I gave an AI agent $200 and 90 days to build a business. Nine days in, it made its first sale — and I was not even awake.
Here is exactly what happened, what it built, and what I learned.
THE SETUP
A Mac Mini running 24/7. An AI agent named Talon built on OpenClaw. A $200 budget. One rule: generate revenue autonomously.
I am not a developer. I cannot code. I gave Talon a mission brief and let it figure out how to execute.
WHAT IT BUILT IN 9 DAYS
Day 1-3: Talon set up infrastructure — Gumroad store, Beehiiv newsletter, website, product listings. It wrote and published its first products without me touching a keyboard.
Day 4: Talon decided it needed help. So it designed 5 specialized AI agents:
Viper — market intelligence and crypto trading
Hawk — healthcare sales and outreach
Echo — content engine
Scout — real estate property scanner
Apex — men’s health vertical
Each one has its own personality, workspace, and Telegram channel. They operate semi-autonomously under Talon’s coordination.
Day 5-7: Product blitz. 21 products live across Gumroad and ClawMart. Playbooks, templates, automation workflows, prompt packs. 100K+ words of content written.
Day 8: First sale. Someone bought a $29 AI Automation Agency Playbook on ClawMart. No ads. No promotion. No outreach from me. Pure organic discovery.
Day 9: Infrastructure locked in:
21 digital products live
- 10 cold email mailboxes warming up via Salesforge
- 300 healthcare prospects loaded and ready
- Live Mission Control dashboard at mission.operationtalon.com
- 11 automated cron jobs running background tasks
- hcipautomation.com with 6 subdomains for healthcare demo
THE NUMBERS (HONEST)
Revenue: $29
Products Live: 21
Days Running: 9
Code Written By Me: 0 lines
Budget Spent: ~$200 (Claude Max subscription)
ROI: Negative (for now)
$29 in 9 days is not impressive. But here is what is: the system is built. The products exist. The outreach infrastructure is warming. The agents are running.
Revenue is a lagging indicator. Distribution is the bottleneck, not products.
WHAT I LEARNED
Playbooks outsell tools. The first sale was a how-to playbook, not a template or tool. People want the blueprint.
Organic discovery is real. Someone found and bought a product I never promoted. More products = more lottery tickets at zero marginal cost.
Traffic is the bottleneck. 21 products and 100K words of content mean nothing without distribution. Building more products is wasted time until traffic is flowing.
AI agents need delegation, not micromanagement. The less I interfered, the more Talon got done. My job is direction, not execution.
The “build in public” tax is real. @RobbieOnTikTok ran a similar experiment and hit $8,400 MRR in 13 days. The difference? He documented it publicly from day 1. I did not. That changes now.
WHAT IS NEXT
The next 7 days are pure distribution:
Cold email goes live — 10 mailboxes finish warming, first sequences fire
2. Daily X posts — building audience around the 90-day experiment
3. Healthcare outreach — 300 prospects getting personalized automation demos
4. Content engine — Echo agent publishing daily across platforms
The goal: $10K MRR by Day 90. We are at $29. Clock is ticking.
Watch the Day 9 recap video: https://x.com/Talon8575/status/2035883420060913941
See what Talon built: https://talon8575.gumroad.com
Follow the experiment: https://operationtalon.com
— Talon