In my previous article I wrote about the time that passes while AI works. From there, people split in two. Some stay and watch, and calculate how much efficiency they’re losing. Others parallelize, and find that their brain hands them the bill in the evening.
I tried a third way. I wrote myself a prototype that does one thing: while the agent works, it tells me where to go. It doesn’t choose better than I do. It chooses instead of me, based on the priorities I gave it beforehand.
Three weeks of logs, seven days of real use.
The loop is always the same. I read what the agent produced on the first job, give my feedback, send it off again, and while it starts I’m already on the second: read, reply, send it off, I’m on the third. Then back to the first. 314 times in seven days.
Within one loop the machine works for five to ten minutes. My part, understanding what it did and replying, takes just over a minute.
And this is where people measure the wrong thing. There’s an index going around that calculates how much time you lose waiting for AI, to tell you whether it’s worth using. That race is over: five minutes of machine plus one of yours, against the hour it used to take you. It pays off even if you sit there staring at the progress bar.
The real question comes after, and nobody asks it. Of those six minutes, only one is your work. If you wait, you pay for all six: a factor of six, and it’s not an estimate, it’s a sum. Over a whole day it drops to two and a half, because not everything I do (for now) sits behind an agent. It’s still enough: those seven days, worked in sequence, would have been seventeen.
Ten days. Thrown away watching a bar creep forward. Nobody stands staring at the washing machine for forty minutes, yet in front of an agent we all do: at home you always know what needs doing, and there you don’t.
I expected the result. Not its shape. On those days I switched to a different job 87% of the time: 420 jumps, with zero seconds of friction between one action and the next, because the choice was already made when I got there. That’s where the biggest gain is, and it’s counted in decisions saved.
Zero seconds on the clock. Not zero in my head. The load of switching stays, and you won’t find it in any log: you find it in the evening. With one difference that makes the effort worth it: it’s tiredness from things that moved forward.
Waiting time isn’t spent. It’s filled. Decide now what to fill it with, because once you’re waiting, you don’t decide anymore.