Product and engineering are worth almost nothing

Product and engineering, on their own, are worth close to zero, and getting closer every year. I say this after more than a decade writing code and shipping products, not despite it.

Building stopped being the bottleneck

Ten years ago, a working product was a moat. Most people couldn't build it, or it took a team and a year, so if you had something that worked, you had a lead that was hard to close. That gap is gone. AI coding assistants, mature frameworks, one-click infra, all of it has collapsed the distance between an idea and a working version of it. Things I used to scope in weeks now take an afternoon, a hidden game in the corner of this site being one of them, and not because I got faster. The tools did.

When everyone can build the thing, the thing stops being the advantage. That cuts both ways, and in security it cuts hard: the cost of running a competent attack collapsed on exactly the same curve.

Illustrative chart: over the last ten years the cost of building a working product falls steeply, while the cost of getting in front of people rises. — Illustrative, not measured.

What's scarce now

If building isn't scarce, attention is. There are more good products than there are eyes to find them, and that ratio only worsens as building gets cheaper. The constraint moved from "can we make this" to "will anyone ever see this." Distribution, an existing audience, a channel that compounds, a wedge into someone else's users, a founder who already has a following, a sales motion that gets you into rooms competitors can't, is the scarce resource now. AI hasn't touched it, because it was never a technical problem. It's a trust problem, and trust gets built slowly, with people, over years. There's no model for that yet, and maybe there never will be.

I've watched worse products win because they had it, and better products die in silence because they didn't. That's what happens when the supply of "good enough" is infinite and the supply of attention isn't.

The Bending Spoons tell

Look at Bending Spoons, the Italian company that recently listed on the Nasdaq. Their acquisition playbook is public and it's blunt: buy a company for its users, not its product, and often not its people. They've bought apps with millions of installs, kept the install base, and rebuilt the product underneath it with their own team. The team that built the thing they just bought is frequently the first thing to go.

Diagram of a Bending Spoons acquisition: the users are kept, since they are what was paid for; the product is rebuilt underneath by Bending Spoons' own team; the original team is often the first thing to go. — What a Bending Spoons acquisition keeps, and what it doesn't.

They pay real money for distribution alone, and treat the product and the people who made it as the disposable part.

That's an accurate read of where the value sits. In a world where anyone can rebuild the product, the only thing you can't rebuild by Tuesday is an audience that already opened the app today. Attention is the business now, and increasingly it looks like the only economic asset left worth buying, selling, or defending.

The other trap: shipping because you can

There's a second-order effect too. When building is nearly free, the temptation is to keep building forever, because every new feature costs almost nothing to ship. It's a trap, and a worse one than doing too little.

The old rule, less is more, is truer now than it's ever been. Every feature you add after the core one has thinner marginal utility than the last, and past some point that delta goes negative. A feature nobody needed adds a decision, a menu item, a place for the user to hesitate, a new way for the thing that used to be obvious to become slightly less obvious. Confusion is a cost, and it compounds exactly like distribution does, just in the wrong direction.

Cheap building was supposed to free you to focus. For most teams it's done the opposite: it removed the one constraint (cost) that used to force discipline.

The exceptions are rare

There are exceptions: technical moats in frontier model training, novel hardware, or infrastructure that takes years and enormous capital to replicate. If you're doing something a thousand engineers with AI tools still can't reproduce in six months, product and engineering excellence still buys a defensible advantage.

But that's a sliver of the market. Most of us aren't training foundation models. Most of us are building an app, a tool, a workflow, something a competent team with decent AI assistance could rebuild in a few months once they knew it worked. For that entire category, and it's most of software, distribution is the whole game.

What I'm doing about it

This changes where I spend my time, and just as much what I say no to. Less time polishing a feature nobody asked for, more time figuring out who this is for and how they'll hear about it, before I write a line of code. Fewer features, because each one I don't ship protects the ones that already earned their place.

Build fast, because you can and so can everyone else, and ship less than you're capable of, on purpose. Then spend what you saved on getting in front of people and staying there, the one part that still takes real work.