New Frontiers

Over the last years, the world has produced a set of models that can outperform the smartest humans in narrow domains: Frontier AI models are beating the best engineers in competitive programming and are solving Millennium Prize problems that had resisted mathematicians for generations.

We believe that achievements of similar scale are possible across other fields. What makes mathematics different is that training problems are abundant, answers can be verified automatically, and models can improve without interacting with the real world. This makes it easy to find weaknesses, generate training tasks, and iterate quickly.

Proximal is building infrastructure that enables this feedback loop in any domain. Instead of relying on manual data curation, we build systems that turn unstructured real-world data, whether it is raw agent traces or artifacts produced from human workflows, into evaluations that expose capability gaps and then use those signals to generate targeted post-training data and improve how models are deployed.

In the past year, we have started with software engineering as a proving ground, and have worked with frontier labs and enterprises to help them train better coding agents. This enabled us to scale to $200M in annualized revenue, assemble a lean team of the most talented engineers and researchers and build world-class infrastructure for post-training and data research.

Now, we are expanding into other domains: we are excited about a world in which AI systems design targeted drugs within days of a new disease emerging, create specialized chips for workloads that today require teams of expert engineers, and rewrite legacy software that still runs hospitals, power grids, governments, and other critical infrastructure. We believe that our work is the most impactful work to get there.

We are a small team of researchers and engineers from Cursor, Google DeepMind, Meta Superintelligence, Prime Intellect, Citadel, and Jane Street, with former founders making up more than half the team.

We’re fortunate to be backed by General Catalyst who led our $15m at a $300m valuation, as well as SV Angel, Diede van Lamoen (Go Global Ventures), Chemistry, and individuals like Liam Fedus (CEO, Periodic Labs), Kevin Weil (former CPO, OpenAI), and Erik Bernhardsson (CEO, Modal).