#LLM
Curated coverage on «LLM»: releases, research, benchmarks, open-source.

Kimi K2.5: Moonshot AI Goes Open Source and Upends Its Rivals' Plans
Yang Zhilin has unexpectedly open-sourced Kimi K2.5. While others close up, Moonshot AI is betting on the community. Why does it need this?

Kimi K2.5: Moonshot AI Builds an Army of Agents on a Budget
While Western giants argue over model power, China's Moonshot AI has gone after speed. Kimi K2.5 turns complex tasks from a tedious queue into a fast-moving swarm.

ICLR 2026: Why Breaking Into AI's Elite Is Harder Than Ever
A 28% acceptance rate turns the top neural network conference into real "Hunger Games" for researchers worldwide. Here's why.

Qwen 2.5-Max: Alibaba Caught Up With Google While You Were Asleep

Moonshot AI strikes first: can Alibaba keep pace with DeepSeek?

Sora for Marketplaces: Chinese Developers Build an Assembly Line for Perfect Sales Videos

Qwen3-Max-Thinking: China's Answer to OpenAI Has Silicon Valley on Edge

The LLM Black Box: Why We Still Don't Understand How They Think
Why modern language models remain a mystery even to their creators, and how researchers try to look under the hood with interpretability methods.

Maia 200: Microsoft Builds Its Own Silicon Fortress, Without Nvidia
Microsoft has released Maia 200, the second generation of its in-house AI chip. It is a direct challenge to Nvidia's dominance and an attempt to make the Azure…

The AGI Race: Hassabis and Amodei Disagree on When the World as We Know It Ends

AWS AppSync Events: Your New AI Gateway Without the Headaches

ComfyUI for Everyone: Stop Fearing Nodes and Take Full Control of Stable Diffusion

You've Finished Andrew Ng's Course: Where to Go Next So You Don't Stay a Perpetual Junior

Clones Flopped: Meta Quietly Buries Its AI Celebrities
Mark Zuckerberg has conceded that chatbots with the faces of Snoop Dogg and Kendall Jenner went nowhere. Meta is shutting down a project that cost millions of…

The End of the Chalk Era: AI Starts Proving Theorems for Humans
Mathematicians may want to start worrying: Numina-Lean-Agent automates proofs. While we argue about chatbots, AI is moving into fundamental science and cyberspace.









