I built the first working version of Onto by myself, in about ten weeks, from a tier-3 engineering college in India while finishing my degree. People assume that's a story about hustle. It's mostly a story about constraints, and why I'd take them again.
The constraint
I didn't have a co-founder, a network of warm intros, or a runway measured in years. What I had was a clear thesis — that AI agents are blind to the live web and someone needs to build the layer that fixes it — and the time to ship something instead of talk about it.
Constraints sound like disadvantages until you notice what they remove. No team meant no coordination overhead. No runway meant no permission to build the wrong thing for six months. I had to find out if the idea was real, fast and cheaply.
Why solo, why fast
Solo, because the slowest part of early building is alignment, and there was no one to align with. I could decide and ship in the same hour. Fast, because the only way to know if external context infrastructure was worth building was to put a real Read API in front of real agents and watch what happened.
Ten weeks in, Layer 1 was live and listed in Anthropic's official MCP Registry, with paying customers. That's not a vanity milestone. It's the cheapest possible proof that the thesis had a market.
What worked
Shipping narrow. I didn't try to build the whole platform. I built the one primitive agents needed most — clean, structured content from any URL — and made it genuinely good. Distribution through the places agents already live, like the MCP Registry, did more than any amount of marketing would have.
What didn't
Plenty. I underestimated how much of the work is reliability, not features. I spent time on things no customer asked for. Doing it solo means every mistake is yours alone, and some of them cost weeks I didn't have.
What's next
Layer 1 (Read) is live. Layer 2 (Watch) — continuous monitoring of the agent web — is shipping to first design partners. Layer 3 (Govern) is the enterprise bet. The constraints are loosening, but the lesson holds: build the next narrow thing, ship it to real users, and let the work compound.
Solo founder, a few months in. Still just getting started.