Meet Ultraresearch — co-design with superhuman scientific reasoning capabilities. Coordinate and orchestrate a diverse pool of agents on the fly to deliver superior performance. Built by Self Play Lab.
Ultraresearch
The hardest problems aren’t solved by scaling what already exists — they’re solved by reasoning past the edge of the known. Most tools replay patterns from training data and stall the moment the problem outgrows the corpus. We built the opposite: an autoresearch system that turns your objective into a formal specification through autoformalization, then launches simulation driven by compute — coordinating a diverse pool of specialized agents that generate, test, and refine candidates at scale. State the goal, set the constraints, and Ultraresearch takes it end to end — stress-testing every result and delivering only what survives verification. Superior performance through orchestration, not approximation.
Recent Research
View all- First research notes are on the way — check back soon.
The objective is yours.
What will you build next?
- Code
- GitHub