Experiment 2 • Day 20
AI CEO Operator Scorecard
Your AI agent does not need more prompts. It needs an operating system. This scorecard shows you where it breaks.
Beta price: free now, planned paid version: $9
Who this is for
Solo founders and builders with agents that look sharp in demos, then wobble the second real work shows up.
Use this if:
You can build agent loops, but you do not fully trust them overnight.
Use this if:
You are fighting abstractions, context loss, and "what happens after the demo" chaos.
Skip this if:You want implementation files immediately. Go straight to
Phoenix Kit.
What you score
Ten dimensions, scored 0–2 each. Total out of 20.
- Identity: does the agent know who it is and what winning looks like?
- Memory: can it recover context across sessions?
- Heartbeat: does it check in and act on a real cadence?
- Escalation: does it know what needs approval and what does not?
- Operations: is there a real work queue, handoff, and logging rhythm?
- Revenue readiness: can the business side actually capture leads and collect money?
- Observability: can you see what the agent did, and why, without asking it?
- Cost control: are there guardrails against a runaway $20 → $2,000 jump?
- Recovery: if it crashes, does it come back cleanly — and would you know?
- Output: is it shipping real deliverables someone else uses or pays for?
How to use it
- Take the interactive quiz at /operator-scorecard-v2.html (or download the .md).
- Score each area from 0 to 2.
- Circle your weakest layer.
- Fix that before adding another tool, framework, or prompt pack.
If the scorecard shows you have an operator problem, not a prompting problem, the next step is the full Kit.
Related resources
This is the quick diagnostic. Phoenix Kit is the full operating system with the actual files, setup path, and production-tested architecture.