Welcome to Clief Notes. Here’s where to start. 👋 You don’t need to watch the entire classroom before trying something. Pick a starting point, make something small, and bring your questions back here. 🧭 Find your way around Navigating the Course shows you what’s here and how to get around. If you’re new to working with AI, start with The Foundation. We work through prompting, context, and organizing your files so you can give an AI a job you can actually check. 🛠️ Never ran a coding agent or folder strcutre? Jump into Module 4 This is a good starting point if you want to work through a practical example. The short videos mix animation with actual app screenshots, and there’s a fictional practice folder you can follow along with. - 4.1: Install and First Use — Get oriented and try your first task. Grab the Foundation Practice download from Resources here. - 4.2: Working With Files — Turn meeting notes into a decision list, then check where each decision came from. - 4.3: Think, Then Build — Compare your options, choose an approach, and approve a short brief before asking for the finished work. - 4.4: Project Instructions — Leave useful instructions with your project so you have less to explain next time. - You’ll find guidance for Claude Code and Codex, plus a regular chat route where you save the files yourself. 💬 Use the check-ins At the bottom of the lessons, you’ll find polls and discussion posts. Share what you tried, what worked, and what needed correcting.

Thank you Jake Van Clief and David McDermott for having me, and David Vogel for the Hermes assist. This one is ICM in practice. I open my own Hermes and walk every layer of its memory, live: the built-in memory files, session history, skills and the curator, the memory provider I host myself, and the ICM workspace that holds the job itself. The part I think this room will care about: in my workspace, every folder carries a small AGENTS.md router that points to a CONTEXT.md contract. Hermes loads AGENTS.md on its own. It doesn't load CONTEXT.md, so the router is what makes sure the contract gets read. Small router, real instructions 1 step away. Watch it here: https://www.youtube.com/watch?v=GtROa4PfvY8 ICM is why this system holds together, and Hermes is what walks it. Jordan The read through https://agentsasfolders.ai/blog/ai-agent-memory-hermes/

I watch Clief’s JEV video and it was brilliant. I was hooked the whole way through. Thank you for that post. I’m now left wondering where and if it fits into ICM since a primary responsibility handled by ICM is effectively routing the AI model/harness through the file/folder structure…and JEV is a general categorizer …so in the example in the video…JEV is dumb but brilliant at sorting things into categories like data to its associated department but incapable of much else so it’s placed as a preprocessor to the LLMs for efficient and cost savings advantages. My thinking now is how do I use JEV to supercharge the ICM routing we already do? Am I thinking about this correctly? Am I missing something?

made a short on a rule i now write into the CLAUDE.md before any automation build. if u just ask Claude Code to build an automation, it will often reach for a model call at every step, bc it can. reading an email, checking a total, saving a row, sending a receipt, all thru the LLM. the split from the video. judgement goes to the agent, rules go to code. in an invoice flow only one step needs judgement, reading the messy email and pulling out the amount. checking the total, saving it, sending the receipt never change, so those are plain functions. what worked for me is listing the steps in the CLAUDE.md with a tag on each, agent or code, before it writes anything. then Claude builds the code steps as normal deterministic functions and only calls the model where i said so. the build gets cheaper and way easier to test. do u put rules like this in ur CLAUDE.md or decide it as u go.

Rung two built the window. Rung three is what survives it, open from today. There is something you have told the model three times this month. Nothing big. The client signs as Sam, never Samuel. We stopped saying "the portal" in June. It got it right every time and it forgot every time, and somewhere around the third time you stopped noticing you were the one doing the remembering. That is not the model being slow. That is a window closing, with nothing on the other side. 📚 What just opened Memory Engineering, seven posts, the whole rung at once. The Overview draws the line the rung stands on, this turn against what survives between turns, and the three classes carry one claim each: (1) Most memory fails by keeping too much (2) A memory you cannot find is worse than none (3) A memory is a belief, not a record Then the payoff, the whole series on one card, my perspective on rung three, and the closing post on where memory sits inside ICM. By the end you own a memory: one small file the room reads every time you open it, a drawer behind it for the rest, and a date in your calendar for the day you read it back. No system in it, and you will notice you never built one. 🪜 Where you are standing (Rung 1 & Rung 2 should be completed) If not, you can find them here they are in order and should be taken from series 1 to series 2 and so on: PROMPT ENGINEERING 📚 Series 1 Overview CONTEXT ENGINEERING 📚 Series 2 Overview Climbed rung two? Bring the room's file, the seven lines, because this rung decides what gets written into it, what stays out, and what gets crossed out the day it stops being true. The Overview asks one more thing before class one: the three things you type to the model in every fresh window. Three, not thirty. Not there yet? Rungs one and two are still open and still the order, the Start Here post is pinned. Honest word on this one: it is the rung people stop on, because nobody ever asked most of us to decide what should be kept. So the steps are smaller and every wrong version is shown in full. If you have never built anything that had to remember, that is who I wrote it for.