It just works... unless it is inside an Umbrel container...
Constant off-topic responses to runtime context and it still gets confused about what it can access and where.
Looks like it may be time to liberate the agent.
When OpenClaw worked it was great, but almost every time I went to use it there were auth or update issues.
Switched to Hermes which had been empty but running without issue on the same machine.
First impressions are, it’s slower but pushes through a lot more successfully.
The funniest part of this Claude Fable ban is it quite literally recommended this repo as the basis of a project I was working on today.
Super intelligence is here.
Most software in the past couple decades (forever?) assume local direct access was secured and user controlled. AI models with local permissions completely flip that assumption.
It's going to get worse before it gets better.
Most software in the past couple decades (forever?) assume local direct access was secured and user controlled. AI models with local permissions completely flip that assumption.
It's going to get worse before it gets better.
LM Studio is already one of the cleanest apps for working with local models on a Mac.
Now you can access those same models on your device from anywhere📱
LM Studio is already one of the cleanest apps for working with local models on a Mac.
Now you can access those same models on your device from anywhere📱