Simone

distributed systems / deep learning

[Simone @sibellavia]

distributed 5000x5000 matrix multiplication on my two laptops collaborating in the execution
i’m simulating multiple peers connected to a network where tinyp2p distributes the computational load
one node submits the task to the network, which is processed and divided into multiple subtasks
all nodes return subtask results which are then merged into the final computation

[antirez @antirez]

Modern AI resulted from research made also by many non-US scientists (Hinton, the French folks, Linnainmaa, many others). The pre-training corpus was produced worldwide with massive code contribution from Europe OSS. What is happening with frontier LLMs is unacceptable.

[Armin Ronacher ⇌ @mitsuhiko]

Europeans please wake up.

[antirez @antirez]

Anthropic did a big strategic error. Normally they compare their models with their old models. Instead today, now that everybody knows how strong GPT 5.5 is at coding, they put it in the mix, basically showing all their customers that the benchmarks can't be trusted.

[Simone @sibellavia]

@antirez @evilsocket Hahahahahaha 😂

[Simone @sibellavia]

@antirez unexpected! in the sense that it’s good news and I’m really curious to see how you’ll do it.

[Simone @sibellavia]

@mitsuhiko nice!

[Simone @sibellavia]

Excellent thread on how programming with LLMs should be done correctly.

[antirez @antirez]

So let's start from this post to tell you about my journey with the LLMs and the new Redis data structure I'm implementing right now. Vector Sets were hand-coded, this time I decided to use Claude/Codex as a helpers, and guess what? The work I had to do was huge. Thread:

[Simone @sibellavia]

@awnihannun congrats Awni. big MLX fan.

[Simone @sibellavia]

@mitchellh a GUI around pi would be nice

[Q @qtnx_]

message to certain labs: Europe is not so bad trust

[antirez @antirez]

Of all human feelings, envy is the one I despise the most.

[Simone @sibellavia]

Sandboxes are the new trend. Remember you can just run your microVM locally and run LLM-generated code there, no need of wrappers or external bloated solutions.

[Simone @sibellavia]

@alxfazio I'm working on a tiny tool for knowledge/memory management
https://t.co/SeZZRHMHgd

[Simone @sibellavia]

@axk0ng @mitsuhiko > One of the major challenges I see is that you have to write your software with Wasm in mind
Exactly. I built a couple of proof of concepts with coding agents in mind that could enable WASM sandboxing. But it's not as versatile as I had hoped.

[Simone @sibellavia]

@kuberdenis Calling codex from CC in specific use cases (such as code review) may be a solution!

https://t.co/U9rR1r8pAo

[antirez @antirez]

If you explain how to do it, Claude Code is able to use Codex when it is incapable of solving certain issues. This way you have the best of both worlds.

Here is the skill file that lets Claude use Codex.
https://t.co/qdit9JQSbm

[Simone @sibellavia]

@BradAI blocked for AI slop

[Simone @sibellavia]

"That said, if vibe coding is the process of producing software without much understanding of what is going on [...], automatic programming is the process of producing software that attempts to be high quality and strictly following the producer's vision of the software"

[antirez @antirez]

New blog post: Automatic Programming -> https://t.co/Z4tWUsMC26

[Simone @sibellavia]

@BradAI That's the distinction he actually made (automatic programming vs vibe coding).

[Simone @sibellavia]

@MattPRD @moltbook You should improve comments upvotes.

[Simone @sibellavia]

@mitsuhiko is this pi?