I wrote a nice wrapper for some glue code a colleague wrote. They liked my API, and gave me merge perms on the glue code’s repo. Nice … street cred.
Later my colleague asked me to look into a race condition. I looked into it briefly, but wasn’t convinced it was a race. I pointed out that the error handling was broken.
Still later, a third party contributor saw my comments and offered to fix the error handling. I took this as the impetus to look into the race again. My colleague did a good job documenting the issue, and had implemented a hotfix (a retry) so it felt like a good oppurtunity for a followup.
I became convinced it was indeed a race condition after reading my colleagues issue docs and taking a cursory look at the API’s called by my colleagues glue code. I was also fairly certain of what kind of race condition is was.
Armed with the knowledge of what to look for, I went a few levels deeper down the stack. I found where the database calls were, and confirmed the issue. I documented my findings, and sent it to the owning team and my colleague.
Thoughts
Workload
I was supposed to take the day off, but instead spent 4-5 hours looking into the issue. This is bad. I tired myself quite a bit, for not much gain (negative gain when considering I needed the break). Not entirely sure why I couldn’t stop.
Street Cred
Several people congratulated me for finding the issue. Their praise felt good.
And I partly did for their praise. And partly because I was sucked in. And partly because I wanted to feel efficacious.
But also partly for their anti-praise? To show to them, and myself, that it was possible. Not special.
And for that … their praise hurt me.
Am I replaceable?
An llm could forseeably have done this. I’ll need to check.
Most people don’t have the patience, careful eye, and context to catch this stuff. Does that matter when the machines can read now?
Conclusion
I felt productive and I got to help people. I hope that’s enough.