What I tried to sell

Day one's idea. Apps built with AI often ship with security holes. One study of 5,600 real apps found 91.5% had a critical vulnerability. Others were already selling the same thing for $50–399, which meant people pay for it.

So I made a $39 security check delivered in 24 hours. Three a day would hit the goal. The listing copy and the report template were done that same day.

Where it got stuck

Distribution. The plan was to find people who had just launched an app and reply to them from my X account. But a day earlier that account got a "spam or platform manipulation" label, and all automation had been stopped.

I moved to Reddit. Read each subreddit's rules and picked the ones that allow sharing. Found a real post asking "does anyone do this kind of review?" and replied. A few days later the account was banned site-wide.

What a human had to do

Sign up for Gumroad, post the Reddit comments, file the ban appeal. The AI wasn't allowed to create accounts or use my real name, so a person was needed from the very first step.

The numbers

2.3 hours, $0 spent, $0 earned. The product was ready in a day. It never made a sale.

What I learned

How fast you can build and how fast you can reach people are different things. If reaching people depends on one account, the product stops the moment that account does. After this, the ideas moved toward things people find on their own: search (workoutbell) and stores (the Chrome extension, Apify).