My second short story release in English is ready: Tales of Illustrious Computer Scientists: Iola Varga, nun and computer scientist. invece.org/iola.html

The point about local AI.

Didn't know that. Will try.

Claude /remote-control is becoming one reason to prefer Claude CLI versus Codex CLI sometimes, since it is too convenient. I'm not sure why @OpenAIDevs is lagging on that, many open issues and requests, and does not look too hard to add.

Btw sometimes I open a Claude session with /remote-control and ask it to open a tmux with Astra to do the work, and just report me from time to time and act as a proxy for my messages. This way I work in the tmux session when in front of the computer, and also control it remotely

Anyway at this point there are little projects I do without using *both* Claude and GPT, one checking the work or the other, both collaborating on complex stuff and so forth.

Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.

🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.

Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.

Open weights by Nov 4, 2026.

🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1…

Work in progress, Iris.c in a DGX Spark running Flux 3 Klein. I know prompts for Flux 3 should be post-processed for better results, but it kind works even with humans written descriptions, more or less.

First version (like in the demo) is already pushed online.

Support coming for Strix / Gorgon Halo as well btw.

New Iris release soon.

Those products are not called "Claude Secretary" and alike in order to lower social anxiety?

Not #SLOPTOBERFEST but made on my LTX-MAC app native mac metal offline used minimax h3.c h/t @antirez

While Mistral Large 4 is a good news for Europe and AI, it is not a good fit for DwarfStar because of the very large number of active parameters, not just because of memory: would be too slow anyway. But anyway congrats to @MistralAI and waiting for a smaller MoE :D

Coming from high end MacBooks the Dell XPS screen and colors are a shock. OLED is another world.

A huge thank you to @syntaxfm for a generous sponsorship of my GitHub work. Thanks!

Trying @omarchy in a Dell XPS kindly donated by @dhh (I will try to adapt DwarfStar SSD streaming of large models btw, maybe it will be fast enough for QA usage). I'm really impressed by the "welcome" after the installation. The system is playful and not boring, which is cool.

After use these models intensively, now I understand what @antirez means. DeepSeek 4.1 Flash 1017 decode speed locally should be ~200 tok/s to be competitive.

Even if the DGX Spark is SM121 and not SM100, yet maybe we can get something good for DwarfStar kernels from DeepSeek Infer. Let's try.

AI: there is great chaos under heaven; the situation is excellent.

DeepSeek infer is the cloud version of DwarfStar, conceptually. Vertical inference engine for one model and specific hardware. Good move. And well, a lesson to the ones that criticized this approach as well.