daften
Dieter has over 18 years of experience with Drupal. Having worked on projects of all sizes and in various technologies, there is no project he can't tackle. As technical director, he is now supporting the operations teams of Dropsolid to go to the next level.
Sessions
Description
Developer community is currently split into three uneven camps:
- The Skeptics who refuse to touch AI on principle.
- The Champions who seem to be rewriting Facebook every afternoon just to prove a point.
- The Silent Majority - those who want to use new tools but feel so far behind the "Champions" that it’s discouraging to even start.
If you belong to that third group, you’re likely suffering from an AI Mid-Life Crisis. You’ve seen the polished stage demos where AI generates a perfect migration script, but when you tried it, you got a hallucinated mess of Drupal 7 hooks and broken YAMLs.
It’s time to talk about the "Failure Stories."
This session isn't a "how-to" on prompt engineering.
It's an honest discussion on what happens when AI fails in the real world of Drupal development - and what we can do about this.
We’ll discuss why feeling "behind" is a trap, why the 10x developer myth is toxic for team morale, and how we can bridge the gap by being more transparent about our struggles.
Three years ago at DrupalCon Lille, we gave a talk on managing technical debt and how modern Drupal helps you tackle it. The core of that talk still holds: technical debt management is about business value, not code. What's changed is the economics.
AI coding assistants have made refactoring faster and cheaper. Tests are easier to write after the fact. Code review has a new automated layer. The execution cost of fixing debt dropped significantly. But the upstream work, deciding what to fix, why it matters, and explaining that to your client, hasn't gotten any easier. If anything, it's harder now that everyone assumes AI can "just handle it."
We'll walk through the Lille framework with a 2026 lens. What still applies unchanged (more than you'd think), what AI actually makes better (refactoring, test generation, MR review), and where it falls short (planning, verification, the client conversation). We'll also get into the 80/90% question: when AI output is good enough for business needs and when engineers are right to push for more.
AI still can't make the plan for a complex refactor or reliably verify its own outcomes. It doesn't replace senior engineers or solution architects. But the time you spend on technical debt remediation drops significantly, and that changes how you think about what's worth fixing and when.