How we built an AI agent that writes and fixes our tests

Ahmed Bashir and Priyadharshan R discuss why DevRev's automated tests sometimes passed even when the app was broken, and how the team rebuilt that safety net with an AI agent.

The session introduces the agent's work across writing product tests, selecting which tests to run, and repairing them as the app changes, with the goal of reducing test-maintenance work for engineers.

Key takeaways

• Examine the gap between a passing test suite and a working app.
• Explore agent-driven test writing and selection.
• Understand how test repair follows app changes.