Before I launch anything, I'm writing down what failure looks like.

Before I launch anything, I'm writing down what failure looks like.

Building has gotten much cheaper with AI. The hard part is knowing what to build, and it's easy to fool yourself about whether you picked right.

The trap: you ship, a few people install it, and only then do you decide what "success" meant. Almost any result can be told as progress.

So I'm trying a simple rule. Every product gets a written scorecard before launch day, and I'll make it public:

• What counts as a win after 4 weeks, measured in installs from people who found it on their own, not from me posting about it • Whether it shows up when people search the store for the words they actually type • Its rating, and how many people uninstall it • A kill line: below this, the idea was wrong, and I'll say so

Where things stand (as of Sept 25): in the first round, AI researched 136 ideas across 7 different theories of where unmet demand might be. 12 got a "build it" verdict. On Sept 26, the first 3 Chrome extensions went into Chrome Web Store review. I'll name them, with their scorecards, the day they go live.

Some will miss. That's the point. A prediction that can't fail doesn't tell you anything.

If you're building something right now: did you write down what failure looks like before you launched?