Letter

Capture Is Not the Job

Why I built Valinor, and what it is actually for

I have used, by rough count, somewhere north of twenty task managers. Paper notebooks, Things, Todoist, OmniFocus, Notion boards, Apple Reminders, the Gmail snooze button, a Shortcut on my iPhone that dictates a voice note straight into Apple Notes. Each of them promised that if I could just get everything out of my head and into the system, the work would follow. Each of them was excellent at the first half of that promise and silent on the second. I could capture anything. I could not, reliably, finish anything.

It took me an embarrassingly long time to see that this was not a failure of discipline but a failure of category. Every product in this space, including the new AI-powered ones, is a capture tool. It solves the problem of remembering. But remembering was never my problem. My backlog could hold ten thousand items with no degradation in performance, and that capacity had nothing whatsoever to do with whether any of them got done. The tools were optimizing a constraint I did not have while ignoring the one I did.

Valinor is my attempt to build the second half. This letter explains what that means in practice, and then the reasoning behind it.

The four steps

What I eventually worked out, by watching what I actually did on days that went well versus days that did not, is that getting from an intention to a closed loop involves four distinct operations, and a to-do list only performs the first.

Capture is the familiar step: the email arrives, the thought occurs, the meeting ends with an action item, and it gets written down somewhere. This is a solved problem. Any app does it; the AI ones now do it for you, reading the thread and extracting "follow up with Sarah." I have nothing to add here except that stopping at this step is the original sin of the category.

Decomposition is where the work actually begins. "Follow up with Sarah" is not an action; it is an intent. To act on it I have to reconstruct the thread, decide what I am asking her for, and open a draft. Only then is there something I can do. The task, as captured, was three tasks wearing a trench coat, and the reason it sat untouched for a week was that I never wanted to pay the setup cost of unpacking it. Decomposition means breaking the intent into pieces that are each small enough and concrete enough to start without deliberation. When the draft is already sitting there with the ask written in, I finish it in ninety seconds. The difference between those two states is the whole product.

Sequencing is the step nobody builds for. Given a set of well-scoped actions and a bounded window of time, which ones, in what order? I have forty-five minutes before my next meeting. The wall of undifferentiated items does not answer that question; it just stares back. What I need is an ordering that fits the window and, ideally, one that starts with something I can close quickly, because a closed loop changes how I approach the next one.

Closing is the step that produces the outcome. The email is sent, the ticket is resolved, the loop is shut. This sounds trivial, but it is the only step that matters, and it is the step every upstream decision should be designed to make more likely. A system that measures its success by items captured is measuring the wrong thing. Valinor is designed to measure loops closed per unit of available time.

The philosophy underneath

Three claims sit behind that design, and I want to lay out the evidence for each, because I did not arrive at them by intuition alone.

The first claim is that an intent is not an action, and the gap between them is where follow-through dies. The clearest evidence for this comes from Peter Gollwitzer's work on implementation intentions. In a 2006 meta-analysis with Paschal Sheeran covering roughly ninety-four studies, Gollwitzer found that converting a vague goal into a specific if-then plan, one that names the situation and the response in advance, produced a medium-to-large improvement in goal attainment, with an effect size around .65 (Gollwitzer and Sheeran, Advances in Experimental Social Psychology, 2006). The mechanism is that a pre-specified plan removes the deliberation step at the moment of action. You do not have to decide what to do; you have already decided, and the situation simply triggers the behavior. A pre-written draft with the ask in it is an implementation intention. "Follow up with Sarah" is not.

The second claim is that the binding constraint on completion is activation energy, not memory. Whether I do a task is predicted almost entirely by whether it is scoped correctly, and scope is a vector rather than a scalar: it has a direction (is this the right next action?) and a magnitude (is it sized to fit under my threshold for starting?). Different people have different thresholds, and mine varies by time of day and by how much of the week I have already spent. A system that watches what I pick up versus what I leave sitting can learn where my threshold lies and decompose accordingly.

There is a second half to this claim, which is that unfinished work is not inert. Bluma Zeigarnik showed in 1927 that interrupted tasks are remembered and mentally rehearsed more than completed ones, and a century of subsequent work has confirmed that open loops intrude on attention and generate a low-grade stress that accumulates. What matters for product design is a finding from E. J. Masicampo and Roy Baumeister in 2011: simply making a specific plan for an unfinished goal eliminates that intrusion, before the task is done (Masicampo and Baumeister, "Consider It Done!", Journal of Personality and Social Psychology, 2011). The plan itself offloads the cognitive cost. This is why an inbox full of captured-but-unscoped items feels heavy, and why the same items, once decomposed into concrete next actions, feel manageable even before any of them are finished.

The third claim is that momentum is real and can be engineered. Teresa Amabile and Steven Kramer analyzed thousands of daily work diaries and found that the single strongest predictor of motivation and engagement on a given day was not a breakthrough or a reward but the experience of making progress on meaningful work, even small progress (Amabile and Kramer, "The Power of Small Wins," Harvard Business Review, May 2011). Closing a loop, however minor, changes the affective state you bring to the next one. The implication for sequencing is direct: a system should arrange the available window so that it opens with a quick close and builds toward the larger ones, rather than presenting the largest, most intimidating item first and waiting for willpower to show up — contra the eat-the-frog school.

Taken together, these three findings describe a coherent mechanism. Decomposition converts intents into implementation intentions, which raises follow-through and discharges the Zeigarnik load. Activation-threshold scoping ensures each piece is startable. Sequencing by small wins compounds the motivational effect of each close into the next. And time, the forty-five-minute window, is the surface on which all of this gets expressed, which is why Valinor is being built to treat the calendar as an input rather than a separate app.

Why the name

In Tolkien's mythology, Valinor is the land in the uttermost West, guarded by the Valar, the blessed powers. It is not a place where nothing happens; it is a place where what happens is protected, where effort is not dissipated by chaos. That is the feeling I wanted from a work system and never got from a list: not a record of everything I owe, but a protected space in which the things I have decided to do are already scoped, already sequenced, and already waiting for me to close them.

Today the ability to scope and sequence your own work is a large part of what distinguishes a senior individual contributor from a junior one. It is a skill we pay for, and we pay a lot. Valinor is an attempt to give that skill to anyone with an inbox and a calendar — and let them close every loop.

Sources