I spent a long time building How Are You?! An Android app that quietly watches over an elderly person living alone. No wearable, no button to press, no "are you okay?" popups. You install it once on their phone and forget it exists. It learns their normal day and if it detect anomaly in the behavior, like not moving in the park, to send notification to somebody.
I was proud of it. Technically it's the best thing I've made.
And I give up. I want to write down why, because I think the reasons are more useful than the app ever was.
The thing I couldn't promise
The whole point of an app like this is one promise: we'll notice if something is wrong, 24 hours a day. That's the only promise that matters. A monitoring app that misses the one night that counts is worse than no app at all, because the family relaxes and stops checking in themselves.
And I could not make that promise honestly.
Not because of my code. Because of the Android devices. Especially the cheaper ones that elderly people sometimes own.
Some Android devices kill background apps aggressively to save battery. Some of them do it silently, days after you install. You do everything right, you get every permission, and then one morning the manufacturer's battery manager just decides your app has been running too long and shuts it down. No warning. The family thinks everything is being watched. It isn't.
I fought this for months. Recovery workers, push wakeups, watchdogs, family emails with two-step fixes, dead-man's switches. I got it very good. But "very good" is not "always," and for this app the gap between those two words is a person lying on a floor while an app that was supposed to be watching sits frozen.
Once I admitted I couldn't honestly promise the one thing the product exists to promise, the rest of the decision got easy.
The part nobody warns you about: you can't sell where it's needed.
The part nobody warns you about: you can't sell where it's needed
Here's the cruel joke. The people who need this most gather in specific places online: caregiver groups, elderly-care communities, dementia support forums. Every one of those places, for good reasons, bans promotions. You cannot walk into a room full of exhausted caregivers and say "hey, I made a thing." And you shouldn't be able to - those rooms would be a swamp of spam otherwise.
So the product that could genuinely help those people is not allowed to reach them. Meanwhile the places where you can post are full of other builders, not customers. You end up shouting your safety app at an audience of people building their own safety apps and trackers.
People can disappoint
I'll be honest about something that stung. When you try to promote something, a lot of people say they want to help. They nod, they're encouraging, they tell you it's a great idea. Very few actually do anything. I gave a lot of free advice and free help to other people along the way, wanting nothing back, and I kept expecting the same in return. That's on me. It made me feel naive. Most people that wanted to help, actually did it because they wanted something from me, for their own apps, or money. I got so annoyed, that every message that I got, I thought it is the next scam proposal, not believing anybody. And this is a pity.
I don't regret being that person. But I've stopped expecting the world to be made of that person. If you go into this needing sincere help from strangers to survive, plan for the version where it doesn't come.
The AI thing
People have a reflex now: if they smell AI in your work, they dismiss it. Not because the work is bad - often they haven't even looked - but because "AI" is enough of a reason on its own.
I used AI heavily to build this. I'm not going to pretend I didn't. And I'll say plainly: not all AI-generated code is slop. If you know what you're doing. if you can read it, break it, test it, and throw away the bad parts. it's a real tool and it makes you faster. The slop comes from people who can't tell the difference, shipping whatever the model spat out. That's a skill problem, not an AI problem.
But the reaction is real, and you have to plan around it, fair or not. It's just one more thing tugging against a solo builder.
AI amplifies what you already have. Both, your strengths and your weaknesses.
Here's the real lesson, and it's the oldest one in the book. Literally — it's in every book about building products, and I skipped it anyway. Always start with your customer. Find it first, before you do anything at all!
I had a good idea. Genuinely good. But a good idea with no specific customer is a hobby, not a business. I did it backwards. I fell in love with the problem, then the architecture, then the clever solutions to the OEM battery wars, and somewhere in all that I never actually sat down with ten real families and asked them what they'd pay for and what would make them trust it.
The order the books tell you and I now understand why is it like this:
* Find the customer first. An actual person you can talk to;
* Talk to them. Understand the problem in their words, not yours;
* Do the boring analysis. SWOT. Who else solves this, why would someone switch, what could kill you;
* Write down requirements from their needs;
* Figure out how you'll reach more people like them;
* Then, and only then, write code;
Code is the last thing you do. Not the first, not the fun middle. The last. I had it as step one because writing code is the part I love and I'm good at. That's exactly the trap.
Nobody care about how good the code is written
My app is well-architectured. Clean layers, real tests, careful handling of a hundred edge cases. And here's what I learned: nobody cares. You cannot promote good technical skills. Users don't buy architecture. They buy the feeling that their mother is safe and they can sleep tonight.
Everything I'm proud of is invisible to the person who'd pay for it. What they can see is the promise, and I couldn't keep the promise, so the beautiful engineering underneath was worth nothing to them.
With thousands of apps launching every single day, 'it's really well built' isn't a differentiator. It's table stakes nobody checks. Beyond a small amount of skill, a lot of who succeeds and who doesn't is honestly luck and reach - being in front of the right people at the right moment. The good ones drown out just as easily as the bad ones.
So here is my advice
Slow down. There's no prize for shipping code this week. Find one real customer before you open your editor. Have the awkward conversations. Do the unglamorous SWOT and the requirements doc that feels like a waste of time. Figure out how you're actually going to reach people, because if you can't reach them, nothing else matters.
I'm not bitter about it. I learned more from this failure than from anything that worked. I just wish I'd learned it in a planning document instead of in half an year of my life.
Build the customer first. Build the code last.
Top comments (0)