starting something new

February was my last month at Cash App. I got some questions about why I decided to leave—and why now—that I thought would be worth writing about.

Nine years ago, I co-founded a company called Sway Finance. We went through YCombinator during what I describe as the last AI wave—this was 2016, and everyone was promising that AI was going to reinvent the future. We promised to use AI to make bookkeeping simple and affordable for small businesses. AI at the time didn't mean LLMs—LLMs didn't exist yet. It mostly meant supervised machine learning, and if you were really ambitious, unsupervised machine learning. There weren't APIs for anything and certainly nothing like OpenAI or Anthropic; we trained our own models and connected Node servers to data pipelines in Jupyter notebooks. It wasn't even the wild west: things those days felt more like being researchers let loose from the lab with scribbled notes and white papers of what should work.

Sway ultimately didn't work out, and we all went on in the world without Sway and generally without AI reinventing our lives.

In 2023, OpenAI released ChatGPT-4, and Square CEO Alyssa Henry bought the entire company enterprise accounts with the mandate that we incorporate AI into our work somehow. This was the first time I had really thought about AI in several years aside from watching my friend demo Dall-E to me the year prior (we made it draw cats in the ocean partaking in a Japanese tea ceremony).

The thing about large companies is that it takes forever to make a decision, but you have the resources and immediate distribution once you do. So my team built one of the first ChatGPT-based AI projects at Block: Glean, before Glean. Somewhat to our surprise, it worked—and it worked well. We—nor anyone else in industry—knew how to benchmark its accuracy in a scalable way and sometimes it hallucinated, but it worked and it didn't require a PhD to build. Funnily, the concerns we had around accuracy and the security approvals we had to overcome to build our internal Glean bot are non-issues already, both at Block and also across industry. At the bleeding edge, time and progress only accelerate.

Which brings me to my AI red pill: nine years is a long time, but the pace that AI is developing and becoming useful to mainstream society far outpaces that. I know I'm not novel or early to this realization, but the older I grow, the more convinced I am in my naive 22 year old's belief that AI will reinvent the future. It's moving at such a fast clip that waiting even one more year to start a company has meaningful consequences. Things take longer than we expect but they also move faster than we could have ever imagined. That much is obvious with each passing year.

So I'm starting something new. I'm excited to share more about what it is soon, and I'll be building this next chapter in public. Subscribe below if you'd like to follow along (or message me for our secret Discord!).

Here's to the future... and to simultaneously moving fast and slow.

march 3, 2025
← back