How we work

Building in the open, without the hype.

Ren AI is in active development. Our principles are the ones we can actually keep at this stage — and the ones we intend to keep as we grow.

I

Evidence over hype

We are early, and we say so. We will not publish benchmark numbers before we publish the harness that produced them. If we cannot measure a claim, we do not make it.

II

Understanding over autocomplete

Ren Code is built to understand a codebase — its architecture, dependencies, and conventions — before it changes anything. Correctness comes from context, not from guessing the next token.

III

Reliability over marketing

A confident wrong change is worse than an admitted uncertainty. We design for review: pull requests you can read, reasoning you can inspect, and control over what merges.

IV

Honest about the stage

Ren AI is in active development, building Astra — our flagship language model — and the products powered by it. We would rather show real progress than stage a finished story. The roadmap is the product right now.

Commitments

What you can hold us to today.

We publish the method, not just the result

When Astra produces a capability number, it will ship with the evaluation harness that produced it. Until then, we describe progress qualitatively and honestly.

You decide what Ren Code can see

Repository access is granted per repo and revocable. Ren Code reads what you allow, and nothing else.

Changes are made to be reviewed

Ren Code produces pull requests, not silent commits. You read the diff, you decide what merges. Autonomy with an audit trail.

Honest about the stage

Ren AI is early. We would rather show a real roadmap than stage a finished story. What you see here is what actually exists.

Get started

Try Ren Code while it is still early.

Create a workspace, connect a repository, and build alongside a product we are improving in the open.