Vitamins Got Free. Painkillers Didn't.

You shipped the app. Usage is fine. The money isn't. So now you're lining up the next batch of features. Stop. You don't have a feature problem. You built a vitamin, and AI just made vitamins free.

When I started Vermillion, my whole business model was "if you build it, they will come." I'd never have said that out loud. It sounds stupid out loud. But it's how we worked. We obsessed over the stack, the architecture, whether the scroll hit 60fps. We never once sat down and asked whether the thing was worth a dollar to anyone.

We built good software. Some of it made nothing.

I didn't have a name for what was missing. The name is product-market fit. It cost me years and real money to learn that it isn't a marketing problem you get to after the build. It's the reason the build exists.

The thing that changed in 2026

Product-market fit is the oldest advice in startups. I know. You've read the essays. So here's why it went from one priority among many to the only one that counts.

Building used to be the gate. Slow, expensive, and that slowness protected you. Spending four months and a chunk of runway forced a few hard questions before you got in too deep. The gate is gone now. A founder opens Lovable on a Saturday and has a working app by Sunday. What used to cost four months costs a weekend.

So everyone is building. And almost everyone is building a vitamin.

A vitamin is the nice-to-have. The clever idea you've always wanted to make. The feature you'd personally enjoy. AI is very good at vitamins, because a vitamin has nothing hard to satisfy. It just has to exist and look decent. The founder ships it, posts it, and then sits there confused about why the revenue never shows up.

It never shows up because nobody was in pain.

AI made vitamins free. It didn't touch the price of a painkiller. A painkiller solves something people already spend money, time, or sanity on, and finding one of those was never a coding problem. So it never got cheaper.

How we figure out which one we have

This is the part founders skip, because it doesn't feel like progress and there's no screen to show for it. It's thinking on paper instead of burning tokens. Three tools, depending on where you are.

The Value Proposition Canvas comes first, before a line of code. One job: tell you whether you're holding a vitamin or a painkiller. You map what the customer is trying to get done, what hurts, and what they want, against what you're offering. Then you go talk to people who have the pain. Not to ask if they like your idea. People lie to be nice. You ask what they already do about the problem, what it costs them, and what they can't stand about it. Do this and you can kill a bad idea on paper for the price of a few coffees, instead of killing it in production for the price of a quarter.

The Lean Canvas is for when there's already smoke. Someone's paying for a workaround. Paying for a competitor and complaining about it. Paying too much. Paying in time and duct tape. There's a signal, and you're hunting the fire. The box that matters most on that canvas isn't the product. It's distribution. It makes you work out, on paper, how this thing reaches a human being, before you spend a week building something with no path to anyone. You can run three Lean Canvases in an afternoon and throw two away. That's the point. Throwing them away is free. Building them isn't.

The Business Model Canvas is for a company that already has fit and is now fighting in the margins. The questions move up the stack and turn into engineering questions. Do we rework the architecture so an LLM can code against it without making a mess? Do we give two engineers real ownership of the parts that touch money? You can't paste those into Claude and get an answer back. Someone senior decides them and lives with the call.

The bridge nobody draws

Most product-market-fit advice stops here, at the worst possible spot. "Go talk to your customers" is true and almost useless on its own, because it leaves the founder holding a pile of notes with no idea what to build on Monday.

The canvas isn't the deliverable. It points at one thing. Your North Star. The single needle that, if it moves, makes the business better this quarter. And moving that needle is more and more an engineering job.

We run Vermillion on EOS and the Traction system. The unit that matters is the quarter. Long enough to shift something real, short enough that nobody drifts. Ninety days, one North Star, everything else demoted. There are always ten competing priorities. There's always one that outranks the rest. The skill is naming it, then having the nerve to ignore the other nine.

What a focused quarter buys

Years ago we picked up PicklePlay. They had real traction and a growth curve that had gone flat. We took one needle and owned it. In 68 days their user base went from 5,000 to 80,000. Same product. We didn't bolt on features. We found what was stopping growth and moved it. That partnership ran four years and carried all the way through their acquisition by Universal Tennis.

That's what a North Star looks like when someone owns it. These are the others I've watched a single quarter move:

• A ten-point lift in checkout or paywall conversion. On a real revenue base that's a six-figure swing, sometimes $200K, with no new app. Someone decided it was the only thing that mattered and went hunting for the friction.
• An onboarding flow you test and tune properly, instead of the one you shipped once and never looked at again.
• Dead code and dead features cut out, so sprint velocity climbs by half because the team stops dragging last year's idea behind every commit.
• Bug reports down a third. That's churn and support load you stop paying for.
• An App Store rating off the floor, 3.4 to 4.4, which changes install conversion on every impression you're already buying.
• The founder out of the codebase and back in the room where the next round gets raised.

None of those is a feature request. Every one is a North Star a senior team can own end to end. That's the work that doesn't commoditize, because it needs someone who knows the business, not just the code.

The rule worth keeping

In 2026 the constraint is never can we build it. You can build anything. The constraint is whether it's a painkiller, and what single thing proves that this quarter.

Use AI to build. Use the canvas to decide what's worth building. Then put one needle in front of a team that owns the outcome, not a queue of tickets.

Before you add another feature, answer one question straight. What's your single needle-mover this quarter, and who owns moving it? If the answer is "all of it" and "me, between fundraising calls," that's not a roadmap. That's a vitamin and a wish.

Forward this to the founder who just shipped something nobody's paying for.