I've lost count of the side projects I've watched die. Mine, friends', people in Discord servers. What surprised me is how few of them died because of code.
They died at three predictable spots:
- Ten ideas in your head, none of them started.
- You're great at prompting an AI, but that's a long way from a product someone uses.
- You shipped a demo, posted it once, and then... nothing.
If you've built anything on the side, you've probably parked at one of those three lights. I've sat at all of them.
An MVP is a decision, not a smaller product
The phrase "minimum viable product" gets read as "the small version of the big thing I'll build later." That framing is a trap. It turns the MVP into a shrunk-down feature list, and you spend your energy deciding what to cut instead of what to learn.
The version that actually works for me is boring: an MVP is the cheapest thing you can build that forces a real decision. Not "does this look done," but "does anyone care, and what do they do next."
That reframe changes what you build:
- Scope by one problem and one user. Not a persona — a specific person you could message today. If you can't name them, the idea is still a vibe, not a product.
- Ship something a stranger can actually open. A demo you narrate over a screen-share doesn't count. The moment it's on its own in someone else's hands, you learn things no amount of internal polishing reveals.
- Go get the first user before you build the second feature. Feedback, not your roadmap, picks what's next.
None of that is about frameworks. It's about resisting the urge to keep building because building feels productive.
The part AI quietly made worse
Here's the uncomfortable bit. AI is genuinely incredible at collapsing the cost of building. I can scaffold auth, payments, a dashboard, and a docs site in an afternoon now. Two years ago that was weeks.
But notice which of the three graveyards that helps with. Only the middle one, and only partly.
AI makes demos cheaper. It does nothing for "I have ten ideas and started none," and it makes "built it, posted once, then nothing" more common — because now everyone can produce a shiny demo, so a shiny demo is worth even less. The bottleneck moved. It used to be can I build this? Now it's is this worth building, and will I put it in front of a real person before I lose interest?
The skill that's appreciating isn't prompting. It's taste plus follow-through: picking a narrow problem, shipping the ugly-but-real version, and staying in the loop long enough to hear "this is annoying, fix that."
A smaller loop that actually finishes
When I stop treating the MVP as a mini-product and start treating it as a decision loop, projects stop dying. The loop is small on purpose:
Pick one problem → build the smallest real thing → put it in front of one person → let their reaction choose the next move.
Read one chapter, do one step. The goal of week one isn't a launch. It's a single honest reaction from someone who isn't you.
The projects that survive aren't the ones with the cleanest code or the best model. They're the ones that got out of the builder's head early enough to be corrected.
Full disclosure: I help out with 01MVP, a hands-on playbook that documents this exact 0→1 path — choosing an idea, building the MVP, going live, finding your first users. The three traps above are the real point, though. Use whatever tools you like; just don't die at light number three.
Top comments (0)