I finally have a working Spring transactional JPA prototype set up with multiple repositories, each with their own transaction management and connections. I needed to figure out the plumbing because that had always been set up by the time I ever worked on a project, so it was all new to me.
I started with Claude 3.5's ideas about how to set it up. It didn't work. So I used Google's search engine for suggestions as to how to make it work as pure Jakarta JPA, and got that working to some degree, but without transaction management.
Adding transaction management the same way didn't work, so I turned to GitHub copilot with OpenAI 4.x, and started asking questions about what I'd done wrong. It took three long days of querying and coding before it finally worked the way I need it to.
So these tools are utterly useless for creating anything significant unless you have the programming experience to know what to ask and the patience to slog through the process of debugging code you don't really understand.
All told it was a no hurry 6-9 week effort (I'm retired and am just coding for fun when I feel like it - no schedule, managers, or users setting artificial deadlines.)
I recently read a quoted post from the Muskoid claiming that xAI employees just paste a file into Grok and let it fix it. Nice theory, but without the context of the application, that flat out wouldn't work.
But what do you expect from an Apartheid backed con artist who doesn't know jack shit about actually writing code?