Selin Yilmaz is a postdoctoral researcher in behavioral economics at a European research university. She publishes, teaches, supervises two master’s students, serves on two departmental committees, and is co-investigator on a grant with external reporting requirements.
She also, for most of the past two years, had a deep work practice that was failing her.
She had the time blocked. She had a quiet office. She had a clear research agenda. What she did not have was a reliable way to actually get started once the time arrived.
The Baseline: What Was Going Wrong
Selin tracked her sessions for four weeks before making any changes. The data was clarifying.
Her 90-minute writing blocks—protected on her calendar three days per week—were producing roughly 55 minutes of actual productive output on average. The first 25 to 30 minutes were being consumed by re-orientation: re-reading the last section she had written, scanning her notes, locating the relevant references, trying to reconstruct the argument she had been developing.
She knew this was happening. She had always assumed it was simply the nature of academic writing—that you spent the opening minutes getting back into the work.
What she did not know was that this was not universal. Some researchers she spoke with described sitting down and beginning almost immediately. The difference, she learned, was preparation—specifically, the presence or absence of a working-memory brief they had given themselves at the end of the previous session.
Selin had no end-of-session handoff process. Every new session started cold.
Version 1: The Handoff Note Alone
Her first attempt was simple: she would write a three-sentence handoff note at the end of every session describing where she left off and what the next move was.
This helped. Her re-orientation time dropped from 25 minutes to roughly 15. But it did not solve the problem fully.
The handoff note told her where she was, but it did not reconstruct the argument she had been building. She would read the note, nod, and then still spend ten minutes re-reading her last paragraph to remember why the argument had moved in that direction.
She also found that the handoff note did nothing for the other major entry friction: she would sit down to write and find herself mentally managing the email she had not answered, the student question she owed a response to, the seminar abstract that was due by end of week. None of these were urgent. All of them were distracting.
Version 2: Adding AI to Gate 1
Three months into her experiment, a colleague mentioned using AI for context loading before writing sessions. Selin was skeptical but tried it.
Her Gate 1 prompt evolved over a few weeks. The version she now uses:
Here are my notes and the last two paragraphs I wrote
on my current paper section:
[paste]
I need a working-memory brief for a 90-minute session.
Give me:
1. Two sentences on the core argument I am developing.
2. The single strongest objection I have not yet addressed.
3. The next specific thing to write—not "continue the argument"
but the actual point the next paragraph needs to make.
4. One reference I should probably check before I go further.The output from this prompt typically takes her two minutes to read and orient to. Her entry time dropped from 15 minutes to under five.
The difference was not just speed. It was cognitive posture. She was arriving at the document as a writer who knew where she was, not as a researcher who had to reconstruct where she was. The mental energy that went into reconstruction was now available for writing.
Version 3: Adding Gate 2 for Interruption Triage
The remaining friction was the ambient anxiety problem. She described it as “carrying the office with her into the writing session”—awareness of pending items that she was not yet done with.
She added Gate 2: a five-minute interruption triage before each session.
Her prompt:
I have a 90-minute writing block starting now.
Here are the things I know are pending and might pull my attention:
[list—typically 4–7 items]
For each one: is there a 30-second action I can take right now
to create closure, a short message I can send,
or can it genuinely wait until after the session?She worked through the resulting list before closing the AI. For student messages, she sent a quick “response by end of day” acknowledgment. For the seminar abstract, she blocked fifteen minutes the following day. For the committee email she had been putting off, she drafted a two-sentence reply.
She described the effect as “setting down the office outside the door.” The metaphor is apt. She was not suppressing awareness of pending items; she was actually resolving or deferring them in a way her mind could accept.
Her re-orientation time with both gates in place: under three minutes on most days.
Version 4: Gate 3 and the Exit Point
The final element she added was an exit-point definition at the start of each session.
Her previous sessions had ended in one of two ways: she ran out of time (the most common), or she lost momentum and drifted. Both left her uncertain whether the session had been a success.
