About PrimeSet Labs
We are an AI lab in San Francisco. We build the environments and benchmarks frontier labs use to teach models expert work.
We build the hardest work AI can’t do yet, and make it verifiable, so frontier models can learn it on the path to AGI.
The prime set
Our mark is the set of primes below ten: four squares with sides of 2, 3, 5 and 7, sharing one corner. Every whole number is built from primes. We build the hard, verifiable tasks that frontier models build expert skill from.
- 2A 2 × 2 square: the first prime, lit
- 3A 3 × 3 square
- 5A 5 × 5 square
- 7A 7 × 7 square
How we work with frontier labs
License what exists, or build around what you need next.
- 01Ready-made environmentsLicense validated environments across domains.
- 02Custom buildsEnvironments around the skills you care about.
- 03EvaluationsBenchmarks and evals for your models.
How we build every environment
The same bar in every domain, from the first task to the last.
- 01
Verifiable
Automatic verifiers check every answer and the evidence behind it.
- 02
Frontier-hard
Calibrated so today’s best models are challenged, and can still learn.
- 03
Uncontaminated
Test sets are held out and never used for training.
- 04
Expert-built
Designed and reviewed by people who do the work.