Traverse.

Reinforcement learning environments for long-horizon agent journeys.

We are building the infrastructure to automate all economically valuable work, starting with software engineering as it is the most verifiable and deterministic domain that exists today. Tests either pass or fail and builds either succeed or break, which makes it the ideal starting point for training agents on real outcomes rather than human demonstrations.

Software also creates a natural flywheel where AI that can write code can help improve its own training infrastructure and accelerate its own development, which is a feedback loop we want to establish as early as possible.

AGI is coming, and when it does the bottleneck will be high quality training data for every industry where valuable work takes place. Building a reinforcement learning environment for legal research, medical diagnosis or financial analysis by scraping the internet is infeasible. Actual domain experts are required as they understand the workflows and edge cases and accumulated knowledge that makes each field what it is.

Our approach is to assemble the absolute best people in every vertical we enter, pairing world class domain experts with world class ML engineers to build training environments that simply cannot exist without that combination of talent, and our bar for what counts as the best is extremely high because mediocre environments produce mediocre agents and we are not interested in that outcome.

We envision a future where Traverse is the company that builds high quality RL environments across all verticals, not one option among many but the clear leader in every domain we touch, and we are building toward that future starting now with software engineering as our first vertical.

The path to general intelligence is to traverse through every domain where work happens.