The Comfort of the Wrong Problem
I spent last weekend rereading "The Last Question." Not for nostalgia. I needed to understand something about the products I'm building, and honestly, the story made me deeply uncomfortable about it.
The premise is simple: humans keep asking a superintelligent computer the same question across centuries—can it reverse entropy? The computer always says it cannot, not yet. But here's what matters for us as builders: the humans never stop asking. They keep coming back. They keep trusting the system with their deepest problem.
This is the opposite of what we do.
- We solve problems nobody has yet, then convince them they need solving
- iterate based on usage data from people who downloaded something on a whim
- We call this "product-market fit" when really we're just getting lucky with retention metrics
- Ship, measure, adjust—that's the mantra
I'm not sure this is the right move, but I see the pattern everywhere. Last year, I built an AI tool that automated contract review for SMEs. Beautiful product. Shipped in 8 weeks using Claude's API. We got 340 signups in month one and thought we'd cracked it. By month four, active users dropped to 28. The problem we solved was real—but not real enough. Not the kind of real that makes someone wake up thinking about it.
The Question Nobody Stopped Asking
Asimov's genius was understanding that the humans in his story had the same problem for 200 years. Same question. Same hunger for an answer. The computer's value wasn't about innovation or disruption—it was about reliability toward a single obsession.
We don't build for obsessions anymore. We build for friction points. We optimize for DAU and engagement loops and we A/B test the language on our CTA buttons until we squeeze out a 3% improvement that probably doesn't matter, I'm honestly not sure why I still do this, and then we call it product excellence.
In 2023, I watched a founder at a Bogotá tech meetup pitch a tool that used LLMs to help HR departments manage employee onboarding. It was technically solid. The demo was smooth. And as I listened, I realized: nobody lies awake thinking about onboarding. It's a problem, sure. It's a cost center. But it's not a compulsion. It's not the reason someone becomes an entrepreneur.
Asimov's humans asked their question because the answer mattered to existence itself. That's leverage.
The Uncomfortable Part
Here's where I get stuck: I don't know if this realization changes what I should be building. It might not. Maybe the world doesn't need more obsession-driven products. Maybe it needs better tooling for the thousand small frictions that make life at work 5% worse than it should be. Maybe incremental value adds up.
- But also: every AI product that gets abandoned after 90 days is a failure of imagination
- We're treating these tools like features when they could be infrastructures
- Stripe didn't become Stripe by making payments 3% faster
- It became Stripe by solving something developers couldn't stop thinking about
I'm working on something new now. Can't say much, but I'm deliberately not starting with user research or TAM analysis. I'm starting with a single person's unresolved frustration—a CTO I met in Medellín who's been fighting the same problem for four years. Not complaining about it. Fighting it. Spending his weekends on it. That's the obsession marker.
The Question That Hangs
Asimov's last scene always hits different when you've shipped products nobody needed. The computer finally has the answer. It can reverse entropy. But by then, nobody's asking anymore because the question matters less than the asking itself did.
I wonder how many of our tools will look like that in ten years—technically correct solutions to problems that stopped mattering the moment we solved them.
