Cognition applies GPT-6 Astra so Devin can test and prove its work

11 Sep 2026: OpenAI published a customer story — Cognition is applying GPT-6 Astra across Devin Cloud, CLI, and Desktop so the agent can test its work and return evidence. Walden Yan’s quotes are on OpenAI’s page. Distinct from Cognition’s SWE-2 model post.

11 Sep 2026: OpenAI published “Cognition helps Devin test its own work with GPT-6 Astra.” That customer story is the filing event. It is not Cognition’s 10 Sep SWE-2 model launch.

OpenAI: Cognition is applying Astra across its lineup — the core cloud agent that is Devin, plus CLI and desktop. The page’s through-line is test-and-prove evidence, not a new coding-model scoreboard.

Walden Yan, Cognition co-founder, on OpenAI’s page: one of the big pieces Astra improves on is its ability to test and prove that its work actually functions the way you expect. Cognition’s named problem on that page is review catching up as teams write more code.

OpenAI’s example: Devin uses Astra to test the iPhone game Otter Run and returns a recording of the game running in a simulator plus a report of checks that passed and areas left untested. Yan: when a customer sends a screenshot of a bug, the team can pass it to Devin through Astra and get back a screenshot of the fixed result. Those workflows and the “review less code / ship more” direction are Cognition’s account on OpenAI’s page — not measured review-time cuts.

A dated OpenAI customer story puts Astra on Devin’s evidence loop across cloud, CLI, and desktop. File the page and Yan’s quotes. Do not fold it into SWE-2. Treat recordings, coverage reports, and faster customer fixes as Cognition’s claims on OpenAI’s site.