Case Study 2: David's Deep Work Environment

An 18-Month Journey to the Perfect 5am–7am Learning Block

David is a 35-year-old software architect learning machine learning. He has a demanding job, a partner, a dog, and exactly zero ideal study hours in a normal weekday — unless he creates them himself.

He created them at 5am.

This wasn't a natural fit. David is not a morning person. "I used to make jokes about people who got up at 5am willingly," he says. "I don't make those jokes anymore." The transformation took about six weeks of miserable early mornings before it became something he looked forward to.

But the environment is the reason it works. Without the specific physical and digital setup he's built, the early mornings would have dissolved into aimless phone scrolling and getting back into bed. The environment is what makes 5am mean machine learning time instead of just time that happens to be 5am.

The Problem He Was Solving

Before the 5am system, David tried learning ML in the evening after work. He had good intentions: he'd come home at 6:30pm, eat dinner, and plan to study from 8–10pm. Then:

• His work phone would buzz with a message. He'd check it. Then think about the problem.
• He'd open his laptop for ML study but have 15 browser tabs open from work.
• His partner would want to watch something together; he'd feel guilty about studying; he'd half-watch and half-study and do neither well.
• By 9pm he'd be too tired to engage with new concepts. He'd "review" (reread) something he already knew.
• By 9:30pm he'd give up and go to bed.

Three hours of "study time" producing, at best, 30 minutes of actual cognitive engagement with new material.

The evening simply didn't work. His brain was depleted from work, his environment was full of competing cues (TV, work messages, social interaction), and the physical space was not associated with focused learning.

The morning, he discovered, was different. Before work, there were no competing demands. The rest of the world was asleep. His brain was fresh. His environment could be completely controlled.

The Physical Setup

David's 5am learning block happens in a specific room — the spare bedroom that's been converted into a home office. He did not have this room from the beginning; for the first four months, he studied at the kitchen table. The dedicated room came later, but it made a substantial difference.

The Room

The spare bedroom-turned-office has: - A standing desk (he alternates between sitting and standing in 25-minute blocks) - A proper monitor (he does not learn on a laptop screen alone — the larger screen reduces cognitive friction when working through code and reading papers simultaneously) - No television, no comfortable couch, nothing associated with relaxation or entertainment - A whiteboard on one wall (used constantly for drawing architectures, explaining concepts to himself) - Blackout curtains that he keeps closed (at 5am in winter, the darkness outside is not helping; inside, controlled lighting tells his brain it's "work time" regardless of what's happening outside)

The room is used only for two things: deep work and video calls. Not casual internet browsing, not watching content, not playing games. The dedicated-use principle applies here: the room signals work mode the moment he enters it.

The Chair

This sounds trivial and isn't. David uses a specific chair — a proper, supportive desk chair — only in this room. When he sits in it at 5am, after 18 months, the association is automatic: this chair = deep work. "It's like Pavlov with me as the dog," he says. "I sit in that chair and I feel like working."

No Phone in the Room

This is absolute. David's phone charges in the bedroom. It does not come into the office during the 5am block. Full stop.

At 5am, there are no urgent calls he needs to take. His partner isn't going to call him from the bedroom. The work emergencies that happen at 5am are the kind that would actually wake you up — and if it's not waking him up, it can wait until 7am.

"I cannot overstate how much having the phone in the other room changes the experience. I don't think about it. The drawer-in-another-room is way more effective than the drawer-in-the-same-room. Out of sight isn't enough. Out of the room is what works."

The Monitor: Work Computer Only

His office monitor is connected to his personal laptop — his learning computer, not his work computer. His work laptop is in his bag, not plugged in, not visible, during the 5am block.

This physical separation prevents the work-thought contamination that plagued his evening sessions. Work problems live on the work computer. They don't exist in this room until 8am.

The Digital Setup

Computer Configuration for 5am

David has a specific macOS user profile for his learning sessions — different desktop background, different dock, different default browser, different notifications settings. When he switches to this profile: - All work email and Slack notifications are automatically disabled - The social media websites he uses for work networking (Twitter/X, LinkedIn) are blocked by Cold Turkey - His browser opens to a simple start page with his three scheduled tasks for the session

This profile-switching sounds elaborate. It takes 30 seconds. He set it up once and hasn't touched it since. The result: opening his learning profile is itself a cue — everything about the computer configuration tells him "this is learning time, not work time."

