The cost of interrupted work (2023)

Some days an interruption will throw me off my train of thought, and I spend the remaining six hours collecting discarded bottles and railway ties for hopeful use somewhere, somehow, sometime.

Other days an interruption costs me pretty much nothing.

I’m still trying to figure out how to tell which of those days I’m going to have and whether to just not log into Slack for the day.

I've found one thing that minimizes interruption cost: pair programming. At one startup we pair programmed all day, every day. Resuming from an interruption was almost seamless. Can't explain it, only experienced it.

If only I wouldn’t prefer stabbing myself in the leg with a rusty knife over pair programming.

When pair programming was a fad in the early 2000s, I tried it with a coworker for a security-critical piece of code that needed two pairs of eyes on it.

It felt horrendously unproductive to have two people at one keyboard but we compared commit rates and the surprising result was that we produced the same rate of changes as working separately.

I find it is slightly slower (maybe 20%) than two individuals alone but the quality is quite a bit higher and the effect of this higher quality compounds over time (i.e. less tech debt -> fewer bugs, faster development on future code).

Im quite credulous of Kent Beck's claim that when categorizing the last ~15 bugs on a project with pairs and singles he found that all 15 were in code merged by an individual rather than a pair.

If it were an application you could just install I think everybody would use it. It demands psychological safety though, which most teams dont have, and is becoming less common these days.