First Principles
When people stop asking how and start asking should
Darshan Yadav, Field CTO at Seclore, on ownership giving way to permission-seeking, the impact of work that never got built, and the verification tax AI moved onto senior engineers.
Darshan Yadav
Field CTO, Seclore
What this conversation is about
Asked what he notices first when a team starts to struggle, Darshan Yadav does not describe a metric. He describes a change in the questions people ask. They move from problem solving and ownership to permissions. The question stops being how do we solve this and becomes what should we do. He says this shows up well before anything is late.
That is a sharper signal than it sounds, because ownership and permission-seeking are not two working styles. They are what the same person looks like before and after the cost of being wrong starts outweighing the cost of not moving. Nobody decides to make that trade. It arrives as a change in what feels prudent. And it happens in the questions long before it happens in the delivery date, which is why the delivery date is the worst thing to watch.
The rest of the conversation holds the same shape. He values the people who kill bad ideas before they reach production or even requirements — impact measured as work that never got built, which almost no reporting system is designed to see. And he is precise about AI: it removed the typing and got people to answers faster, and it introduced a verification load that did not exist before, paid by senior engineers and reviewers. The same people whose value is deciding what not to do are now absorbing the largest volume of things that must be checked.
Three things to take from this
1.
Listen to the grammar of the questions, not the status.
How do we solve this and what should we do are different cognitive states, not different phrasings. When a team’s questions shift toward permission, something has changed in how the people asking are weighing risk. It is free to notice, it arrives weeks before any dashboard moves, and it is still cheap to act on at that point.
2.
Count the work that never got built.
The highest-leverage contribution on most teams is a bad idea stopped at requirements. Nothing in a delivery metric can represent it, which means the people doing it are systematically invisible in exactly the reviews that decide who matters. If you lead a team, you have to name that contribution out loud, because no system will name it for you.
3.
Find out who is paying the verification tax.
Generating got faster. Checking did not. That difference lands on reviewers and senior engineers as a load nobody scheduled and nobody counted. Before you call an AI rollout a win, ask the two or three people at the review end of it what their week actually looks like now.
“People move from problem solving and ownership to permissions.”
— Darshan Yadav
Host’s note — Dharma Ramasamy
Darshan named the cleanest behavioural marker anyone in this series has given me, and I do not think he knows that is what he did. He was describing a delivery signal. He was describing a nervous system.
Under sustained threat load, the weighting between two costs quietly inverts. The cost of being wrong starts to outrank the cost of not acting. Nobody experiences this as fear and nobody decides it — it arrives disguised as good judgement, as checking first, as making sure. What it produces on the outside is precisely what he described: the question in someone’s mouth changes from how do I solve this to am I allowed. Agency contracts before output does. Which is why permission-seeking is not a maturity problem or a culture problem to be trained out. It is a load reading.
The part I find most interesting is what he asked for at the end. He wants a dependency map — visibility into what a decision hangs on. But his own earliest signal already is one. When someone starts asking permission, they are telling you where the dependency sits, in behaviour rather than in structure. The map he wants exists. It is drawn every day, in the questions, by people who have stopped believing they can move without cover.
Be on the series.
Ten minutes, two cameras, nothing to prepare, and you see the edit before it publishes.
Get in touch →The map he’s asking for.
Darshan wants visibility into what decisions actually depend on. That dependency is already being drawn — in who has stopped moving without permission, and where the checking load quietly concentrated. Making it visible before it reaches the delivery date is the work I do.
How the diagnostic works →