# We Are About to Get Very Good at Building Bad Ideas



For years, engineering scarcity saved us from ourselves.

Every product team I have run, at Rappi, at Revolut, at Yuno, carried a backlog several times larger than its capacity. We treated that as a problem to manage: prioritization frameworks, quarterly planning fights, the endless negotiation over what makes the cut. But that scarcity was doing quiet, unpaid work for us. A bunch of mediocre ideas never got built because there simply wasn't time. The cost of building forced us to argue and to choose.

We are losing that protection.

I see it in my own work every week. Things that used to take a squad a quarter now take a small team days (in some cases they take me an afternoon, alone, with agents). Building is not free yet, but it is falling toward free fast enough that the difference stops mattering.

Most people read this as an unambiguous win. I think it is more dangerous than that. We are about to become extremely good at building bad ideas quickly.

## The simple part [#the-simple-part]

The central question of product development is changing, and once you see it the rest follows.

For twenty years the question was "can we build this?" Feasibility was the gate. Entire disciplines grew around answering it: estimation, capacity planning, technical discovery, the roadmap itself. When building is free, "can we" is almost always yes. And a question that is almost always yes stops filtering anything.

The questions that replace it are "should we build this?" and "how are we going to make it successful?"

That second one matters more than it looks. Building the product used to be the whole job, and it gave product people a special place: we sat at the door of a scarce resource. Now everyone can build. So the job moves to everything around the build. Knowing what will actually move the needle, which is taste, and no model has taste about your specific customers. And then launching, driving adoption, and proving the thing made a difference. A shipped product is the starting line. Making it successful is the discipline.

None of this is controversial. Say it in any product meeting and everyone nods.

## The complicated part [#the-complicated-part]

Nobody nods at what comes next.

When building was expensive, products died of natural causes. They lost the roadmap fight, starved of engineering time, and faded. Scarcity did the killing for us, and we never built the muscle ourselves.

When building is free, nothing dies on its own. Every experiment, every side bet, every "quick win" lives forever unless someone deliberately ends it. And ending it requires the hardest sentence in product management: "I was wrong, this didn't work."

How do you fight your own ego to say it out loud? You championed the idea. You celebrated the launch. So instead of saying it, we keep the thing alive and tell ourselves it is strategic.

The result is zombie products. Technically alive, moving nothing. They take space in your codebase, where someone has to maintain and patch them. Worse, they take space in your team's head: every roadmap discussion and every architecture decision now has to route around them. At Yuno we orchestrate 1,000+ payment methods across 190+ countries; I think about portfolio weight every single day, and I can tell you the heaviest company loses.

The mistake is treating the kill decision as an act of individual courage. Courage is not a system. If sunsetting a product requires someone to volunteer for ego damage, it will not happen often enough.

So here is my proposal: kill criteria written at inception. Before building anything, agree what success looks like. A number and a date, not aspirations. If the product misses it, sunset is the default, and keeping it alive is the decision that needs a defense.

Notice what this flips. You are not admitting failure when you kill it; you are honoring an agreement your past self made when they were objective. The ego fight happens at inception, when it is cheap, instead of two years later, when it is impossible.

Two more pieces would complete it:

1. Outcome reviews, not launch reviews. We celebrate launches and almost never reconvene ninety days later to ask if the thing moved the needle. That second meeting is the one that matters, and it should be as mandatory as the first.

2. A real sunset process. Killing a product well is a skill: migrating users, communicating honestly, deprecating cleanly, and harvesting what you learned. Teams that do it well make the next kill easier. Teams that do it badly teach everyone to never try again.

## What I'm not sure about [#what-im-not-sure-about]

I don't know yet if kill criteria survive contact with reality. A number agreed eighteen months ago can be genuinely wrong: the market moved, the strategy changed, the metric was badly chosen. There will be pressure to renegotiate every criterion the moment it is about to trigger, and if renegotiation is easy, the whole mechanism is theater.

I also don't know where the line is between a zombie and a bet that needs more time. Some of the best products looked dead at month twelve. My current answer is that the difference is whether anyone can say, concretely, what would change the trajectory. A bet has a thesis. A zombie has a defender.

I am going to try this. I will write about how it goes, including the criteria that turn out to be badly set.

Until then, one question, and I actually want answers: what is the oldest zombie product you are still maintaining, and what would it take for someone to finally turn it off?
