United StatesTechnology and telecom2025 – ongoingCognition
Devin for internal software development and code review
Cognition merged 659 Devin pull requests in a week and used AI review for every pull request
Problem
Investigating bugs and preparing changes interrupted engineering work. As agents produced more code, reviewing changes became a bottleneck.
Approach
Cognition used its own Devin agent to investigate repositories and prepare pull requests from employee requests. Staff could start work through Slack or issue trackers. The company also used Devin Review on every pull request to organize changes and flag possible bugs.
Results
Takeaway
The reported pull-request count measured accepted output. Cognition did not provide a controlled comparison of engineering time or defect rates.
Categories
- Industry
- Technology and telecom
- Business function
- IT and R&D
- Innovation type
- Process innovation
- AI technology
- Generative AIAI agents
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
Cognition described its internal development and review workflows using Devin. It reported 659 merged agent-generated pull requests in the preceding week and company-wide use of Devin Review for pull requests.
Full details are in the original sources.