Reflections on Toronto

I leave Toronto on August 22, after sixteen years here and three years at university. Each section below is one thing I got wrong, or one thing that turned out differently than I expected.

I think determination is really cool

In second year I went to every office hour, for every course, whether or not I was stuck. I was tutoring across the city two and three times a week the whole time, an hour each way on the subway, and paid my own tuition with it. I wasn’t technical, so most things took me about twice as long as they took the people around me, and I did them at that speed instead of pretending otherwise. I went to everything I could get to, a dozen or so events in a year and most of the hackathons.

I was in the office twelve hours a day for most of the winter I spent at Mozilla.

The version of this people usually want is that I got strategic, or learned to ask for help. I did ask for help constantly and it mattered. It isn’t the cause. The cause is that at some point I started caring, and once you actually care the hours stop being a cost you’re paying.

I don’t know whether caring is something you can decide to do. Mine showed up in a year where I failed most of what I attempted, and I’m not sure what would have happened without that year.

The position that made everything else possible paid nothing

My first real engineering work was a part-time volunteer spot on a university team maintaining an autograder. I got it because I went to an event, heard someone talk about their internship, and emailed a professor the next week. I was assigned tickets and worked through them. It was the first time I’d touched a real repository, a pull request, or a code review.

Every technical thing I’ve done since ran through those four months. Mozilla, where I wrote the privacy paper and shipped C++ into Gecko. Shopify, where I spent this summer on search, building the IDF and BM25 retrieval pipeline behind help search. Neuralink infrastructure, which starts in three weeks.

I had no way to evaluate whether that position was good, because I didn’t know enough to evaluate anything. What made it work was that it was real: a real codebase, real reviews, and people who would tell me when the code was wrong.

I got the opposite version wrong a year later, when I took an internship for the office and the company name rather than the work. I knew within days it was a dead end. The signal I should have been reading was never the brand. It was whether anyone there would look closely at what I built.