Lakhdar Meftah

PhD in Computer science

[Lakhdar Meftah @m3ftah]

@forgebitz Good one

[Lakhdar Meftah @m3ftah]

@Mr_Derivatives She is always fantastic!

[Lakhdar Meftah @m3ftah]

@MatthewBerman MTP + small model quality improvement (2 years) + Taalas chip

[Lakhdar Meftah @m3ftah]

@heyeaslo Feels good

[Lakhdar Meftah @m3ftah]

@jimcramer They do not need to prove things, they said they will show on the earnings

[Lakhdar Meftah @m3ftah]

@adcock_brett @Figure_robot So freaking!

[Lakhdar Meftah @m3ftah]

@sickdotdev Won’t be long until it’s back again anyways

[Lakhdar Meftah @m3ftah]

@joyjiao12 @scaling01 When?

[Lakhdar Meftah @m3ftah]

@sama That started with an uppercase, new device?

[Lakhdar Meftah @m3ftah]

@andrewdfeldman @cerebras You say fearless, but you are also saying small models are good, work on bigger models, please!

[Lakhdar Meftah @m3ftah]

@cerebras Interesting, what quantization did you use, and did you use MTP or support it?

[Lakhdar Meftah @m3ftah]

@chamath You keep bringing private datasets, those are bot needed on the base model, they only need to learn intelligence and patterns, only general patterns, once they are smart they can read the dataset and learn from it on runtime.

[Lakhdar Meftah @m3ftah]

@alibaba_cloud Nice!

[Lakhdar Meftah @m3ftah]

@scaling01 Seems better than GPT-5.5

[Lakhdar Meftah @m3ftah]

@Designarena @Zai_org Impressive if true

[Lakhdar Meftah @m3ftah]

@willahmed Agree

[Lakhdar Meftah @m3ftah]

@Zai_org @ClementDelangue Highs are close

[Lakhdar Meftah @m3ftah]

@DanKulkov And chatty

[Lakhdar Meftah @m3ftah]

@TheRonnieVShow After the fact

[Lakhdar Meftah @m3ftah]

@TheAhmadOsman Maybe