I probably spend 100$ a month on AI coding, and it's great at small straightforward tasks.
Drop it into a larger codebase and it'll get confused. Even if the same tool built it in the first place due to context limits.
Then again, the way things are rapidly improving I suspect I can wait 6 months and they'll have a model that can do what I want.
I believe this. I've been having the forgetting problem happen less with Gemini 2.5 Pro. It does hallucinate, but I can get far just pasting all the docs and a few examples, and asking it to double check everything according to the docs instead of relying on its memory.
Similar to the function documentation provides to developers today, I suppose.
This helps the model to focus on a subset of codebase thst is relevant to the current task.
(I built it)
But I guess it could be interpreted differently like you said.
The promise of AI is I can spend 100$ to get 40 hours or so of work done.