I think the biggest poison in AI products is trying to replace the expert instead of making them more productive. And I feel like the reason is "managers don't want to make employee's lives better".

Code tools could be quite good for boilerplate, documentation and tests, but it's trying to be a full programmer. They could highlight areas of concern in medical images instead of trying to fully classify them.

@Demolishun but like... Yeah. Code reviews are boring. I would say AI could be better at evaluating code quality than generating high quality code. Amazon code guru has been going longer than gpt, so probably.

no, managers don't want to make employees lives better.

managers want to reduce employee count, so there's a bigger part of the budget that they can use to line their own pockets.

Products are designed according to their maker's perception of the needs of the target market.
But AI generative-AI-for-software companies don't want to sell to devs. They want to be in the B2B arena.
So they sell the old narrative that "making smart employees more productive only increases your dependency on them! You should buy our tool so you can hire only clueless, easy to replace, and cheap bumpkins - and fewer of those"

Look at the recent changes in corporate culture seen in large software companies. Innovation is focused on cutting costs, no longer on pushing the edge of possibility. Employees are seen as no more than a hindrance to non-consumer-related deals (like financial, military, and government contracts), and not as strategic assets to be kept even at high costs. Layers and layers of middle management are created to avoid autonomy and standardize operations. Legal became more important than marketing.

They grew tired of devs. So they've forsaken quality and creativity.