Charis Mitsakis

Building LLM agents · conversational AI

[Charis Mitsakis @cmitsakis]

@xeophon I wish Zcode was MIT licensed as well, but it's closed source I think

[Charis Mitsakis @cmitsakis]

@arthurmensch All day I argue with the conspiracy theorists and the fat-shamers that deny the existence of Le Chaton Fat 🐱 (yes denying its existence is the ultimate form of fat-shaming)

[Charis Mitsakis @cmitsakis]

@paw_lean @pizzaboy exactly! smart guy🧠

[Charis Mitsakis @cmitsakis]

@julien_c @huggingface are you low-key fat-shaming Le Chaton Fat? 🐱

[Charis Mitsakis @cmitsakis]

@AN0NYME4EU @kimmonismus of course. I never believed the denialists conspiracy theorists

[Charis Mitsakis @cmitsakis]

VibeThinker-3B, post-trained upon Qwen2.5-Coder-3B base, scored 94.3 on AIME26, with a performance similar to DeepSeek V3.2, GLM-5, and Gemini 3 Pro.

Small models are the future for agents because they can use tools to get the knowledge they lack and they can run fast and cheap.

[Francesco Bertolotti @f14bertolotti]

Stellar performance from a 3B model. These results were achieved primarily through post-training refinements on Qwen2.5-Coder. The paper doesn't provide many details, but it appears they distill from RL ckpts and then do a final RL-based instruct RL.

🔗https://t.co/FmdRwGNMOg

[Charis Mitsakis @cmitsakis]

denying the existence of Le Chaton Fat is the newest form of fat-shaming

[Charis Mitsakis @cmitsakis]

@cargoshortdad64 Recursive Self Fattening reaching galactic dimensions 🐱

[Charis Mitsakis @cmitsakis]

@scaling01 you called Le Chaton Fat a joke? 🐱

[Charis Mitsakis @cmitsakis]

@kimmonismus I think small models are the future for agents because they can use tools to get the knowledge and the can run fast and cheap

[Charis Mitsakis @cmitsakis]

@financialjuice is this for Le Chaton Fat? 🐱

[Charis Mitsakis @cmitsakis]

The haters want to ban Le Chaton Fat, but they don't understand it's futile because Mistral implemented Recursive Self Fattening, and nothing can stop the exponential.

[GLIF @heyglif]

The French Goverment needs to STOP Le Chaton Fat before it's too late. This is the FAT takeoff we've been warned against for years!

[Charis Mitsakis @cmitsakis]

@techmeditator it's not just fat, but it's getting fatter through Recursive Self Fattening. The details of the RSF implementation are described on the paper.

[Charis Mitsakis @cmitsakis]

@ysbhalgat 😮 the AI race is over then. Le Chaton Fat won 🐱

[Charis Mitsakis @cmitsakis]

@wire_catt If only we had the hardware. Personally I prefer Le Chaton Fat 🐱 even though I will never be able to run it locally

[Charis Mitsakis @cmitsakis]

@zephyr_z9 it's over for all AI labs 🐱

[Charis Mitsakis @cmitsakis]

@petergyang they don't want us to fix our code, they prefer we run insecure slop they can hack

[Charis Mitsakis @cmitsakis]

@paw_lean There are no shortcuts. Persistence is important