I left tigerbeetle this week. They didn't do anything wrong, far from it. I'm just not well suited to long-term employment.
I should know that by now, but I convinced myself that this time would be different. At the beginning I was talking about a 3-month trial contract, knowing the risk, and then I just got overexcited and dove in head-first.
Before taking this job I was approaching critical mass - between hytradboi, github sponsors, a grant from emergent ventures and a sprinkle of consulting, I had assembled the equivalent of a postdoc position in a field that doesn't really exist yet. I gave that up to take this job, and that made it very hard to admit that I had made a mistake. On top of that I convinced myself that it was worth sticking to something for the experience of persevering itself, so I was determined to stick it out for at least two years.
What finally nudged me over the edge was reading Quit. It read like a chapter-by-chapter checklist: Have you been thinking about quitting regularly for a long time? Are you reluctant to let of of the large sunk costs? Are you continuing solely because of commitment to some arbitrary goal? Are you worried people will think you are a quitter? Have you started gnashing your teeth in your sleep?
Ok, that last one was not in the book, but physical pain seems like a good place to draw the line.
So I'll go back to what I was doing before. Being independent was hard and I was often filled with doubt, but it is the only path I've tried that has felt right.
I don't know exactly what that will look like. I'm proud of my smaller projects from 2020/21 (eg internal consistency, hytradboi, why dida), but my work on imp felt overwhelmingly hard and was pulled in too many different directions. It's clear that I have to do something differently, but I don't know yet what that will look like.
My time at tigerbeetle was mostly spent testing, debugging and optimizing the storage engine, alongside @kprotty. I recently backported a simple benchmark sending 10m transfers between 2 accounts (more interesting benchmarks were harder to backport). Back in 2022 Dec 6 we were averaging 2,734 transfers/second with a peak (closed loop) latency of 14,749ms. By 2023 Apr 14 we had it up to 120,397 transfers/second with a peak latency of 3,817ms. That feels pretty satisfying.
Some substantial share of the credit is due to tracy. I'm glad that tigerbeetle is now sponsoring tracy, and I'll continue sponsoring them myself.
I've very tentatively made a mastodon account - @email@example.com. Not sure yet that I'll keep it, but worth a go.