The answer staring the OP right in the face.
> You've allowed the AI to train you to use the wrong word through a shared delusion, haven't you.
What an awful interpretation, phrased in the most demeaning manner possible. No, I double check everything the AI suggests. That's basic competency with the things.
> Me: I'm trying to use a cooperator in my schematic, how do I hook it up? Also what's a suitable cooperator for 3.3v logic level?
> Gemini: It sounds like you might be thinking of a comparator, not a "cooperator," in your schematic. Comparators are electronic circuits that compare two voltages [...] LM393: This is a low-power, dual comparator that can operate from a 3.3V supply. Its output can be directly connected to most 3.3V logic circuits.
They might be able to help stitch together modules (like sensor boards plugged into microcontrollers) and definitely can write code to get things going, but they fall flat on their face hard for any kind of bare bones electronics design. Like 5% success rate and 95% totally incorrect hallucinations.
The models can also output code that can be turned into a schematic through an interpreter, but there is virtually zero training data for this because humans always use and work with schematics.
It's a weak jack of all trades: it knows a fair amount about the sum of human knowledge (which is objectively super-human), but can't go deep on any one thing, and still seriously lags behind humans in terms of reasoning. It's an assistant that all book smarts and no street smarts. Or maybe: it's a search engine for insanely specific things.
Rote work, as well. Things like porting an enum from one programming language to another: past the source language into a comment and start it off with one or two lines in the target language. Dozens of tabs are surely faster than manual typing, copy paste, or figuring out vim movements/macros.