# Divergent Labs

> A research lab building reasoning systems around divergent, nonlinear thinking. We publish the harness before we train the model.

Divergent Labs builds "adhd", an open-source reasoning harness for coding
agents: N candidates generated in parallel under isolated cognitive frames,
one critic pass to score and prune, survivors deepened into inspectable plans.
Evaluated head-to-head against a single-shot baseline across six open-ended
engineering problems (adhd won five of six). Founded by Udit Akhouri.

## Primary pages

- [Home](https://divergent.sh/): thesis, the harness, evals, founder
- [Writing](https://divergent.sh/blog): research notes

## Research

- [Repo](https://github.com/UditAkhourii/adhd)
- [npm package](https://www.npmjs.com/package/adhd)
- [Preprint](https://adhdstack.github.io/)

## Posts

- [Reading the adhd evals honestly](https://divergent.sh/blog/reading-the-evals-honestly): The harness won five of six problems. The sixth is the one worth writing about, and the ratios are the number most likely to mislead you.
- [Shared context is shared bias](https://divergent.sh/blog/shared-context-is-shared-bias): Tree-of-thought widens the frontier and carries the same anchor down every branch. Isolation is not a prompt you write, it is a property you enforce.