I've worked at stripe for two years. I learned a lot about scaling organizations and software. I've also made a lot of friends! Over my time there, I became less of a software engineer and more of a systems engineer. I learned to think about systems holistically - the fleshy users & engineers are part of the system.
If I had to take one thing from stripe, it's that it's really easy to make large impact at scale. Just fix stuff. Just.. look!
Why leave - shedding responsibility
I decided to leave. Having to narrow the stuff I build to "economically viable for stripe" didn't feel like the highest return for myself at the moment. It's true that I got to build stuff around LLMs there - I was able to get Stripe mentioned on the GPT-4 blog post! But even if I was automating things with LLMs, it wouldn't have been the goofy capabilities research I really wanted to do. At stripe, there were more important things to take care of than.. automating a weed whacking robot.
Ah, the latter stage of a software engineer's career. Responsibility.
I couldn't justify spending all of my time there experimenting. And even though stripe is fast, I need to move faster still. Scaled orgs get capped by Amdahl’s law. I can’t afford that right now.
Fortunately, I was able to buy myself some time! I found out that working as a contractor part time is, in fact, viable. And better still, there are places where the "most economically viable activity" is closer to what I want to do! I'll have two days a week to build stuff on the side. I've already got a gig for 3 days a week that I’m really excited about!
(If you would like some system built, please reach out to me, and we'll find out a way to make it happen!)
What I'm doing next
I'm going to do indie research & development. I'll design and ship AI systems that are maximally useful for humans. I'll share my findings through code, blog posts, and demos.
The objective of my research is to understand how much utility we can extract out of small consumer grade AI models. There are software patterns & primitives that will emerge, and I want to help uncover them. I'll figure out what those patterns are by shipping a lot of different useful stuff for myself and others. Something common will emerge!
User happiness will be my eval!
Thanks for reading!
yacine.ca for raw notes
Are my notes/blog posts useful to you? Throw me a one time ⬆money upvote!⬆
moving at the speed of light, we shall seek the stars
Makes sense, similar reason to why I left Mailchimp. Great folks, got to start going towards the NLP & LLM route but the Mark Rober "squirrel backyard obstacle course" type stuff was never going to fly.