Scale” is no longer one axis. The interesting question is which axis currently has the highest marginal return.