MVP now means maximum viable product, not minimum

MVP used to give you permission to ship something rough. Minimum viable product. The whole point was that it could be ugly and half-finished, because the idea was to test whether anyone wanted it before you spent real effort making it nice. For years that was good advice. Ship the embarrassing version, learn, iterate. I believed it too.

I don't think it holds anymore, and I want to be honest about why, because clinging to the old version of this advice will quietly sink your first version in 2026.

The bar moved while we weren't looking

The quality bar for software jumped, and it jumped because building got easy for everyone at once. When anyone can generate a clean, working, decent-looking product in a weekend, "decent-looking and working" stops being a milestone and becomes the floor. Your rough MVP is no longer being judged against nothing. It's being judged against the polished weekend project some stranger shipped the same day, and against the dozen others that landed this week, and against whatever launches tomorrow, because something better always launches tomorrow now.

That's the part that really changed. It's not just that the bar is high. It's that it keeps rising every single day, because the cost of clearing it keeps falling. Your "full product," the thing you're proud of, could genuinely be someone else's casual MVP if they happen to have good taste and a good stack. That's a brutal thing to sit with, but it's true, and pretending otherwise doesn't help you.

So MVP flipped meaning. It's not minimum viable product anymore. It's closer to maximum viable product — the most polished, considered version you can manage for the scope you've chosen. Not maximum in features. Maximum in care. The scope can and should still be tiny. Do one thing. But the one thing you do has to clear a real taste-and-quality bar from the very first version, because people decide in seconds now and they have infinite alternatives.

Why people won't tolerate slop

The reason is simple and a little unforgiving: people are exhausted by AI slop. The generated, default, seen-it-a-thousand-times look has flooded everything, and people have developed an instant allergy to it. The moment something reads as low-effort or generic, the trust is gone before they've even tried it, because the signal it sends is "nobody cared here, so why should I."

And that's the thing — quality isn't really about looking pretty. It's a signal. A first version that's been genuinely cared for tells someone that you sweat the details, that you'll keep sweating them, that this is a real thing made by a real person with standards. A sloppy one tells them the opposite, and in a world of infinite alternatives the opposite is fatal. The polish is the message.

This is exactly why taste became a moat. When the baseline product is free and uniform, the scarce thing is the judgment to make something that doesn't feel like the baseline. Taste is mostly restraint — the discipline to do one thing fully and well instead of five things roughly, and to refuse the generic default even when it's right there for free. It's the one part of building that didn't get commoditized, because the AI hands everyone the same competent middle, and climbing out of that middle on purpose is still entirely human work.

What I actually do about it

With Blankdot I've had to make peace with shipping less, but shipping it cared-for. The temptation is always to add another feature so it looks substantial. The better move, almost always, is to cut scope and raise the finish on what's left. A smaller surface done with real care beats a bigger one that feels generated, every time, because the small careful one earns trust and the big generated one leaks it.

Concretely, that means I let the agent generate the boring scaffolding fast, and then I spend my actual attention on the handful of decisions that make it feel like a person built it. The one interaction that's genuinely smooth. The copy that sounds like me and not like a template. The deliberate choice — a real typeface, a real color, one detail that's clearly a decision and not a default — because that's what separates my thing from the flood. The generated baseline is free. The taste on top is the whole job.

So if you're about to ship your first version, don't ship it rough and call it minimum. Shrink the scope until you can make it genuinely good, then make it genuinely good. Clear the bar on a small thing rather than miss it on a big one. The era where ugly-but-it-works earned you a pass is over. Maximum viable is the new minimum, and taste is the part of it nobody can hand you for free.