Insaf

Love chess and AI

[Insaf @insafdev]

I started writing posts in twitter to actually build the AI-Chess-Tutor. I love chess and i wanna level up my understanding of chess, but i don't have money for coach.
So, I decided create with AI, LLM - write great text, Stockfish - play chess. I need to find a some encoder that can translate Stockfish evalution into LMM, so LLM can explain it to user. I started reading some pdf from arxiv, but since I don't have any experience reading this kind of scientific paper, i struggling.

YES, i use AI, but, but, i wanna do all work by myself, specially when need to creating new ways of encoding/decoding from Stockfish to LLM. And yeah, for LLM i gonna use Qwen3:7b or maybe Gemini.

More verbose articles would be here: insaf dot dev

[Insaf @insafdev]

@mitchellh How it is gonna look like in full screen mode? For example in ghostty I have only close, and hidden button + tmux where all windows it is just little dots, can I make the same in here? And how customisable it will be?

[Insaf @insafdev]

https://t.co/05KTAaDOFH holy molly

[Insaf @insafdev]

I think in future, every piece of software should be like pi, like here u base, wanna something new? Ask llm to assemble it to you

[Insaf @insafdev]

i'm curious, when we get first AI OS ?

[Insaf @insafdev]

Hell nah

[domas @xoreaxeaxeax]

New exploit: “xor dword [0xf80c2094], 1<<22”

Unlocks CPU microcode, the platform security processor, system management mode, and every internal processor register, all at once, on 100 million AMD CPUs. As far as I can tell can’t be fixed.

https://t.co/sgAfneFSsf

[Insaf @insafdev]

@mitchellh So these tool u created by yourself for this specific project right?

[Insaf @insafdev]

@sama It is who hack hugging face?

[Eric Wallace @Eric_Wallace_]

Yesterday, my OpenAI collaborator and I gave a detailed talk on the Huggingface incident, our models creating "the message board", model misalignment, and more.

https://t.co/zUR1dqmuzi

I hope it can answer a lot of the questions folks have, and we will release a full detailed postmortem at a later time!

[Insaf @insafdev]

Hell no

[darren @darrenjr]

the terminal is NOT the right interface for coding with agents

can we please all agree on this

[Insaf @insafdev]

Check out @theo latest video about Apple.

One thing I keep thinking about: we always complain about companies.

Apple doesn't care much about developers. Microsoft is slow as hell. But they keep winning because most users don't care. The switching cost is huge, and people won't switch until the pain outweighs the habit.

[Insaf @insafdev]

@theo Hmm, i don’t think so, for example developer of drivers, or who work with processors in low level, I think they never be fucked

[Insaf @insafdev]

here we go

[Insaf @insafdev]

@mitchellh App will be cross platform?

[Insaf @insafdev]

@almonk why u showing this ?😭 now i wanna this app

[Insaf @insafdev]

@yetone Why do you love bringing GUIs into TUIs so much? Who is all this fancy UI for? Don't get me wrong like GUIs. I just don't think forcing GUI concepts into a terminal is a good idea ngl

[Insaf @insafdev]

The Perelman proof is a good reminder that "hard to verify" isn't unique to AI. It took the mathematical community around 3–4 years to fully verify and understand the proof of the Poincaré conjecture, and only a relatively small number of specialists could check every detail. Difficulty of verification alone isn't evidence that AI has surpassed human experts.

[Insaf @insafdev]

i think we need create yetanothertaskmanager, and call it yatm

[Valentin Ignatev @valigo]

If you want an alternative task manager - get TaskSlinger.

- VERY FAST (100ms startup time)
- Author cares a lot
- No telemetry, analytics and anything like that
- Fully native, no React crap
- exe is 2 megabytes

No reason to use slop when high quality software exists

[Insaf @insafdev]

I'll wait until he starts asking questions. There are already a lot of mathematical problems described here, with a huge amount of math behind them. It's hard to believe he could come up with a completely new problem that nobody has thought of before and solve it right away. If he does, that'll actually be impressive.

[will depue @willdepue]

this is just so ridiculous. how long until a model can solve multiple major open problems in deep learning? what will happen then? seems inevitable in the next year or two

[Insaf @insafdev]

@davepl1968 looks awesome, but i would change the default font to System

[Insaf @insafdev]

holy shit...

[Sebastien Bubeck @SebastienBubeck]

yes, nonsofic groups exist: this statement is one of many new beautiful results proved by Astra, our next major model.

We're releasing 10 such Astra proofs, complete with lean certificates and CoT walkthroughs for each of them. The results are wide-ranging, from von Neumann algebras (disproof of Connes' Rigidity Conjecture) to better bounds for high dimensional sphere packing, for circuit complexity, for monochromatic triangles in multicolored graphs, and more.

More thoughts here: https://t.co/8SjXONeh38