Setting aside that model means something different now … MDD never really worked because the tooling never really dealt with intent. You would get so far with your specifications (models) but the semantic rigidity of the tooling mean that at some point your solution would have to part way. LLM is the missing piece that finally makes this approach viable where the intent can be inferred dynamically and this guides the implementation specifics. Arguably the purpose of TDD/BDD was to shore up the gaps in communicating intent, and people came to understand that was its purpose, whereas the key intent in the original XP setting was to capture and preserve “known good” operation and guard against regression (in XP mindset, perhaps fatefully clear intent was assumed)

If you add why you think so we might learn something.