Post #1396622
2026-02-06 12:04 UTC
@dimillian If you have the model both implement and test the ticket, how do you know the result is actually correct? Where I work, we make a point of having a separate QE actually verify the ticket and close it. Fresh pair of eyes and all that.
Replies (1)
-
@dimillian@mastodon.social 2026-02-06 15:12
@slothdude Using a different model is good, but even fresh agents/context with different rules can do the job. The review tooling is the bottleneck now, IMO. We're not there yet with that.