[justinclift]

> especially with AI.

While AI models still hallucinate commonly, this might not be the best use of it.

The downsides of the occasional hallucinated answer for building standards seems like it could be pretty bad. Seems much worse than the upsides?

[kyrra]

I think it's not just the answer you get, but using AI as a stochastic search engine over the rules to point you to potential answers.