Starting a Company at the Beginning of the Exponential
Last updated: April 21, 2026
The following is a historical perspective of how it felt to start a company in November 2025. I'm writing this down so that we can all look back a year from now and relive the beginning of AGI.
November - Boundless Optimism
It's a warm fall day. My brother Mayank and I have been exploring company ideas. One of our good friends introduces us to these hedge fund legends, a storied PM and a killer Partner.
On paper, we had everything. Technical ability, domain expertise, and warm intros to customers. The kind of setup where the only question was how fast.
Cursor was fantastic. Within a couple weeks, we built heavy data pipelines to filter the world's data (Kalshi, Polymarket, Twitter, YouTube, News, etc.).
What took me a couple months to build a year prior now took a couple weeks.
December - Hard Problems are Easy
Opus 4.5 changed everything.
Product iteration felt like cheating. We started every week with a grand redesign of the product, with high hopes that this iteration would solve it. By the end of each week, we'd have 3 customer conversations and 7 full app redesigns.
What took me a week to implement in November now took a day.
January - Wait a Second...
Codex 5.2 drops.
Coding is no longer a manual task.
Our team starts to have at least one existential crisis a day. Did the interface matter anymore? What could we build that still mattered?
We completely dropped the UI. It was time to go all in on agents.
Maybe we were on to something, but it still felt off.
What took me 1 day of effort in December now took a few hours.
February - It's Over
Opus 4.6 and Codex 5.3 were released on February 5th, 2026.
Coding became 100% automated.
We meet an AGI-pilled customer who had just started using Cowork:
“Wait why won't this just be done by Claude?”
“Well, the filtering logic is actually pretty non-trivial, there's a lot of domain-specific context, you'd need... right?”
We finished the sentence. The customer didn't respond. The call ended.
We go on a walk outside to talk about life post-AGI. Never had the future been so uncertain, but it was clear that our software company had little time left. As we stared across the Hudson, we realized: software is dead.
What took me a few hours to build in January was now a single prompt.
Reflections & The Future
It's hard to describe how an exponential feels until you're living through one. I consider myself AGI pilled, and even I was caught off guard by the rate of progress. Builders felt it first. The rest of the world is about to. I'm excited and terrified to watch it happen.
Enterprises will keep buying software, and there's real value in solving their problems. But now that building it is a commodity, I want a harder challenge.
It's time to search for long horizon problems that can't be one-shot by AGI.