Cognition claims its GPT-6 Astra integration improves Devin’s software testing and delivery of recordings, screenshots, and test-scope reports, potentially reducing engineers’ manual code-review burden.

state: seedheat: lowuncertainty: mediumconvergesscott: mediumcoding-agents agent-verification frontier-modelsCognitionWalden YanOpenAI

What is this?

Cognition says it is integrating OpenAI’s GPT-6 Astra into Devin’s coding-agent harness, claiming more comprehensive tests, clearer reports, and easier-to-follow video evidence. Its announcement reports leading performance on an internal testing benchmark; OpenAI’s launch page quotes Cognition research executive Silas Alberti describing the improvements. Availability is ambiguous: the launch quote says integration happens on launch day, while Devin’s blog body says Desktop and CLI support is coming soon and Cloud support is being added. These are vendor claims, not independent validation, and the supplied snippets do not establish screenshot or test-scope reporting improvements, measured reductions in manual review, or Walden Yan’s involvement.

Why it matters to Scott

Cognition’s emphasis on better tests and inspectable video evidence converges with Scott’s Verification Cost position, opening a publishing comparison with Superlever’s independently validated release gates: what evidence actually lets engineers delegate review? The vendor claims do not establish independent verification or reduced review effort, so this is a consequential product-direction convergence rather than validation of his stronger claims; no supplied radar hit tracks this integration itself.

ip:concept.verification-costip:concept.mechanically-different-verifiersdev:project.superleverdev:concept.validated-release-preview-boundaryradar:coding-agent-self-report-failure-blindnessradar:proofrun-local-agent-verification-receiptsradar:concept.agent-verification

queries asked of Scott's wikis

coding-agent verification harnesses acceptance tests
human review bottleneck agent-generated code
agent evidence artifacts recordings test reports
self-testing agents independent verification trust
model selection engineering benchmarks cost per rollout

Measured heat

now 0 pts/hpeak 0 pts/hcomments 0/hmomentum: steady1 platformsage 720h

points/hour across evidence · reading as of 2026-10-12 02:59:37.977291+11:00 · deterministic, not a model opinion

How the heat travelled

09-11 16:00	⭐ origin directly observed	Cognition helps Devin test its own work with GPT‑6 Astra OpenAI on openai	—
09-12 02:20	our radar first saw it · +10.3h	discovery anchor: openai.article.e3ea4b1263030c88c18725fb	—

pace: p0 vs 1032 stories at the 336h mark (now 720h old) — behind aafp-commons-signed-agent-notebook (0.0x)

Evidence (1) — ⭐ canonical anchor

source	object	author	score	comments
🟧 openai ⭐	Cognition helps Devin test its own work with GPT‑6 Astra Retrieved article excerpt Open article · Retrieved 2026-09-12T02:20:37.552544+00:00 September 11, 2026 # Cognition helps Devin test its own work with GPT‑6 Astra GPT‑6 Astra improves Devin’s ability to test software and show that it works, with the goal of helping engineers review less code and ship more. [Start building](https://openai.com/startups/) Company size: Startup Region: North America Industry: Technology Products: API Loading… Share Cognition is the company behind Devin, an autonomous software engineer used by businesses around the world, from big banks to tech-native startups. As Cognition’s engineering teams write more code, reviewing that work has become a challenge. Cognition sees GPT‑6 Astra’s ability to test its work and show the results as a way to make code review more efficient. > “One of the big pieces that Astra improves on is its ability to test and prove that its work actually functions the way you expect.” Walden Yan, Co-founder, Cognition ## Testing software and showing the results Cognition is applying GPT‑6 Astra across its product lineup. “We’ve been using it to make improvements across our product, including the core cloud agent that is Devin, but also our CLI and desktop products,” says Yan. In one example, Devin uses Astra to test Otter Run, an iPhone game, and returns a recording of the game running in a simulator, alongside a report identifying checks that passed and areas left untested. The recording shows the application’s behavior, while the report documents the scope of the testing. Engineers can use those outputs to inspect how the software works and understand what still needs attention. Astra is also helping Cognition get back to customers “much quicker,” says Yan. When a customer sends a screenshot of a bug, the team can pass it to Devin using Astra, which fixes the issue and returns a screenshot showing the result. ## Working toward less manual code review Cognition sees stronger testing with GPT‑6 Astra as a path toward a more efficient review process. Having Devin test its work and provide evidence of the results could help engineers evaluate changes with less manual examination of the code. > “We expect over time that we have to manually look at less code and end up shipping more at the end of the day. This is one of the things we’re really excited about when it comes to GPT-6.” Walden Yan, Co-founder, Cognition ## OpenAI <3 startups [Join the community](https://openai.com/leads/startup/)[Start building(opens in a new window)](https://openai.com/startups) ## Keep reading Perplexity customer story art card [Perplexity trusts GPT-6 Astra with end-to-end systems Sep 14, 2026](https://openai.com/index/perplexity-improving-accuracy-with-astra/) Scaling Storage for 1 Billion ChatGPT Users (Part I) card image [Rapidly scaling online storage to serve over 1 billion ChatGPT users EngineeringSep 11, 2026](https://openai.com/index/scaling-storage-one-billion-users-part-one/) How a researcher uses Codex and ChatGPT to search for new antimicrobial molecules — card image [How a researcher uses Codex and ChatGPT to search for new antimicrobial molecules Applied AISep 10, 2026](https://openai.com/index/using-codex-chatgpt-to-search-for-new-antimicrobials/)	OpenAI	—	—

