AI coding tools can produce implementation faster than most teams can review it. That raises a practical question: do we need limits on how much AI-generated code we accept?
This session explores the risks of high-velocity AI-assisted development in Drupal teams: oversized changes, weak architectural understanding, shallow review, and “vibe coding.” We’ll look at whether AI-code percentage is actually a useful metric, what better guardrails might exist, and how practices like POCs, small tickets, focused merge requests, testing, static analysis can help.
The goal is not to ban AI or slow teams down for its own sake, but to use it without letting implementation velocity outrun human understanding.
