You could prob feed it the most recent USPAP pdf.
It doesn't take it long at all to parse the entire 2024 standard - yes I already purchased the electronic PDF version of the standard. ChatGPT converts everything, including standard by standard in to executable Prolog.
Prolog can become a nightmare for many programmers to understand, because it uses backtracking - unlike any other programming language. However, when it comes to standards like USPAP, the logic is relatively straightforward. The Prolog primarily consists of statements of "fact" or "requirements" for each of the standards and the preamble. These are pretty easy to understand.
A real standard, with tight requirements covering the details of appraisal, would start to delve into more complex logic, involving numerous ifs, ands, and buts. Additionally, you have backtracking and a technique called
unification. When I developed the Prolog language "DataBridge" for Red Pepper, and they had training for it, I would say we wound up with just one programmer from India who was bright enough to do advanced coding with it. Others could handle the bulk of coding required to embed SQL statements into the code. However, there is typically a relatively small percentage of programmers who can handle the more complex logic. However, note that I only had backtracking in DataBridge, not unification. I am pretty sure that adding unification and using it in the code would have been even more difficult for the in-house programmers - so, it was a good thing I left it out. When PeopleSoft bought out Red Pepper, they finally decided to replace DataBridge code with their "PeopleSoft" code, which had similar embedded SQL constructs. The reasoning was that PeopleSoft was a very large software company at the time, with PeopleCode programmers worldwide. It would have been a significant effort to train them in DataBridge, so they converted it to PeopleCode instead. But I remember the DataBridge scripts we had: They were so straightforward to read (since so much of the logic was hidden in the backtracking).
It took ChatGPT less than 20 minutes to do everything initially, including a proposal for a new concise USPAP. And I have, of course, all the USPAP requirments as executable Prolog code.
