[solraph]

I have found Copilot to be useful enough to use if someone else pays for it. In some cases it was useful (writing Golang functions that perform basic SQL CRUD functions). Others times, not so much.

As I said above, I'm not claiming it's a massive increase, only that it is there. If AI moved the productivity needle 2%, and a given market has 100,000 software engineers, then that's possibly 2000 devs that are spamming the heck out of every job out there.

> Either it's complete PR BS (most likely), or C-level execs are putting the cart before the horse - preparing for "productivity gains" that haven't actually materialized.

All of the above is also a real possibility.

[kimixa]

Honestly, I suspect the industry I'm in (System Software/Driver Development) is a poor fit for the current state of "AI Assistants" - there's too much specific that's just not got the depth of training data to "learn" from, it's a mature system so not much "boilerplate" needs to be generated, and much of the real complexity is managing the state of the hardware interface and expectations from a constantly-changing documented-mostly-in-some-hardware-engineer's-head-only specification. Or time debugging why the stated specification doesn't actually match the behavior of the hardware.

I don't think anyone in our group who was asked to trial it had much positive to say about it in that situation.