First Principles
Ten minutes to write. Ten people to review.
Anoop Tripathi, Chief Technology Officer at Sauce Labs, on why a quarter became a week, why generated documents cost more than they save, and the capacity signal that remote work took away.
12min Watch
Anoop Tripathi
Chief Technology Officer, Sauce Labs
What this conversation is about
Anoop Tripathi runs product engineering and technology at Sauce Labs. Partway through this conversation he describes a piece of work his team estimated at a quarter and delivered in under a week. He is careful about why. The work itself was small — a bit of UI, a bit of infrastructure, a bit of everything else. What made it a quarter-long project was that it crossed several teams and several time zones, and each small piece had to wait for someone else’s sprint. What collapsed was not the work. It was the waiting.
Then he turns the same lens the other way. Engineers who used to write half a page now produce ten and twenty pages in minutes. One-slide updates have become ten-slide decks. The documents contain errors, so ten people spend real time redlining what one person generated in ten. Producing got cheap. Checking did not. Nothing was removed from the system — it was relocated onto people who did not choose when it would arrive.
Those two stories look like a win and a cost. They are the same variable moving in opposite directions: who controls the timing of the work in front of you. Waiting on another team’s sprint is demand you can see and cannot act on. An unannounced ten-page review is demand that arrives without warning. High demand with low control is the condition the human stress response is built for, and it is not built to hold it for long. Which is why the detail that matters most in this episode is not the ninety days saved. It is what Anoop noticed afterward — that his team got visibly happier, and that the reason was they could finish things.
Three things to take from this
1.
Measure what the work waits on, not how long it takes.
A quarter became a week without anyone writing code faster. If your estimates are dominated by handoffs across teams, time zones, and sprint boundaries, then most of what you are estimating is queue time — and queue time is the part that is now compressible. Before you attribute a delivery win to a tool, find out how much of the old number was people waiting on other people.
2.
If generating got ten times cheaper and reviewing didn’t, you moved the cost onto someone else.
A ten-page document produced in ten minutes is not a productivity gain if ten colleagues each spend an hour correcting it. The person generating feels faster. The people downstream absorb the difference, usually without a say in when it lands. Before you send it, ask what it costs to read, not what it cost to write.
3.
Distributed work hid capacity, not effort.
It is easy to see whether someone appears busy and nearly impossible to see what they could actually produce. Those are different measurements, and the second one mostly stopped being taken. If the only signal you have about someone’s load is what they tell you, you are not measuring capacity — you are measuring willingness to report it. That is a system design problem, not a character problem.
“The joy of being able to finish your work actually makes people happier.”
— Anoop Tripathi
Host’s note — Dharma Ramasamy
Anoop describes happiness arriving with completion, and he is describing something physical. Work you cannot finish because it is parked in another team’s queue does not sit quietly in the background. An unresolved obligation you have no power to close is a low-grade, continuous demand on the stress system, and the body treats it as unfinished business rather than as waiting. Remove the blockage and what people feel is not primarily relief about time. It is the return of control. That is why his team’s mood changed faster than any workload metric would predict.
The same mechanism explains the other half of his answer. Ten pages arriving unannounced is demand with no notice and no ability to refuse. Predictability and control are the two things that determine whether demand registers as manageable load or as threat, and generated volume erodes both at once. This is the part that rarely shows up in an adoption dashboard: the tool can genuinely take work off a team and simultaneously raise the physiological cost of being on it, because the load did not vanish, it changed hands and lost its schedule.
His closing answer is the one I keep returning to. He can see a team’s average throughput. He cannot see an individual’s actual capacity, and he says plainly that he is still working on it. Capacity is not a productivity number. It is what a person has left. Almost nothing in the modern stack is built to show it, which is how organizations end up discovering depletion at the same moment they discover the slip.
Be on the series.
Ten minutes, two cameras, nothing to prepare, and you see the edit before it publishes.
Get in touch →The measure he says is missing.
Anoop can see what his teams deliver. He says he cannot see what any one person actually has left to give, and that he is still working on that. Most leaders find out the same way — after the fact. If that is a gap you recognize, this is the work I do.
How the diagnostic works →