We still create a ton of end-to-end tests and run them a lot. What changed for us is the review process. While our whole team uses agents to do reviews, we also ask other team members to review the coding session itself rather than the diff. That team member can ask the coding session questions about the architecture decisions, scalability, edge cases, ..etc.
Disclosure: I'm building AQ (aq.dev), which is partly why I'm deep in this. We wrote up the session-review practice here: https://aq.dev/guides/how-to-review-an-ai-coding-session/. The practice works with any agents too, nothing tool-specific about it.
Comments
We still create a ton of end-to-end tests and run them a lot. What changed for us is the review process. While our whole team uses agents to do reviews, we also ask other team members to review the coding session itself rather than the diff. That team member can ask the coding session questions about the architecture decisions, scalability, edge cases, ..etc.
Disclosure: I'm building AQ (aq.dev), which is partly why I'm deep in this. We wrote up the session-review practice here: https://aq.dev/guides/how-to-review-an-ai-coding-session/. The practice works with any agents too, nothing tool-specific about it.