Context Switching Is Stealing Your Engineering Hours

Your most expensive engineers are spending their days in shallow loops between Slack pings, and you are paying senior rates for junior output.

The hook

Last Tuesday I opened our Linear board to understand why a feature that should have taken three days was dragging into its second week. The engineer, one of our strongest, had exactly one PR open. Not blocked. Not waiting on review. Just incomplete. I checked her calendar: four "quick syncs," two incident triage threads for a service she does not own, a Slack huddle about a migration she had context on from six months ago, and a standup that ran 25 minutes because someone used it as a design review. She told me, "I had maybe 90 minutes of real coding time yesterday." She was not exaggerating.

हरीच्या जागरणा · The Vigil for the Lord

हरीच्या जागरणा । जाता कां रे नये मना ॥
कोठे पाहशील तुटी । आयुष्य वेचे फुकासाठी ॥
तुका म्हणे बरा । लाभ काय तो विचारा ॥
Life is being spent for free. The person who does not audit where their hours go will look for the loss and find it everywhere.

What I keep seeing

I have watched this pattern at three companies now, and it gets worse as engineers get more senior. Seniority makes you the default person to pull into every room. The more you know, the more rooms want you. The more rooms you enter, the less you ship. Nobody plans this. It just happens, one "quick question" at a time.

Cal Newport's research at Georgetown put a number on it: the median knowledge worker checks communication tools every six minutes. Microsoft's Viva Insights data showed the average Teams user context-switches 15 times per hour. Everyone agrees interruptions are bad. Almost no one structures their org to prevent them. I have been guilty of this myself. I have pulled engineers into rooms "just for five minutes" and cost them an afternoon.

The mechanics

Context switching is not an annoyance. It is a measurable cognitive cost. When you pull an engineer out of a problem, the state they were holding in working memory (variable names, data flow, the shape of the function three calls up the stack) does not pause. It evicts. Gloria Mark's research at UC Irvine measured the recovery time: roughly 23 minutes to return to the same depth of focus after a single interruption. Not 23 minutes to start working. 23 minutes to reach the same cognitive depth.

Four interruptions in an afternoon do not cost four times 23 minutes. They cost the entire afternoon. The engineer never reaches depth. They operate in a shallow loop: read message, lose thread, re-read code, start rebuilding mental model, get pinged again. A junior engineer with three unbroken hours will outproduce a staff engineer who gets tapped every 40 minutes. Software is not typing. Software is holding a complex model in your head long enough to see the flaw in it.

Slack makes this worse by design. The green dot is a promise of availability. Channels create ambient obligation. "Quick question" is never quick for the person being asked. Only for the person asking.

Where Tuka comes in

Tukaram's frame is not about discipline. It is an accounting argument. "आयुष्य वेचे फुकासाठी": life is being spent for free. Fuka means without payment, without return. He is not calling you lazy. He is saying you are being robbed and you have not checked the ledger.

The vigil (जागरण) is the structure that protects attention. It is not about trying harder to focus. It is about choosing what you stay awake for, and then refusing to leave. The question he poses, "लाभ काय तो विचारा" (consider what the gain actually is), is a CTO's question in devotional clothing: what is the return on this meeting, this ping, this "quick sync"? If you cannot answer it in one sentence, the meeting should not exist.

What I would actually do

Institute a no-meeting, no-Slack window of four hours every morning, org-wide, enforced at the infrastructure layer: Google Calendar Focus Time, Slack scheduled Do Not Disturb, Linear notifications paused. Standups move to async Loom updates, 90 seconds per person, no exceptions. Any synchronous meeting requires a written agenda posted 24 hours in advance or it gets cancelled automatically. Measure managers not on how many meetings they run but on how many unbroken two-hour blocks their reports get per week. This is not a productivity hack. It is a structural decision about what your most expensive people spend their time doing. Right now, most orgs are paying staff-engineer rates for someone to read Slack in 23-minute fragments. That is not an engineering problem. It is an accounting problem, and Tuka spotted it four hundred years ago.

Chetan Dhandal