rekursiv.ai
Scale AI scientists whose own breakthroughs accelerate the next.
About rekursiv.ai
We're scaling self-improving AI scientist teams to ideate/experiment/discover, generating new knowledge autonomously. Their discoveries reduced ARC-1/2 costs by 10,000× while maintaining state-of-the-art accuracy and yielded material advances on combinatorial ML problems. We believe that AI is bounded not by compute, but ideas. Scaling to millions of sessions of discovery enables leveraging its own discoveries and presents a novel training source capable of ushering a new era of self-reliant foundational models.
Public traction evidence
Each signal links to the public source used for attribution.
- X
@BenjaminDEKR Now you know why I left Google to join Luma. I was in the team that developed Veo early on but knew it would never be shipped to the masses for quite a long time, same as Sora. Not until a company like Luma forces their hand,…
@BenjaminDEKR Now you know why I left Google to join Luma. I was in the team that developed Veo early on but knew it would never be shipped to the masses for quite a long time, same as Sora. Not until a company like Luma forces their hand, that is (at least I hope, gimme access...
- LinkedIn
Starting rekursiv.ai with Joshua Dillon
I'm starting a new venture as a co-founder. Joshua V. Dillon and I have formed rekursiv.ai and are building teams of autonomous AI scientists.
- X
Great to see training next-token-prediction on video unlocks emergent understanding.
Great to see training next-token-prediction on video unlocks emergent understanding. We tried something similar with VideoPoet in 2023 but it failed. Key unlock seems to be higher VAE compression, more semantic latent space, and cleaner data. Well done!
- X
We raised $5M from Y Combinator to scale AI scientists whose own breakthroughs accelerate the next.
We raised $5M from Y Combinator to scale AI scientists whose own breakthroughs accelerate the next. In days, they matched top models on ARC-AGI at up to 10,000× lower cost, every gain from methods they invented themselves. How it works below:
- GitHub
GitHub signal from rekursiv-ai/sagent
rekursiv-ai/sagent: A coding-agent CLI and strongly-typed Python library -- self-mutating, hot-swapping, multi-provider, with async tool calls and bidirectional recursive spawn. (Python).
- X
This is a core problem we found when conducting autonomous research: without a clear benchmark or baseline, it has a hard time making progress.
This is a core problem we found when conducting autonomous research: without a clear benchmark or baseline, it has a hard time making progress. However, we found these issues can be substantially improved with the right harness design. More to come 👀
- X
Reading documentation written by Claude hurts my brain, while GPT doesn't.
Reading documentation written by Claude hurts my brain, while GPT doesn't. The cognitive load just seems way too high. Anyone else feel the same?
- GitHub
GitHub signal from rekursiv-ai/trackinizer
rekursiv-ai/trackinizer: Epistemological database for agent and human efforts, beliefs, and findings. (Python).