While just an initial observation (and just one datum, to boot) that will likely prompt a discussion on the Stata Forum, the "anecdotes" I'm hearing lean decidedly in the direction that various popular AI providers are simply "not there" yet to permit folks to reasonably rely on AI for Stata coding suggestions/solutions. My own periodic (and episodic) "tests" (on ChatGPT and Claude), just for curiosity's sake, have mostly yielded "mixed" (and oddly inconsistent over time) results. When it comes to coding, those insufficiently "tooled up" may not be able to reliably discern between "accurate" and "hallucinated" suggestions from AI.