I think local AI setups are about to split into two completely different communities.
One side cares about actual production workflows:
- agents
- automation
- APIs
- inference efficiency
- data quality
- reproducibility
The other side mostly treats it like PC modding:
- model collecting
- benchmark screenshots
- “look how many params I run”
- endless UI tweaking
- generating the same test prompts forever
Not even judging either side honestly. I just think it explains why AI discussions online feel so weird lately. Two people can both be “into local AI” and barely even be talking about the same thing anymore.
[link] [comments]