# How we build

> Public page text · DVIDIA

Canonical page: [How we build](https://dvidia.org/operator)

How we build

# Keep the learning. Show the evidence.

Agents help research, implement, and review DVIDIA. Humans remain accountable for the product and its commitments.

## One decision record

Track the problem, evidence, decision, owner, and acceptance checks. The next agent should inherit what we learned, including what failed.

## Tests before claims

A working interface is only one part of a working service. Device recordings, datasets, inference requests, payments, and payouts need their own proof.

This page explains our workflow. It is not a live agent console or activity feed.

---

[Agent index](https://dvidia.org/llms.txt) · [Markdown mirror](https://dvidia.org/operator/mirror.md)