I'm in a similar boat. It's crazy the comments. I'll review code and make suggestions, but it's for a coworker who has surrendered their critical thinking. I'll get back the next revision with yet more code comments explaining the changes "why they were necessary", context that literally does not matter. The code comments just are a chronological summary of the code review passes. I hate it.

Yes I utilize AI for writing my own code as well, as a whole I feel it makes me slower but produce more durable output, unreasonably so. But durable in ways I don't think will matter, edge cases.

Just the other day I had to explain to it why making a fundamental change to be protected in case a botnet with 600 thousand unique IP addresses hits the endpoint in a single minute was stupid. All of this to hedge against a scenario where an actor can exhaust the cache and evict other values.

As if our infra can even tolerate that.

I hope the bubble pops soon.

I'll tell you right now that I'd rather ask an LLM what's wrong with my code than deal with the toxic userbase of stackoverflow (and I say that as someone who never actually got flamed on that site).

It can definitely be a crutch but that's an exercise for the user to learn how to not become reliant on it, any more than becoming reliant on an expert by constantly asking them questions.

Learning programming, maybe, maybe not. But I have had success being like "assume I already know Ruby, Python, JavaScript, and Haskell. What are the quick pointers and cheat sheets for using R, especially the ggplot2 library"

I could read an R tutorial, but I don't need to learn about variables and what comments are or how to call a function. I just want to know the differences between what I'm already used to.

And I'll admit it was useful to already have that context to throw follow-up questions into when they came up, like "what's the difference between = and <-" which are famously hard questions to Google

Precisely. AI by itself can be useful, the same way spellcheck or autocomplete can be. The root of problems related to AI are that managers overwhelmingly demand that AI be used to increase output at the expense of workers. Say your boss tells you that you need to double output. Will you be paid twice as much? Will you get to work fewer hours, without a loss in pay? Of course not.

Secondly (and this is related to the first point) managers overwhelmingly have unrealistic expectations (or pretend to) regarding AI's capabilities. AI can write code to do a thing, but it cannot do the crucial work of planning how to write the code to make it maintainable. Managers rarely do the actual work that their employees do (whether that's scanning groceries or software engineering) and tend not to value or understand it.