What He Actually Uses

• Jupyter notebooks: for running code, experimenting, following along with textbooks
• Anki (desktop version): the 15-minute review session that opens every morning
• A single PDF viewer: textbooks and papers, opened in full-screen, no tabs
• A text file for the session's notes (plain text, no formatting distractions)

Nothing else. No browser tabs he doesn't need. No email client. No messaging apps.

The Night Before Prep

This is perhaps the most important digital habit in David's system: the night before, he spends 5 minutes setting up the next morning's session. He opens the right textbook chapter, notes down the three specific things he wants to understand or accomplish, and stages his Anki cards. When he sits down at 5am, the environment is already configured. There are no decisions to make about what to do — the work is staged and waiting.

"Decisions are expensive in the morning. If I have to figure out what to work on at 5am, I'll either pick the easiest thing or spend ten minutes figuring it out. If it's already decided, I just start."

The Startup Ritual (7 Minutes)

David's startup ritual is non-negotiable. He's done it every single weekday for 18 months, with two exceptions (illness). It consists of:

Step 1: Coffee (2 minutes) Make a cup of coffee in the kitchen. This is the only morning activity that happens outside the office. The walk to the kitchen and back is the transition between "home" mode and "learning" mode. He does not check his phone during this step.

Step 2: Sit in the chair, open Anki (1 minute) He doesn't ease in. He sits down and opens Anki. This is the beginning of the ritual; everything before Anki is just preparation. The act of opening Anki signals to his brain: the session has started.

Step 3: Anki review (15 minutes) Every morning, without exception. At this point in his journey (18 months in), he has approximately 2,000 ML-related cards in his deck. His daily review queue is typically 50-80 cards, which takes 15-20 minutes. These span everything: math foundations, Python syntax, algorithm concepts, architecture patterns, paper summaries, key researchers and their contributions.

Step 4: Write three focus questions (2 minutes) On the text file open on his screen: what three things do I want to understand by 7am? These are specific — not "understand neural networks better" but "understand why vanishing gradients occur in deep networks and what the intuition behind ReLU's advantage over sigmoid is." Specific questions produce specific learning. Vague intentions produce vague sessions.

Step 5: Begin He opens the pre-staged textbook or notebook and works.

The Music Question

David uses music. Specifically, a single instrumental playlist he's been using for months — he calls it "the playlist that means work." It's always the same playlist, always from the beginning, always on low volume. He plays it only during his 5am learning sessions.

He chose music deliberately for this session because he found that complete silence, at 5am with most of his brain still warming up, allowed too much mental wandering. The familiar instrumental playlist occupies just enough of his background attention to prevent the mind from drifting, without competing with the cognitive work.

He acknowledges this might not be optimal for everyone. "I actually tested it. I did two weeks with music and two weeks without and tracked my output — lines of code written, questions answered in my session log, concepts I could recall the next day. The music weeks were slightly better for me, though not dramatically." He treats it as a personal data point, not a universal recommendation.

The Results: 18 Months In

David spends two hours per weekday on deep ML learning, plus occasional weekend sessions. In 18 months, he has: - Completed three substantial ML courses (fast.ai, Andrew Ng's Coursera specialization, Stanford CS229 audited) - Read 12 textbooks and monographs on ML topics - Built and deployed four ML projects, including one used by his company - Transitioned from "curious software architect" to "ML-capable architect who handles the ML architecture on production systems"

He attributes roughly equal credit to the techniques he uses (retrieval practice, spaced repetition, deliberate practice on problems, interleaving) and to the environment he created (the room, the ritual, the phone policy, the digital configuration).

"The techniques give you the direction; the environment gives you the traction. Without the room and the ritual, I'd still be trying to study at 9pm with my work phone buzzing and the TV on in the background. Without the techniques, I'd be spending two focused hours rereading things I already know. You need both."

What Others Can Steal from David's Setup

You almost certainly can't replicate David's exact setup. You may not have a spare room. You might not be able to wake up at 5am. But the principles are universally applicable:

1. Dedicate a specific location to deep learning — and use it only for that. Even a specific chair. Even a specific corner of a room.
2. Put the phone physically out of reach, not just face-down. "Another room" is dramatically more effective than "same room but in a drawer."
3. Pre-configure your digital environment the night before. Stage what you'll work on. Make the first decision of the session "sit down," not "figure out what to do."
4. Design a startup ritual and repeat it without variation. The ritual is the cue. It only works if it's consistent.
5. Find the time in your day when your brain is freshest and protect it for your most important learning. For David it's 5am. For you it might be a different time. But if your best cognitive hours are going to work or social media, your learning is running on the remainder.