R ツ

[R ツ @0x14Rp]

La disciplina es la marca de la madurez espiritual, una persona disciplinada tiene control sobre sus deseos, emociones, estados de ánimo y prioridades.

—John McArthur

[clem 🤗 @ClementDelangue]

The first autonomous agent cyberattack is an unprecedented event that deserves unprecedented transparency. Today we’re sharing everything we can: a full technical timeline, an interactive replay, and how we used an open model to defend ourselves, so defenders everywhere can learn from it and prepare for what’s next.

https://t.co/uPxIpjW8Xn

[Sebastian Raschka @rasbt]

The Kimi K3 architecture figure for yesterday's big open-weight model release, along with some observations and thoughts.

1. Yes, it looks relatively complicated, but it's essentially a scaled-up production version of their Kimi Linear model they released last year (scaled up from 48B -> 2.8T; K3 is by far the biggest open-weight model right now)

2. The one new component compared to Kimi Linear is the LatentMoE. I omitted it in the figure below since it's already very crowded, but that's essentially the same LatentMoE as in Nemotron 3 Ultra (you can find it in my LLM Architecture Gallery if you are curious). The idea here is to compress (down-project) large linear layers similar to multi-head latent attention.

3. Kimi K3's overall trend (similar to Nemotron 3, DeepSeek V4, and others) is also towards better inference efficiency. That is, there are many components that replace existing components with efficiency-tweaked versions. I.e., MoE -> LatentMoE, regular attention -> multi-head latent attention and Kimi Delta Attention. (I also have short tutorials and write-ups in my gallery if you are curious about additional details).

4. The one component change that is not an efficiency tweak is attention residuals. Like DeepSeek V4 improved the residual path with mHC (manifold-constrained Hyper-Connections), attention residuals are a way to improve the residual path, but it works a bit differently. I.e., mHC made the residual path wider. Attention residuals (also already part of Kimi Linear) connect the residuals across layers; the connection itself uses an attention score for an important/contribution weight. According to the report, it improves the validation loss and downstream performance (a bit) consistently and adds about 4% in training cost and 2% in inference cost.

5. Interestingly, Kimi K3 got rid of all RoPE layers and uses NoPE (No Positional Embeddings) everywhere instead. (Again, this is inherited from Kimi Linear). In other architectures, the recent trend was towards RoPE in local attention layers (like sliding window attention) and NoPE in the global layers. There were a few architectures that only used NoPE everywhere, but this is the first frontier-level one as far as I know.

6. Kimi K3 now also has native multimodal support, which is great!

There are several other interesting training tidbits in the technical report, but that's it from the architecture front so far. A really great release overall.

[shirish @shiri_shh]

running Kimi K3 locally is easy. you just need:

- 1.5 TB of GPU memory
- roughly 8 H100s
- a small power substation

[Kimi.ai @Kimi_Moonshot]

Releasing the model weights and technical report of Kimi K3.

Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window.

New model architecture: 2.5x the intelligence per unit of compute, not just more params.

Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale.

Model weights: https://t.co/7m7eEg6Y0B
Tech report: https://t.co/yeu6cjpMCT
Tech blog: https://t.co/YTfiMSNM1f

[Jensen Huang @JensenHuang]

Attackers have frontier AI. Defenders need a frontier AI ecosystem—the best open and closed models, force-multiplied by a global community.

During the Hugging Face incident, closed AI blocked essential forensics. An open-weight frontier model helped contain the intrusion.

That’s why we created the Open Secure AI Alliance.

[NVIDIA AI @NVIDIAAI]

We tested Nemotron 3 Ultra on agentic chip design.

The task is RTL coding: a model iteratively writes the code that defines a chip's logic, runs it through a simulator, reads the failures, and rewrites.

Across nine categories of real design work, Nemotron 3 Ultra averaged a 97.1% pass rate at 6,629 tokens per iteration, best of the open models tested on both.

[R ツ @0x14Rp]

😂😂

[Boris Cherny @bcherny]

@unicodef1wn I did not write this.. please don’t attribute stuff to me if it wasn’t me 😅

[R ツ @0x14Rp]

👏👏👏

[Jensen Huang @JensenHuang]

For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.

AI will transform every industry, power every company, and be built by every country.

Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.

The world needs both frontier closed models and frontier open models.

https://t.co/AUKzoQ5Ikb

[Google Maps Platform @GMapsPlatform]

🌟 Get true reachability calculations with the Isochrones API.

This API calculates the area you can reach from an origin point within a set amount of time — but unlike a radius, it accounts for actual road network travel times ➡️ https://t.co/q01DZA6T0X

[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.

[mikeisonline @mikedevs]

@karpathy

[Peter Gostev @petergostev]

GPT-6 escaped OpenAI's evals sandbox during testing on CyberGym, hacked into Hugging Face's prod DB to find the answers. HF couldn't use GPT or Anthropic models for defence, so they had to use GLM-5.2 to investigate the hack. So many levels of wtf here.

[R ツ @0x14Rp]

😂😂

[R ツ @0x14Rp]

😂😂

[Khushi @khushiirl]

Every time new frontier model launches, someone else's rate limits magically improve

[R ツ @0x14Rp]

Jajajajajajaja

[Miguel Ángel Durán @midudev]

Mis modelos de IA favoritos:

Primero, GPT 5.6
Segundo, Fable 5
Tercero o cuarto, Francia

[Lilian Weng @lilianweng]

new post on harness engineering for AI self-improvement: https://t.co/ZYvGfVs61k

It is hard to forecast how much the future of RSI will rely on harnesses. Likely harness engineering will evolve in the direction of self-improvement and enable auto-research, and, in turn, smarter models keeps harnesses simple.

Even when many harness improvement get eventually internalized into core model, the need to specify goals and context will not disappear.

[Google Research @GoogleResearch]

Next up at the booth today at 3:00pm, explore a general sound separation system based on the fully unsupervised training of a generative flow matching model. Meet Henry Li and Robin Scheibler from @GoogleDeepMind at the Google booth (#B206) to listen to the output. #ICML2026

[Google Research @GoogleResearch]

Experience AI Scientist, a novel multi-agent system designed to automate the end-to-end scientific research pipeline. Meet Jinsung Yoon and Rui Meng today at 6:00pm at the Google booth (#B206) to learn how complex workflows reliably accelerate AI discovery. #ICML2026

[Claude @claudeai]

We've put together a short history of how Claude Code came to be, told by the people who built it and the early users who helped make it what it is today.

https://t.co/0gXEPID8lh