Watsonx for i observations

Folks:

For the past couple of days, I've been playing around with Watsonx for i.

I'm no AI expert, but I do have some observations ...

1. It seems quite slow.
2. Sometimes it throws in non-RPG code. For example, for some bit
manipulation, it suggested using a '<<' operation. In another block of code
it suggested using 'IF condition THEN;' along with 'PRINT'. Very basic (the
language) looking.
3. It tried to assign the result of %LOOKUP to an indicator variable.
4. It suggested that “Instead of using fixed-form dcl-proc and dcl-pi,
adopt free-form declarations for better readability and flexibility”.
5. Conversely, it suggested using callp to invoke a procedure.
6. The summary function seems to be the most useful function.
7. I ran into token limits a few times. A very large program was too big
for it to process. This was frustrating because one of the primary things
I would use AI for is to gain comprehension of large, legacy, programs.
8. A few times it just plain errored out with no useful diagnostics. "Oops,
something went wrong. Please try again after some time"

Some of the issues seemed quite odd ... Considering this LLM is supposed to
be trained specifically for RPG, it demonstrated some rather marked lack of
understanding about RPG.

That aside, I really don't like using VS Code ... it's very awkward and
seems (to me at least) that the IBM i paradigm just doesn't fit in the IDE.

Maybe it's because I'm using Rocket/Aldon for change control and its VS
Code extension isn't nearly ready for prime time.

david