Hey everyone,
I’ve been tinkering with different local setups for autonomous agents lately, and I’m curious how others are experiencing it.

For me, the biggest pain point hasn’t been the model itself it’s the “agent logic” going rogue. Sometimes it over-optimizes something totally useless, sometimes it just loops forever, and sometimes it does something smart and I have no idea why it worked that time and not the last ten tries.

So I’m wondering:
What’s the biggest challenge you’ve personally run into when playing with autonomous agents locally?

Is it:

• the planning loop?
• tool usage?
• memory going wild?
• debugging the chain of thought?
• or just compute limitations?

No right or wrong answers I’m just trying to see what problems people here are actually facing so I can sanity-check whether I’m the only one fighting these weird edge cases.

Looking forward to hearing your chaos stories. 😅