JUST GOT MY BOTTLENECK REPORT Took the AI Bottleneck Finder at charlieautomates.com/diagnose and here are my results: STAGE: The Operator PRIMARY BOTTLENECK: Burning Tokens Because the System Has No Memory Dave, you said it straight: wasting tokens AND redoing stuff with Emery over and over. That's not a tool problem, that's a prompt architecture problem. Your system is stateless. Every session starts from scratch, which means you're paying (in tokens AND time) to re-establish context that should already be baked in. HIGHEST LEVER: Build a master system prompt (a living context document) for Emery that gets loaded at the start of every session. This is the single highest lever because it kills both problems simultaneously: token waste drops when Emery stops asking clarifying questions she should already know the answer to, and your output gets consistent because the instructions are locked in writing, not floating in your head. You shift from re-training Emery every session to managing a system that already knows what you expect. QUICK WIN: Open Claude Code right now. Paste in the last 3 conversations where Emery gave you inconsistent output or you had to repeat yourself. Ask Claude to identify every piece of context you re-explained across those sessions. Then ask it to consolidate those into a single system prompt under 800 tokens. That's your Emery starter doc. Load it at the top of every future session. Do this in the next 90 minutes. What does yours say? Take the quiz: charlieautomates.com/diagnose Interesting some of the questions I donβt think I fit in the actual answers so I tried to do the best answer I could. Having other and allowing you to fill in details, might be super helpful and more accurate