So it's great for single answers, or to get a bit of depth on a topic - but it has massive flaws at the moment for career specific things like programming.

I've used both Chat GPT and CoPilot for front end and back end engineering and it's shite for anything that consists of anything more than a personal project. It has no context of scope, or of any new frameworks that have been released in the last 5 years. It gives answers that are mixed between functional and object orientated programming. It's not consistent either, for example, it will bounce between using the composition API and the options API in the Vue framework.

When giving advice on using Google API endpoints it will often target deprecated tech. On more than a few occasions, I shit you not, if it doesn't know the answer it will just lie and make up API endpoints - it's fucking hilarious. It needs to avoid data sources from unreliable and old frameworks like Stack Overflow.

Copilot is much, much better, but is still completely clueless for anything that goes beyond the scope of a simple piece of programming. If you want a piece of global functionality on a large application it'd be absolutely clueless. Now, you could say that this is indicative of it being in it's early stages, or because GPT doesn't have any data from the last year - but I think this is more of a problem with taking data from inaccurate and outdated sources. The more frameworks and programming evolves (which at the moment is A LOT), the more outdated the AI will become. I could spot an AI project from a mile off.

However, great for small one / two lines of code and to help emulate paired programming.