Robert Allen

I’m adding Mostro to COINCUBE!

Yeah, it seems to always use the Kimi K2 model when I tested it. I think the routing logic is not as good as expected.

Two steps forward, one step back. I first set up a Mac Mini and that was going okay after I figured out my LLM fallback (was hitting rate limits with Claude), but then I noticed I spent $100-150 over the weekend so I decided that was not the way forward.

Which LLM are you using?

I’m building an OpenClaw which I have named “Gandalf.” It is running on a Mac Studio (M2, 64GB RAM). I have installed Kwen3-Coder-Next and have that as the primary LLM: https://dev.to/sienna/qwen3-coder-next-the-complete-2026-… I’m trying out Cognee for graph-based memory: https://blog.dailydoseofds.com/p/openclaws-memory-is-brok… I added a Ralph Mode skill: https://clawhub.ai/richginsberg/ralph-mode He just created a plan for a feature and is proceeding to tackle it in Ralph Mode with the goal of creating a PR for me to review in the morning when I wake up. Kwen is not as good as Claude so far, but with Ralph mode, it shouldn’t matter too much. Supposedly this Qwen model is close to Claude Sonnet 4.5 in terms of coding ability. Either way, I’m pretty sure agents with local LLMs are the future. Now I’m thinking about buying a kitted out Mac Studio… 🤩

Same. I deleted Twitter from my phone and stopped watching YouTube videos (for the most part).

Nature is healing.

Wow, that is another level. 💪🏼

Not enough to worry about it.

😂

Feeling like this needs to be once a quarter moving forward!

Thanks. 😊 I used to just water but it was always rough for me around day two. Like my muscles would start to cramp. This time feels better. I’ve done lots of low carb/intermittent fasting over the last few years. So, I believe my metabolism is pretty well fat adapted.