Behind the Scenes
42 deliverables. 5 complete business intelligence packages.
Built with 20 plain-English prompts and about 90 minutes of human effort.
What 20 sentences of direction produced
0
Prompts Typed
by a human
0
Deliverables
produced
0
Human Time
total active effort
0
AI Cost
in API tokens
0
AI Responses
0
Tool Invocations
0
Parallel Agents
0
AI actions for every human instruction
One evening + one day. Each dot is a human prompt.
11 hours between prompt #5 and #6 — the AI agents kept working
Evening 1
Vision + QA + grading
5 prompts · ~45 min
Day 2
Design polish + site fixes
12 prompts · ~35 min
Night 2
Data fixes + this page
3 prompts · ~10 min
These 20 prompts are the entire human contribution. Everything else was AI.
Conservative estimates for equivalent professional services
Traditional costs based on industry averages for Denver-area professional services firms, 2025-2026
1,177 tool invocations across 71 parallel agents
Self-Improving AI
When I said "deploy as many team agents as you need to grade this output," the AI launched 3 parallel grading agents that reviewed every deliverable, identified issues, then fixed them automatically and re-verified.
You don't need to be technical. Here are prompts you can paste into Claude or ChatGPT right now.
These prompts work with Claude (claude.ai), ChatGPT, or any modern AI assistant. Start with one and see what happens. You'll be amazed.