Jianxin Shen
Notes

The email I didn't send

I held back the angriest line of the dispute. I was let go anyway.

Jianxin Shen

In September 2025 I drafted a sentence to a research colleague and didn’t send it: “We are not your support staff, and we don’t take orders.” Holding it back took real effort. In April 2026 I was let go. The dispute was probably one of the reasons. How much it weighed, I can only guess. I wasn’t part of the decision.

I led the engineering side of a trading research system. A shared data package sat underneath the research team’s strategies, and it was maintained by an intern I had hired. For five months, the dispute between the two teams kept changing its subject.

In July it was about who fixes things. Initialization errors, missing data and mismatched timestamps all landed on the package, and the question was who repairs them and how to avoid repairing the same thing twice.

In September it was about whose definition of correct counts. A lookback calculation failed on one historical date because the underlying data for that day was incomplete. We argued about whether that was the function’s fault or the data’s.

In November it was about who may change production code. I rewrote a data fetch that ran in roughly a quarter of the time and produced identical output on the existing tests. The same day it had been moved out of the production file and into a test file with my name in its filename.

Each time, I thought more evidence would end it. It never did, because by then evidence wasn’t what we were disagreeing about.

What I was actually thinking

Here is my thinking at the time, unedited:

If you can’t do it better than me, shut the fuck up. If your proposal works locally and is garbage globally, fuck off.

A year later I can see that this was two sentences fused together. The second is a technical judgment, and I mostly still stand by it: a fix that serves one strategy and degrades the shared system is a bad fix. The first is a reaction to being challenged. It treated a question about my code as a question about my standing. Because I couldn’t separate the two, every technical disagreement turned into a contest over my position, and every additional benchmark raised the stakes instead of lowering them.

What we were missing was not persuasion. It was a distinction between four kinds of authority that we had never pulled apart:

AuthorityThe question it answers
RequestWho may ask for a fix?
PriorityWho decides what gets fixed first?
MergeWho decides a change is good enough to land?
ProductionWho accepts the risk of running it?

The research team clearly held the first. Neither side had a settled claim to the other three, so each one was decided by whoever pushed harder that week.

The part that was mine

Before that intern, the package had another owner: an earlier intern, also my hire. Pressure from its direct users bent the code out of shape, and I didn’t step in. The second intern inherited both the bent code and, eventually, the dispute.

I was leading people for the first time. The person I was arguing with on the research side was a fresh graduate. Apart from the person we both reported to, I was the only one on the team with prior work experience, and that person wasn’t technical. Nobody had assigned the four kinds of authority. The team had grown without a design.

That vacuum was probably my opening. I was the one person who could have written those four lines down and made them stick, and I didn’t. Twice, I let ownership slide to the least protected seat on the team.

What I can’t resolve

I keep saying I’d handle it differently now. When I try to say how, I find I haven’t actually worked it out.

And yet I’m the one who paid for it. I spent a month on garden leave and had a new job before it ended, and it is a better one. That came from my own work, not from the episode. The episode itself was a loss for everyone in it: for me, for the company, and for the team.

I know where the opening was. I still don’t know what I’d have had to become to take it.