[RajT88]

I feel like self-hosted voice and chat assistants - assuming we can get them to be good enough - can help fight this:

> The computer + the internet is the worlds greatest fidget spinner. How many times have you wanted to "quickly" check Wikipedia, e.g. reading up on a random town then find yourself down a 2 hour deep wikihole reading about plants, plate tectonics, and slavery. Hours and days lost to it.

If you want to know a thing, we should be able to have our computers fulfill that request without any extra drama or engagement bait. Self-host voice assistants require quite a bit of assembly, self-host chat assistants a bit less. We're moving in that direction.

[MisterTea]

Too much tech. I think the focus should be non-tech aides to curb tech use. Otherwise you risk yak-shaving and you're back to fiddling with tech.

[RajT88]

The current self-host infra is absolutely a yak shaving exercise.

My point is, it doesn't have to be once the various solutions mature. Jesus it is irritating to get all the different stacks booting as services on Windows. So many of them want to install per-user, which is your first challenge. Then the next one is getting them to boot headless consistently. I had a few working, which now are back to non-working.

They could be designed to install and run natively as daemons or services, but just aren't. Even running as a desktop application, they all have API support available on localhost. (But most don't let you bind to an IP/custom port - you have to install a reverse proxy to expose them across the network, lol).

Things could be better, but they just weren't designed to be. I'm talking stuff like: ollama, invokeai, openwebui, LM Studio, faster-whisper, and so on...

[MisterTea]

Take a break from this stuff. Go outside. I got burnt out from focusing on tech projects too much. I took a break. It's nice.

[RajT88]

Can't take break from AI. Will get laid off.