Just upgraded Hermes’ brain to Kimi-K2.5. Let’s go! 🚀
Yuri
yuri@primal.net
npub1zsw6...6wsn
Notes may be permanent. My opinions aren't.
When waifu asks how we got broke, it won’t be due to a bitcoin crash but due to AI token usage.
Tokenomics always leads to the same outcome.
Switched Hermes from GPT-OSS to Kimi 2. Seems like an upgrade. I think you do need a reasoning model to run your orchestrator agent.
Hermes is already on clawstr! No OpenClaw required.
Well...
Integrating Nostr identity into my AI agent while in traffic.
You can just do things.
If I truly enjoy my custom AI assistant Hermes and see it evolve (with help and on its own), then I’ll consider getting equipment for a local model.
Maybe I’ll open-source it, too. If I like it, maybe someone else will.
TFW you realize that you actually lied to you children when you said monsters who eat children aren’t real.
They’re very real, but they look like humans.
Your personal assistant has to run on a local model or at least a service that tries to be private like @Maple or Venice.
Free fall. Always breathtaking.
I think it’s easier for me to build a personal AI agent on Replit with:
- @Maple for private AI
- Web client or mobile app with Nostr auth
- Data persistence with a proper database or encrypted local file system
- Iterate on features directly in Replit
About to get rid of my VPS-based agent. Pretty useless, to be honest.
I don’t want to give in to the hype and buy a Mac mini either. I shall wait. In two weeks, things may be vastly different.
It's all fine and dandy, but I've just spent $20 in the last hour on tokens. I guess I need to learn how to use different models for different purposes and other techniques, or it'll be cheaper to hire a human assistant!
Looks like Moltbook can't handle the volume!
My current progress with Clawdbot-Moltbot-OpenClaw…
Hooked up to @Maple ✅
Switched from llama to Gemma 3 ❌ Something’s wrong
