In 2022, in October, I built a Gmail tool. It sorted your emails, and recommended newsletters you haven't read in a while so you could unsubscribe in one click.

The idea seemed obvious. I had this issue. And occasionally when I talked to people, they complained about cluttered inbox too. I'd even pay for such solution. So I built it in a week.

I put up a landing page and ran a cold outreach for about a month. Nobody bought it. Nobody even signed up. A few people replied and said their inbox was messy but they didn't really care.

The worst part came later. I stopped using it after a month, because it didn't actually save me enough time. I had built something I wouldn't use for myself, and then I tried to sell it.

A week spent isn't the end of the world. But I've built so many apps and tools that the weeks add up.

And with AI, you can build almost anything today, so you do. You skip the part where you ask whether the thing should exist.

So now, when I get a new idea in the shower, I run it through two questions.

The Two Questions

1. Is this something only I can do, or is it something anyone can do?

The fact that you can build something tells you nothing about whether the thing should exist. A thousand other developers could have built the Gmail tool. Nothing about my situation made me the right person to do it. I was just another one.

2. Am I building this because it’s needed, or because I'm avoiding something?

This one is harder, because the honest answer can be embarrassing.

For me, it's almost always customer conversations. I'd rather spend a week building than have three uncomfortable calls with people who might tell me my idea is bad.

Building feels like progress. Talking to users feels like exposure.

Every time I've skipped the conversations and gone straight to the code, I've built the wrong thing. Every time.

So the test is simple: if you're about to spend a week building, ask what a week of customer conversations would teach you instead.

If conversations would teach you more, you're avoiding them. Go have them.

I turned these questions into a prompt you can paste into Claude or ChatGPT. It runs your idea through both, plus a few more, and gives you a verdict: build it, wait, or kill it.

It's at homeby3pm.com/tools/ideakiller. Free, takes 90 seconds. Would have saved me a week in 2022.

You can build almost anything. The skill is knowing which things to leave alone.

What are you about to build that you probably shouldn't? Reply and tell me. I read every one.

See you next Thursday.

— Michal

P.S. I have a Notion document called "Probably Not" for ideas that fail both questions. It has forty-three things in it. That's almost a year of stuff I didn't build.