Latest

Jev is a decision model, not a chatbot. It returns probabilities instead of prose. Here's what that's worth when an LLM agent is doing the deciding, with real numbers from four days of running it inside my Hermes email pipeline.

The gap between a workflow that works in a chat window and one your business runs on isn't an AI model problem — it's infrastructure. Here's the three-legged stool; n8n for orchestration, OpenRouter for model routing, Fly.io for hosting and data, plus a private React Router v8 frontend.

Comparing two leading self-hosted AI agents — Hermes and OpenClaw — from the perspective of running them on your own infrastructure. Covers stability, auth models, memory systems, skill ecosystems, and real-world operational tradeoffs.

Hermes is an open-source AI agent that runs in the cloud 24/7 and chats with you through apps like Telegram. This post now leads with the setup I'd recommend today—Nous Portal's $20/month hosted Hermes Cloud, Composio as an MCP server, and Honcho for real memory—followed by the original Railway + OpenRouter walkthrough.

Initially I was impressed by GPT-4o-Mini's ability to read text from images, but I quickly realized that getting consistent, reliable results required some trickery.