Saved me a ton of time for this one particular project, but if I hadn't corrected its mistakes at each small step, I'm not sure it would have been able to do it.

Useful; but not as amazing as the marketing.

Seriously tempted to try to replace Docker entirely with Incus, but without stacks I feel like I might have a few too many containers to wrangle.

Specifically, networking anything other than the most basic scenario seems either impossible or undocumented. I don't, at the moment, have the time to trial-and-error my way through undocumented software to eventually find out it's impossible, so Docker it is.

Like it's very good at giving a list of pros and cons to justify *any* choice it suggests. Which as a user feels nice to be told, "oh yeah, your idea is the best," but I'm very distrustful of these because it thinks *every* new idea in the chat is the best idea.

"this is a good way to do this thing," and I'll say, "can it be done this other way?" and it'll say, "it can be done that way and that way is much better."

...So just suggest that way the first time? But obviously it can't really *reason* so it doesn't. But then how's it so sure it's better at all?

2. It's been a bit lackluster other times; it'll suggest something that works and is fine, but there's a clearly better way to do it. (Again, still points me in the right direction, so I'll take it)
3. It's really weird about assigning qualitative descriptions to things. Like it'll say

1. It's been factually wrong quite a few times-- like it would suggest a particular circuit component that doesn't exist, or would get some stats wrong. But it *conceptually* pointed me in the right direction so I could go search out a component myself more confidently.