the bad news is both companies are in their enshittification stage, paywalling features and trying to sell you crappy ai shit

Awh, I was really hoping this was just Docker advertising itself as good for deploying AI. Running neural networks usually means running Python, and Python package management with CUDA is so hard that (in practice) you'll just see people providing a Docker container. Some kind of "Docker AI Hub" would make sense as a product.

Looked it up, nope, it's "we'll generate your Dockerfile for you!" which is very very stupid :(

but when i've tried using podman for postmill development, it crashed, and for deploying services, i couldn't get ansible to work with it. which is weird, because ansible is also a red hat product.

Oh yeah, this was my experience too (sans ansible). I wanted to use Podman for the same reasons you listed.

Actually, according to my notes, first I tried setting up a Conduwuit server to get connected to the Matrix using Podman. So, that's two marks against it. Which is strange, since it should be the same API around the same Linux tools, where the only difference is the containers can't use root, which they shouldn't be doing anyways...