Everyone building agents hits the same wall: the system only really exists inside a context window, and the moment that window rolls over you rebuild it from scratch in a fresh chat. So I moved the entire system state onto disk and built a loop that keeps it consistent. The maps layer is the part I would point at first. Five product maps plus a workflow map, every line anchored to a path-and-line citation. A job never mutates a map. It emits a structured flag, and a separate interactive pass re-verifies that flag against the live tree before it is promoted to a map line. Isolating that verification is what stops the maps drifting into the exact thing they exist to catch. The payoff is a queryable index over the whole system that I actually trust, and that trust is most of the sanity. The loop runs on two clocks. A deterministic kickoff fires once per task: registers the active worker, regenerates the tree, assembles its context. Zero model judgement, identical every run. Independently, a continuous harvest loop runs on a thirty-minute cycle, routing each job's record into the right surface, retiring finished workers, and flagging where a folder's own contract came up thin. Those surfaces are flat markdown, and that is the point. LIVE_STATE.md is the active set. ISSUES.md, PENDING_MAP_UPDATES.md and CONTEXT_FEEDBACK.md sit alongside it, each a single-writer file with one job, committed and fully diffable. Every executor reads the folder map before it touches anything, so it grounds its working set and writes to the right path instead of inferring one. It only works because the loop can read its own state back on every pass, and that state is the filesystem, not a window that forgets. Full video walkthrough below. How have you all layered on top of your ICM setups? Curious what you've added, what extra routers or surfaces you run, anything specific to your own structure you think is worth sharing.

🦦 Otter AI Project, Building My Own Local AI Dashboard! Over the past month or so I've been building something I'm really proud of which I call Otter AI, a fully self-hosted AI command center that I built from scratch for my own personal use. The idea? A private, local alternative to ChatGPT or Claude that runs entirely on my own machine. No subscriptions, no data leaving my network, no cloud dependency. Just a powerful AI dashboard built exactly the way I want it. ✅ What's built so far: • Multi-turn AI chat with streaming, voice input, and file/image uploads • An integrated suite of terminals connected to your device such as PowerShell, CMD, and SSH with Cisco switch and Ubiquiti Dream Machine Pro support as well as other devices. • Gmail integration with AI email summaries and draft replies • Multi-persona AI debate panel with the ability to create new personas (yes, I made Moobly the Greedy Goblin and Aldric the Arcane Sysadmin to argue about how to solve my IT problems) • Deep research mode that autonomously searches the web and writes cited reports • RAG document indexing, the AI can read and reference any files I give it and past conversations. I uploaded all of my conversations with ChatGPT and Claude into it to get it up to speed. • n8n workflow automation integration • Full settings panel with 16 color themes, custom branding, and voice input • Weather widget, draggable and customizable dashboard, and a ton more 🔜 What's coming next: • Electron packaging, turning it into a real .exe desktop app • Client Manager panel for MSP work • Script library for PowerShell/bash • Image generation support • Multi-user profiles so friends can run their own instance The goal is to eventually package this up and include a how to document to setup some other dependencies so anyone can run their own private AI dashboard, no technical knowledge required. Built with Node.js, Ollama, SearXNG, and a lot of late nights ☕