Twitxy Saved

snimu (Sebastian Müller)

Verified account Prime Intellect

@omouamoua

Leading the RL Residency and doing research @PrimeIntellect
Munich, Bavaria snimu.github.io Joined December 2022
710 Following 4.2K Followers 2K Posts

Latest posts 5

snimu (Sebastian Müller) Verified account @omouamoua
Really proud of all the residents who have done amazing work the last few month!
And we're not done yet.
Prime Intellect @PrimeIntellect 17h
Our RL Residency gives independent researchers compute grants and mentorship to build open evals and training environments for overlooked problems.

In 2026, residents shipped projects spanning robot control, medical imaging, gameplay, and GPU kernels. Some highlights:
View quoted post on X
7 replies 8 reposts 73 likes 3.5K views View on X
snimu (Sebastian Müller) Verified account @omouamoua
Mazebench is so good that Anthropic already trains on it (which on the flip side means you shouldn’t take their scores on this excellent eval very seriously tbh, or any claims based on such benchmarks)
💺 Verified account @patience_cave 4d
Fable 5.2 is so good, it solves puzzles almost suspiciously well. In a difficult 3D level, it gave the solution to an old version of MazeBench.

This level was updated a couple weeks ago, so the old solution no longer works. For some reason this caught Fable 5.2 off guard.

In some cases it was able to figure out the recent changes to MazeBench, but in this case it was not.
View quoted post on X
2 replies 5 reposts 62 likes 2.4K views View on X
snimu (Sebastian Müller) Verified account @omouamoua
people talking about how sth fundamental is missing for true creativity in LLMs: how is this not just a question of data and compute?

Humans spend unbelievable amounts of compute, and see incredible amounts of data for this (whole lifetimes over centuries of billions of people). Our advantage is that we see the producers of the creativity (humans) as inherently valuable, and are thus willing to spend on the resources they need: food, water, housing, energy, whatever.

Achieving the same level of long-term creativity and long-term planning in machines without significant spending seems like a pipedream, so allegations of something fundamental missing are pretty shaky imo (allegations of efficiency gains being left on the table otoh are very legit, we can do much much better than humans or current LLMs).
6 replies 0 reposts 24 likes 1.9K views View on X
snimu (Sebastian Müller) Verified account @omouamoua
> a false claim we caught in a Navier-Stokes paper

@OpenAI
Sinatras Verified account @myainotez Nov 13
We built a fieldscale Flywheel graph for math and are sharing the first results from our meta-harness going wild over it. We have 60+ followup results on OpenAI’s 721 papers: improvements on several of them, and a false claim we caught in a Navier-Stokes paper check it out
View quoted post on X