Verifiable data for
recursive self-improving AI
Verifiable data for recursive self-improving AI
Backed By
Jigsaw is a research lab focused on recursive self-improvement. We develop rigorous methods for turning AI R&D into post-training artifacts that advance self-improving systems at frontier research labs.
RSI will be the most impactful mechanism in the mission to achieve ASI. Our work makes its progression verifiable and its discoveries grounded.
RSI will be the most impactful mechanism in the mission to achieve ASI. Our work makes its progression verifiable and its discoveries grounded.
Frontier data for RSI,
the fastest path to superintelligence.
Our bet is that AI research can be mapped onto a closed set of verifiable problems, and that leveraging those problems will yield comprehensive gains in model performance.
We distill frontier AI R&D into gradable tasks, and develop rigorous methods to evaluate verifier quality and measure post-training outcomes.
We distill frontier AI R&D into gradable tasks, and develop rigorous methods to evaluate verifier quality and measure post-training outcomes.
Frontier data for RSI, the fastest path to superintelligence.
Our bet is that AI research can be mapped onto a closed set of verifiable problems, and that leveraging those problems will yield comprehensive gains in model performance.
We distill frontier AI R&D into gradable tasks, and develop rigorous methods to evaluate verifier quality and measure post-training outcomes.
