Two months in Tokyo, on a student visa
An internship at a Japanese bank taught me more about writing code other people can read than any course did.
In December of my third year I got on a plane to Tokyo to intern at MUFG, having never been further from Bangalore than Dubai. I had a suitcase, a phrasebook I did not open, and a mental image of the job that turned out to be wrong in every particular.
The codebase was older than I was
I had expected to be handed something new. Instead I was handed a Java service with commit history going back further than my age, comments in two languages, and a set of conventions nobody could point me to a document for. They existed. They lived in the heads of people who had been there fifteen years.
My first week I wrote a function I was quietly proud of. It was compact, it used a stream chain that collapsed four steps into one expression, and it worked. In review, a senior engineer asked me to expand it into a loop.
I pushed back a little. He was patient about it. His argument was roughly: this file will be read by someone at 2am during an incident, in a language that is not their first, under time pressure, and the loop is the version that person can read. Cleverness is a cost you pay later, and you are not the one who pays it.
That reframing has outlasted almost everything else I learned that year. I still write the clever version first. I just do not ship it any more.
Things I got wrong before I got them right
- Assuming silence in a meeting meant agreement. It often meant the disagreement would arrive later, privately, and more thoroughly considered than anything I would have said out loud.
- Optimising my own throughput. I measured a good day by how much I shipped. The team measured it by how little disruption my changes caused to everyone else's week. The second metric is harder and more correct.
- Treating the commute as dead time. Forty minutes each way on the Chuo line, twice a day, six days a week. I read papers on my phone standing up and got through more of them in two months than in the two semesters either side.
The part that was not about work
I was there through February, which meant I got winter. Cold enough that the walk from the station was a decision rather than a stroll. I ate at the same place near the office often enough that they stopped asking what I wanted, which is a form of belonging you cannot request.
I understood almost nothing anyone said to me for the first three weeks, and it was the least lonely I had been in years. Being unable to participate in the small talk turned out to be restful.
I did the tourist things badly and on the wrong days. What I remember is smaller: the particular quiet of a train carriage where nobody is on speakerphone, the vending machine on the ninth floor that dispensed hot corn soup, the way the whole office emptied at exactly the same minute and refilled at exactly the same minute.
What I took home
I came back with better Java, marginally worse posture, and one durable idea: the hardest constraint on most engineering is not the machine, it is the next person to read what you wrote. Japan made that constraint visible because I was, for two months, permanently the person who could not read the room.
I have kept the habit of writing for that person since. It is the closest thing I have to a style.