The Command Vault: An Execution System for Getting Things Done

Most people who cannot focus can focus fine. Put them in front of something genuinely interesting with no way out and they will disappear into it for three hours. What fails is not the capacity. It is everything around it: the task was never defined precisely enough to start, the block was never protected, and the work was interrupted at the exact point where it became difficult.

The Command Vault holds deep work, execution, focus systems, and creative output. It is the vault where things actually get finished, and it is the one most likely to be full while everything else sits empty, because this is where the deadlines live. That makes it a strange vault to write about: the problem is rarely getting yourself to open it, it is getting real output out of the hours you already spend inside it.

The definition problem

Almost every failed focus session was doomed before the timer started, and the cause is nearly always the same: the task was not a task. It was a topic.

Work on the report is a topic. Write the section comparing the two approaches, four paragraphs, no editing is a task. The difference is not pedantry. When you sit down with a topic, your first ten minutes are spent deciding what to do, and deciding is the most expensive cognitive state to be in at the start of a session, because it is the state where the browser tab opens. When you sit down with a task, you start.

A real task passes three tests. You know what the first physical action is. You know what finished looks like specifically enough that you would not argue with yourself about it. And it fits inside one session. If any of those fails, the item is a project, and the useful move is to spend one short session converting it: write down the finished state, then list the first three sessions. That conversion session is often the single highest-leverage thing you will do all week, and it is the one people skip because it does not feel like real work.

Protecting the block

The second failure is environmental. You can define a task perfectly and still lose the session to a fifteen-second interruption, because the cost of an interruption is not the interruption. It is the reconstruction afterward, and reconstruction takes far longer than the thing that caused it.

This is why partial protection does not work. A block with notifications silenced but the phone face-up on the desk is not protected, because you are still spending attention on not looking at it. The protections that hold are physical and boring: the device in another room, the door closed, the browser windows unrelated to the task actually closed rather than minimized, and a single visible note stating what you are doing.

The other half is temporal. A block that anyone can book over is not a block. If you work with others, the block has to exist somewhere they can see it, and it has to be reliably in the same place, because a moving block is one that everyone learns to ignore including you.

Where you put the block matters more than how long it is. Almost everyone has a window in the day when hard thinking is cheaper, and almost everyone spends that window on email because email is what is in front of them when it opens. Spend one session identifying your window, then defend it specifically. Two protected hours in the right window will out-produce five hours in the wrong one.

Session length and what to do inside it

Twenty-five minutes is the default for a reason: it is short enough that starting is not a negotiation and long enough to reach the part of a task where something happens. Use it for anything you are resisting, anything unfamiliar, and anything you have already postponed twice.

Fifty minutes is for work with a long runway, which mostly means writing, designing, building, analyzing, and anything where the first fifteen minutes are spent loading context that you do not want to load twice. If a task requires holding several things in your head at once, do not put it in a twenty-five-minute container. You will spend the whole session assembling and none of it working.

Fifteen minutes is not a lesser session. It is the correct container for admin, for triage, for the specific task you have been avoiding, and for days when you are depleted and the honest choice is between fifteen minutes and nothing. Fifteen minutes and nothing are very different numbers over a month.

Inside the session, three rules do most of the work. Keep a scratch page for intrusive thoughts, so the idea about something else gets written down in four seconds instead of becoming a detour. Do not research mid-draft; write a placeholder and keep moving, because looking things up is the most respectable-looking way to leave a task. And when the timer ends, stop, then write one line about exactly where you are and what comes next. That line is the difference between the next session starting immediately and the next session starting with ten minutes of rereading.

Finishing things that span multiple sessions

Single-session tasks are easy to manage. The real difficulty in this vault is work that takes eight sessions across three weeks, because that is where things quietly die.

The mechanism of death is always re-entry cost. You do a session, leave it in an ambiguous state, come back four days later, and spend twenty minutes reconstructing your own reasoning. Do that twice and the project starts feeling heavy. Heavy projects get postponed, postponement increases re-entry cost, and the loop closes.

Two habits break it. First, always stop at a place you can describe. Not a clean stopping point, which is a myth, but a describable one. Second, end every session by writing the next session's task in full, task-shaped form, before you close anything. You are handing a specific instruction to a version of yourself who will have forgotten the context. Write it for that person, not for the one who currently has everything loaded.

There is also a scheduling point. Multi-session work needs frequency more than duration. Four sessions in a week beats one long session per week, even when the total hours are identical, because the context stays warm and re-entry costs almost nothing.

A Command Vault session menu

Concrete tasks sized to a timer. The fifteen-minute entries include the ones worth running when the real work feels impossible, because a small finished thing restores momentum better than a large abandoned one.

• 15 minutes: Convert one project into tasks. Write the finished state in a sentence, then list the first three sessions. Do not start any of them.

• 15 minutes: Clear the input you have been avoiding. One inbox, one message thread, one form. Finish it inside the timer and close it.

• 15 minutes: Write tomorrow's first task in full task-shaped form, with the first physical action stated. This is the cheapest way to make tomorrow morning work.

• 15 minutes: Kill something. Go through your active commitments and formally drop or postpone one, in writing, including telling anyone who needs to know.

• 25 minutes: Draft one section of a longer piece with editing explicitly forbidden. Placeholders for anything you would otherwise stop to look up.

• 25 minutes: The single task you have postponed three times. Do not attempt to finish it. Attempt to reach the point where you understand why you were avoiding it, which is usually an undefined step in the middle.

• 25 minutes: Audit last week's sessions. What did you actually finish, what stalled, and what was the stall about. Fifteen of those twenty-five minutes should be spent on the stalls.

• 50 minutes: The main creative or analytical block on your most important project, in your best window, phone in another room.

• 50 minutes: Build the reference document you keep rebuilding from memory. The checklist, the template, the standard response. Once, properly, so it stops costing you time every month.

• 50 minutes: Finish something at ninety percent. Not new work, the thing that has been nearly done for two weeks. Nearly done and done are different states and only one of them counts.

What a neglected Command Vault looks like

This vault fails in two directions, and only one of them looks like failure.

The obvious version is emptiness: nothing substantial has moved in weeks. Days are full of responses. You are busy from morning to evening and could not name a single thing you produced. Every meaningful project is at thirty percent. You keep starting and the starts do not accumulate.

The less obvious version is overflow, and it is more common among people who care about this. The Command Vault is the only vault with entries in it. Every other domain is at zero and has been for a month. This looks like productivity and is actually a slow trade of health, money management, home systems, and learning for output, at an exchange rate that seems fine for about a quarter and then does not. Overflow here is what starves the Growth Vault in particular, because learning is the first thing to go when execution expands to fill everything.

Both versions share a symptom worth watching for: you can no longer distinguish between a good day and a busy one. When those two words start meaning the same thing, the vault has stopped measuring anything useful.

The smallest possible first step

Take the one thing that matters most right now, and spend fifteen minutes not doing it. Instead, write down what finished looks like, then write the first three sessions in task-shaped form, each with a stated first action. That is the whole assignment.

Tomorrow, run the first of those three. Twenty-five minutes, phone in another room, one line written at the end about where you stopped. If you are tracking sessions in TaskTreasury, that focus time earns credits like anything else, but the more useful record is the count of finished sessions on one project across one week, because that number predicts whether the project ships far better than how motivated you feel about it. Four sessions in a week on one defined thing will move more than a month of intending to.