Ido Hakimi

@idohakimi

Chief Scientist @ Tzafon (ex DeepMind). The future of intelligence isn't knowing more. It's learning faster.

Joined December 2018

491 Following

237 Followers

Replies Reposts Media

Pinned
Ido Hakimi
@idohakimi
Jun 17
Hacker wife, good life ❤️
Shunit Haviv Hakimi
@ShunitHH
Jun 17
People around me are building really cool thing, so I built /brag, a Claude Code skill that turns your project into a short, shareable launch video. Just tell Claude: "Hey, let's /brag about this." The video below was made with /brag, to brag about /brag.
Ido Hakimi
@idohakimi
Aug 26
Frontier AI labs checking their models for harmful behavior
Shash
@pokershash
Aug 26
This is not a meme anymore
Ido Hakimi
@idohakimi
Jul 3
Fable is currently unusable IMO, not because it can't answer, but because it seems optimized for tokenmaxing
Ido Hakimi
@idohakimi
Apr 7
1/ cool paper from @JD_Corporate on self-distillation. the failure mode is kind of funny once you see it. if the privileged teacher gets to steer the gradient direction, the policy starts learning from information it will never have at test time and slowly cooks itself.
DailyPapers
@HuggingPapers
Apr 6
RLSD: RLVR with Self-Distillation Unifying on-policy self-distillation with verifiable rewards to fix information leakage and instability—using token-level policy differences for fine-grained updates while leveraging environmental feedback for reliable directions.
1
Ido Hakimi
@idohakimi
Mar 24
MolmoWeb by @tanmay2099 and @allen_ai is a glimpse of where CUA is headed. Truly useful CUA agents are close, and 2026 is going to be the year of CUA.
Ai2
@allen_ai
Mar 24
Today we're releasing MolmoWeb, an open source agent that can navigate + complete tasks in a browser on your behalf. Built on Molmo 2 in 4B & 8B sizes, it sets a new open-weight SOTA across four major web-agent benchmarks & even surpasses agents built on proprietary models. 🧵

Ido Hakimi

Ido Hakimi’s posts