I Stopped Having One Workday

This is not advice. It is a field report from an experiment I am still running.

Let me get the numbers on the table. Core sleep runs 22:00 to 02:30. Three naps follow: 07:00 to 07:20, 12:00 to 12:20, and 17:00 to 17:20. Total planned sleep is five and a half hours a day, and yes, it is deliberate.

I want to kill the obvious misreading early, because it is the one this kind of article always attracts. This is not a grindset experiment. I am not trying to win some founder sleep deprivation competition, and I have no interest in the story where sleeping less proves commitment. If anything, this schedule exists because I take recovery more seriously than I used to.

The interesting question was never how little I can sleep. It is whether I can control the shape of my cognitive day instead of accepting the default one.

The Default Day Is One Long Degrading Block

The conventional day has one waking block. You wake up, you work, fatigue accumulates, and eventually you collapse into bed. Whatever state you are in at three in the afternoon is the residue of everything since morning. By evening you are running on fumes and calling it dedication.

I know that day well. I have written about what short sleep does to me: I miss obvious bugs, I take small feedback personally, I read a neutral message as a hostile one. That analysis was right. The prescription was timid. I was still treating sleep, work, and training as three independent schedules that occasionally collided, and the whole goal was to reduce the collisions.

The better framing fell out of my day job. I spend my working hours designing systems, and a day is a system. Sleep, cognitive work, training, and interruptions all change one another. Optimizing any one of them in isolation gives stupid results, the same way optimizing one service in isolation gives stupid architecture.

So I stopped asking how to fit recovery around work. I started asking what the day would look like if I designed the whole thing at once.

The Actual Schedule

Plain facts, because the rest of this essay is commentary on them:

• Core: 22:00 → 02:30
• Naps: 07:00 → 07:20 · 12:00 → 12:20 · 17:00 → 17:20
• Total planned sleep: 5h 30m

What falls out of this is the part I actually care about. Instead of one waking block of seventeen hours, I get four shorter ones: 02:30 to 07:00, 07:20 to 12:00, 12:20 to 17:00, and 17:20 to 22:00. Each block is short enough that I never reach the degraded tail of a long day. Each nap closes one block and opens the next.

Why 2:30 in the Morning

This is the paragraph where I am supposed to tell you that successful people wake up early. I will not. The 5am club is a cargo cult: copy the visible trait, hope the invisible ones follow.

The appeal of 02:30 is not virtue. It is territory.

From half past two until about seven, almost nobody wants anything from me. No messages, no meetings, no notifications pretending to be urgent. The ambient pressure that shreds a normal morning does not exist at that hour. I can load a large technical problem into my head, the whole dependency graph of it, and hold it there for hours without a single context switch.

That first block is where the hardest thinking of the day goes. Not because I am disciplined at dawn. Because the environment is finally quiet enough to think in.

Naps Are Boundaries

This is the intellectual center of the thing, at least for me.

The old workday had no internal structure. It started when I woke up and ended whenever I finally gave up. Every hour asked the same question: keep going or stop? And every hour, that question got answered by a version of me slightly more tired than the last. Tired me keeps working. That is how you end up arguing on Slack at midnight, not because the work needs it, but because stopping is a judgment call and judgment is the first thing fatigue takes.

A long-running process with no checkpoints is hard to reason about. State accumulates, nobody writes it down, and the only way to know the system's condition is to be inside it. My old day was that process. The naps are checkpoints. Every four or five hours the day halts, state gets flushed, and the next block starts clean.

The twenty-minute constraint matters as much as the nap itself. A short nap fits between things. It does not ask the day to be empty first, and a rest you can actually take beats an ideal one you cannot.

Training Sits in the Same Budget

One insight from my old essay survives intact: exercise is not a reward you unlock once work is done, because work is never done. The new framing sharpens it. Work, training, and sleep are not three hobbies competing for hours. They are inputs into the same recovery budget.

My training right now is deliberate rather than maximal: Heavy Duty style strength work, running, and recovery work when the rest of the system asks for it. Heavy Duty already assumes recovery is the binding constraint rather than effort, which makes it a natural fit for a schedule built out of recovery windows.

The question stopped being whether I have time to train. It is what today's budget looks like. A missed nap, a hard session, and a long evening block are not three facts. They are one fact wearing three hats.

Where This Goes Wrong

Without this section, everything above is biohacking propaganda.

A meeting that lands on a nap usually wins, and the block after it runs degraded. A call that crosses the 22:00 boundary is worse. Pushing the core sleep later does not steal an hour from the night; it breaks the next morning's first block, which happens to be the best thinking time I own. "Almost done" remains the most dangerous phrase in my vocabulary. Staying up for it borrows against tomorrow at terrible rates.

Then there are the softer failures. Travel wrecks the timing. Social obligations do not care about my nap windows. An interruption turns a twenty-minute nap into no nap at all. And the most dishonest one: feeling fine. Feeling fine is not evidence. Fatigue accumulates quietly and shows up as judgment, not yawns, right up until I catch myself taking a shortcut I would normally talk myself out of.

The biggest failure mode is subtler than all of these. Four clean blocks produce real surplus capacity, and the temptation is to promise all of it to somebody. If every recovered hour gets sold back to work, the system defeats itself. The surplus is the point. Some of it has to stay unspent.

Discipline Is Not Rigidity

The schedule needs hard boundaries to function at all. That is not the same as worshipping the clock.

When something genuinely important breaks the architecture, I break it consciously. Name the trade, keep it short, recover afterward. What I refuse to do is let exceptions quietly become the schedule. One late night is a decision. Three late nights is a new system, and usually a worse one.

What I Am Actually Testing

The experiment is not whether I can survive on five and a half hours of scheduled sleep. Survival is a low bar. People clear it all the time while slowly getting worse at their jobs.

The experiment is whether the shape of the day is mine to design. What I am watching: decision quality inside each block, whether the old afternoon collapse still exists, irritability, training performance, nap reliability, what breaks first when the schedule gets disrupted, and whether any of this stays compatible with an actual life. These are observations, not measurements. I am not going to invent a dashboard for them.

I also do not know whether I will still be running this exact schedule a year from now. That is part of the experiment. If the data turns, I will turn with it.

What I Actually Manage

The most useful thing this experiment has done is force a reframe I should have had years ago. I do not manage time. I manage state.

A calendar can tell me I have an hour. It says nothing about which version of me shows up for it: the one who can hold the whole problem in his head, or the one who rereads the same paragraph four times and calls it work. Sleep, training, and work are just the dials that decide which one arrives.

The schedule is the current implementation of that idea. It will change. The idea will not.