What's professional? It's easy.

?

A project changes hands - either the client decides they want to work with a new integrator or the property is sold and the original integrator is a distant memory.

A new integrator comes onto the scene and, for argument's sake, the project is lacking in the documentation that we all take great pains to provide (maybe it /was/ provided, but it's been lost/thrown away).

There's an intermittent fault on the keypad - perhaps it's always been like that.? The integrator looks at an unbranded enclosure with an Arduino in it connected to the keypad, they check for basic wiring faults, and... now what?

If we take things to an extreme where ChatGPT generated code, it was pasted into the Arduino IDE and uploaded, tested on-site and it appeared?to work. (Yay!) Even the person that installed it may not understand how it worked.

?

My argument to that isn't about "can you do xyz" (that's often just a function of time/money).? It is about whether "you should do xyz" in a supportable, maintainable way. If the client understands that they're getting code that the person (being paid for their services) may not really understand; well fine - that's on the client (provided they WERE given full disclosure).? That's why customers pay a premium for products like Crestron - because there's a massive support infrastructure and integrator community behind it (admitting it's not perfect, and sometimes even dysfunctional).

?

As I said before, I'm not knocking the solution at all or suggesting it wouldn't work.? I'm not arguing about the quality of gear either - whether a keypad is $25 or $400 isn't the point.

?

Going back to my original point; which is about ChatGPT (and similar) giving a false sense of security to developers (in general - not exclusively automation integrators).? I feel (again, my personal opinion) that LLMs shouldn't be used as a substitute or crutch for actual understanding.

?

Sorry - I've gone *way* off topic.? I'll mute myself now :-)?