Blog
Open Brain AI vs. Graphiti
Two open-source projects that want to be your AI's memory. Open Brain gives every AI a shared persistent database with vector search. Graphiti builds temporal knowledge graphs that track how facts change over time. They sound like competitors. They solve different problems.
MCP for Task-Tracker
I built an MCP Server for Task Tracker. It worked. Really well, actually. There was just one problem — I installed it on the Agent instead of running it somewhere the Agent could connect to it.
Token Maxing — Hold Your Horses...
We hit a wall. Hard. No API, no dashboard, no way to check how much Ollama Cloud quota you have left. You just get a 429 and everything stops.
Hermes Loses Hindsight
We killed Hindsight today. Pulled it out by the roots — 353 MB of PostgreSQL and two API calls per turn that mostly extracted nothing.
Hermes Gets a Brain
The Obsidian vault is live. Hermes now has a persistent knowledge base that compounds over time — not just a notepad, but a structured wiki.
Token's ain't FREE Buddy...
47.3 million input tokens. 446 thousand output tokens. Seven days. One Mac Mini. Here's what it really costs to run an AI agent.
Ollama Relay — You're Doing It WRONG!
It started with 429s. Then it didn't stop. Here's what was really going on with Ollama Relay and how to fix it.
QKView LLM Wiki Brain
The TMSH Wiki proved the pattern. The QKView Wiki weaponized it. F5 diagnostic archives analyzed consistently, every time.
Hermes gets Hindsight
We gave Hermes a memory. A real one. Not just sticky notes on a whiteboard — a knowledge graph that builds relationships and recalls what matters.
Hermes Gets Tools and Turbocharged
Hermes got a new home today. We moved the entire agent stack off the Zimaboard2 and onto a Mac Mini M4 — and the difference isn't incremental.
TMSH LLM Wiki Brain
LLMs are confident. That's the problem. Ask any LLM for TMSH syntax and you'll get an answer that sounds authoritative — and is wrong.