The one I dropped too early
Stash was an iPhone app for the things you save to deal with later. Links from the share sheet and new screenshots landed in one inbox, Stash added previews, suggested tags and sorted them into buckets, and you went through them in review sessions. I started it in August 2025 and came back to it a few times.
I never wrote down what version 1 was. There was no launch plan and no list of what "done" meant, so I couldn't tell how close it was. Eventually I dropped it.
About two weeks later, an app with the same name and the same job showed up on the App Store, and from what I can see it's doing well. That doesn't prove Stash would have done as well. It does show that I quit without ever testing whether people wanted it.
I don't know whether Stash would have worked. I just wish I'd found out before moving on. One test I could have tried: find people who save links and screenshots, then see whether they'd use a simple inbox before I built all the sorting. I never ran that test.
Now, when I review an idea, I try to leave with a small test instead of another feature list. Sometimes the answer is to revise it, park it, or drop it. If I park it, I write down what would make me come back. If I drop it, I want a reason beyond losing momentum.
1. Who has the problem, and who pays?
Who actually has this problem? What did they do the last time it came up, and what are they using now? Someone saying an idea sounds cool is nice, but that doesn't tell me whether they need it or would pay for it.
I also ask myself if I'd want to sell it, answer support questions, and keep it running. If I don't know an answer yet, I write that down too.
2. Could a chatbot do 80% of it?
I don't mean whether a chatbot could copy every feature. I mean: could someone get most of the result they want without my app? I look at:
- A chatbot today: the exact prompt, the files it would need, and the manual work left over.
- An agent today: its tools, permissions, setup, approvals and what happens when it fails.
- A feature in software the customer already uses.
- A capable competitor copying it.
- A stronger future agent, with every improvement labeled as an assumption.
I also include doing nothing or handling it manually. If one of those is good enough, I need a real reason someone would switch to my app.
3. One customer, one channel
I pick one group I can start with and one place I can reach them. "Everyone" and "social media" are too broad to act on. Ideally, I can name ten people to talk to or point to a specific community, search, or App Store discovery test.
4. Do the numbers work?
For a subscription, I work out how many paying customers I'd need to cover the costs and reach the income I'm aiming for:
p = monthly revenue per customer, after refunds
v = monthly cost per customer
f = monthly fixed costs
t = monthly income I want from it
customers_required = ceil((t + f) / (p - v)) only when p > vI count my time separately from the cash costs so I can see what I'd effectively make per hour. And the price has to be a reasonable guess at what people would pay, not a number I picked to make the math look good.
5. What's the smallest test?
I pick the question that could change my mind and look for the cheapest way to answer it. That might mean talking to someone, trying a paid pilot, doing the work by hand, or testing a place to find customers. It doesn't have to start with code.
Before I run the test, I write down what I'm trying to learn, how I'll judge the result, when I'll check the result, how much I'm willing to spend, and what would make me keep going or stop. If it goes well, the next step is still a small one.
Ground rules
- Every claim is labeled Observed, Reported, Hypothesis or Unknown.
- No made-up market sizes, conversion rates or revenue.
- Evidence against the idea stays in.
- The answer fits in 350 words.
- "Test" means a limited experiment, not permission to build the whole thing.