source

object

author

score

comments

🟧 openai ⭐

Cognition helps Devin test its own work with GPT‑6 Astra

Retrieved article excerpt

Open article · Retrieved 2026-09-12T02:20:37.552544+00:00

September 11, 2026

# Cognition helps Devin test its own work with GPT‑6 Astra

GPT‑6 Astra improves Devin’s ability to test software and show that it works, with the goal of helping engineers review less code and ship more.

[Start building](https://openai.com/startups/)

Company size: Startup

Region: North America

Industry: Technology

Products: API

Loading…

Share

Cognition is the company behind Devin, an autonomous software engineer used by businesses around the world, from big banks to tech-native startups. As Cognition’s engineering teams write more code, reviewing that work has become a challenge. Cognition sees GPT‑6 Astra’s ability to test its work and show the results as a way to make code review more efficient.

> “One of the big pieces that Astra improves on is its ability to test and prove that its work actually functions the way you expect.”

Walden Yan, Co-founder, Cognition

## Testing software and showing the results

Cognition is applying GPT‑6 Astra across its product lineup. “We’ve been using it to make improvements across our product, including the core cloud agent that is Devin, but also our CLI and desktop products,” says Yan.

In one example, Devin uses Astra to test *Otter Run*, an iPhone game, and returns a recording of the game running in a simulator, alongside a report identifying checks that passed and areas left untested.

The recording shows the application’s behavior, while the report documents the scope of the testing. Engineers can use those outputs to inspect how the software works and understand what still needs attention.

Astra is also helping Cognition get back to customers “much quicker,” says Yan. When a customer sends a screenshot of a bug, the team can pass it to Devin using Astra, which fixes the issue and returns a screenshot showing the result.

## Working toward less manual code review

Cognition sees stronger testing with GPT‑6 Astra as a path toward a more efficient review process. Having Devin test its work and provide evidence of the results could help engineers evaluate changes with less manual examination of the code.

> “We expect over time that we have to manually look at less code and end up shipping more at the end of the day. This is one of the things we’re really excited about when it comes to GPT-6.”

Walden Yan, Co-founder, Cognition

## OpenAI <3 startups

[Join the community](https://openai.com/leads/startup/)[Start building(opens in a new window)](https://openai.com/startups)

## Keep reading

Perplexity customer story art card

[Perplexity trusts GPT-6 Astra with end-to-end systems

Sep 14, 2026](https://openai.com/index/perplexity-improving-accuracy-with-astra/)

Scaling Storage for 1 Billion ChatGPT Users (Part I) card image

[Rapidly scaling online storage to serve over 1 billion ChatGPT users

EngineeringSep 11, 2026](https://openai.com/index/scaling-storage-one-billion-users-part-one/)

How a researcher uses Codex and ChatGPT to search for new antimicrobial molecules — card image

[How a researcher uses Codex and ChatGPT to search for new antimicrobial molecules

Applied AISep 10, 2026](https://openai.com/index/using-codex-chatgpt-to-search-for-new-antimicrobials/)

OpenAI

—

Interpretation history

2026-09-12T02:24:13Z
grounded: converges/medium — Cognition’s emphasis on better tests and inspectable video evidence converges with Scott’s Verification Cost position, opening a publishing comparison with Supe
2026-09-12T02:21:38Z
case created — This first-party account documents a distinct Devin integration with concrete testing outputs, but review savings remain prospective and the historical import does not warrant urgent follow-up.

Decision trace

10-10 13:48review_dormant28 days without material information; scheduled checks stopped
09-12 12:24groundCognition’s emphasis on better tests and inspectable video evidence converges with Scott’s Verification Cost position, opening a publishing comparison with Superlever’s independently validated release
09-12 12:21createThis first-party account documents a distinct Devin integration with concrete testing outputs, but review savings remain prospective and the historical import does not warrant urgent follow-up.