Her Gate 3 prompt:
My session goal is to work on [specific section/argument].
I have 90 minutes.
Define a specific, observable deliverable that would constitute
a successful session. Be concrete—name the actual paragraph
or argument that should exist by the end.The exit point changed her experience of the session in a way she did not fully anticipate. She described it as “the difference between running a loop and running to a destination.” With a destination, she made decisions during the session differently—she was more willing to skip a tangent that would not advance her toward the exit point, and more able to recognize when she had arrived.
Her sessions became shorter on average—closer to seventy-five minutes than ninety—because she was finishing rather than filling time. She also produced output she rated as higher quality on a simple self-assessment she tracked alongside session length.
The Stable State: Eighteen Weeks In
Selin has been using the full three-gate runway for eighteen weeks. Her current data:
- Average re-orientation time: under three minutes (down from 25–30)
- Sessions hitting their exit point: approximately 70% (she initially estimated she had been hitting a clear goal in maybe 20% of previous sessions)
- Subjective quality rating (1–5): average 3.8 (up from 2.6 under the old approach)
- Total AI time per session: 6–8 minutes (before session), 0 minutes during
She also uses Beyond Time to log session metadata alongside her daily planning, which lets her review patterns across weeks without maintaining a separate spreadsheet.
What She Would Tell Another Researcher
Three things, she says:
The entry problem is real and solvable. She spent two years assuming that slow starts were intrinsic to complex work. They are not. They are a context-loading problem with a specific solution.
The interruption triage is not optional. She almost skipped Gate 2 when she first read about it, thinking it was for people with more chaotic environments than hers. Her environment felt managed. In practice, the five minutes of triage made a measurable difference. “You think you have parked the pending items. You have not. You are still monitoring them.”
Close the AI. She tried running sessions with the AI window minimized for the first few weeks. She found herself opening it. Minimized is accessible. Closed requires a deliberate act. The discipline of closing it before starting is not incidental to the process—it is the process.
The Lessons That Transfer
Selin’s situation has specific features—academic research, multi-role fragmentation, writing-heavy work—but the mechanism generalizes.
The entry problem is not academic. Any knowledge worker doing complex, cognitively demanding work in an environment with competing demands faces the same friction: arriving at the session without the context loaded and with the mental weight of unresolved items.
The runway solves that friction. The specific inputs change. The gates remain the same.
Start Here
If you recognize the first thirty minutes of a session being consumed by re-orientation, the single most effective change is adding a Gate 1 prompt before your next session.
Paste your notes into AI and ask for a two-sentence brief and one specific next action. Do that before your next session starts. That single change will show you whether the rest of the runway is worth building.
Related:
- The Complete Guide to Deep Work with AI Assistance
- The Complete Guide to Deep Work Scheduling with AI
- The Complete Guide to Time Blocking with AI
Tags: deep work, case study, researcher productivity, AI workflow, attention residue
Frequently Asked Questions
-
What is attention residue and why does it matter for researchers?
Attention residue is the cognitive persistence of previous tasks into new ones, documented by researcher Sophie Leroy. For researchers who context-switch between teaching, email, administration, and writing, residue from earlier work degrades the quality of focus available for deep thinking. Managing it is one of the highest-leverage interventions available to academic knowledge workers.
-
How did AI change the case study subject's session entry time?
Before implementing the runway, Selin averaged 25–30 minutes of re-orientation before productive writing began. After implementing the three-gate runway, her entry time dropped to under three minutes on most days. The change came from AI handling context loading and interruption triage rather than her doing it manually through re-reading.
-
Is this case study generalizable to other researchers?
The specific numbers will vary, but the underlying mechanism—reducing context initialization cost and attention residue through structured pre-session preparation—applies broadly to anyone doing complex intellectual work in fragmented environments. Researchers face an unusually high version of this problem due to multi-role work patterns.