It was 2am. I was in bed, scrolling, half asleep.
Somewhere between posts I caught a passing line about how Urdu came to be. I’m Pakistani. Urdu is in the air I grew up breathing. I had never actually asked the question.
I opened ChatGPT out of curiosity.
What I found was less like a birth and more like prolonged friction. People in crowded places who needed to be understood right now. Soldiers giving commands. Traders quoting prices. Locals trying to answer. Migrants trying to keep up. Nobody waiting for grammar to be approved. Nobody pausing to be polite. They borrowed what worked, dropped what slowed them down, and kept moving.
I’m not here to litigate the linguistics. The story might be partly romance. The pattern is what I couldn’t put down.
Urdu wasn’t designed. It was negotiated. Grammar from whoever was numerous. Vocabulary from whoever had power. Rhythm from whoever showed up every day. The poets and the courts came later, after the language had already proven it worked. Correctness arrived second.
I sat with that for a while. Not because of the linguistics.
Because I have watched it happen to every product I have ever built.
I keep trying to design things. Every time, I open a doc, I write a vision, I draw the system, I name the parts. By month three the vision is a relic. The product that exists is the one that survived twenty small arguments with reality.
Arlo started as a tool for one person. Me. I built it because I was tired of explaining the same context to my AI tools every morning. It was supposed to be a personal memory layer. Quiet, single user, mine.
Then I sat across from my co-founder at the first real enterprise conversation Arlo ever had. I went in expecting to pitch. I spent most of it listening. By the end of the call, Arlo wasn’t the same product. It had a second life, one where the memory belonged to an organization instead of to one tired founder.
I didn’t design that. The conversation did.
Continuity did the same thing in reverse. I shelved it a year ago because the ecosystem wasn’t ready. Then the ecosystem moved, my own workflow buckled under too many agents who didn’t know each other existed, and the shelved thing got rebuilt because reality forced the issue. Not because I had finally finished designing it.
Pulse, Presence, Reconcile. Same pattern. The plan was one thing. What survived was what the friction allowed to survive.
We meet products the way we meet languages. After they have names, rules, defenders. After the case study has been written and the deck has been polished. The friction that actually built the thing is invisible by then. The compromises read like decisions. The accidents read like strategy.
It is comforting to believe that founders design products. It is more honest to say products are negotiated into existence between a founder, a few collaborators, the customers who show up, and the customers who quietly leave.
The vision deck is the poetry that gets written after the language already worked.
I have spent years trying to be a better designer of products. I am starting to suspect the higher leverage move is to be a better negotiator with reality. To stop defending the original plan. To notice faster when the friction is telling me what the product actually wants to be.
Urdu didn’t begin with consensus. Neither did anything I have ever built.
The refinement comes later. The friction does the work.