There is a particular feeling that arrives at about eleven at night. You are not working. You are half watching something. And a sentence surfaces, fully formed, from four days ago.
I said I would send that.
Not a vague unease. The exact sentence, the exact person, the exact promise. You know precisely what you owe and to whom. What you cannot explain is how it got to be four days.
The obvious diagnosis is that you forgot. It is also, almost always, wrong. You did not forget. You remembered it on Tuesday morning, moved it to Wednesday, saw it again on Wednesday, and moved it to Thursday. By Friday it had become slightly embarrassing, and a thing that is slightly embarrassing is much harder to pick up than a thing that is merely undone. The commitment did not fall out of your memory. It died somewhere between remembering and doing, in full view, over several days.
That distinction matters, because the entire productivity industry is built on the other diagnosis.
Capture is a solved problem
Look at what the last fifteen years of tools have actually optimised. Quick capture. Global hotkeys. Email to your inbox. Voice notes. Browser extensions. Natural language dates. Every one of those improvements attacks the same failure: the thought existed and did not get written down.
That failure is worth attacking, and by now it is essentially solved. Capture in a modern tool takes under three seconds. If getting things written down were the bottleneck, this problem would be over.
It is not over. Ask anyone who has run the same well organised system for five years whether they still drop things they care about. They do. Their list is immaculate. Their capture is instant. And there are still three items on it that have been rolled forward since March.
So the bottleneck is somewhere else, and it has been somewhere else for years, and the tools have kept optimising the part that already worked.
A list is a record, not a relationship
Here is the uncomfortable structural fact about every task manager ever built: it is passive. It waits.
You open it, and it shows you what you told it. You close it, and it does nothing. If you never open it again, it never objects. The most sophisticated system in the world, with perfect tags and nested projects and a beautiful weekly review, will sit there in complete silence while the single most important thing in it rots.
A notification is not an exception to this. A reminder fires once, at a time you chose, in a moment you did not choose. You swipe it away. It has no opinion about being swiped away. Swipe it away nine times and the tenth is identical to the first: same wording, same tone, same weight. The system has no memory of having asked, and no capacity to care that you keep saying no.
Now compare that with the one mechanism that reliably makes people follow through: another person who knows.
If your co-founder knows you promised the client a proposal on Friday, you send the proposal on Friday. Not because you are more organised on Fridays, but because a second person is holding the other end of it. When you slip, they notice. When you slip again, they notice differently. The pressure escalates, silently and socially, and it escalates whether you like it or not.
That is not a productivity system. It is a relationship. And every study of commitment devices, from the classical ones to the modern behavioural economics versions, keeps landing on the same finding: the mechanisms that work are the ones you cannot quietly opt out of.
Your task manager is one you can opt out of at any moment, invisibly, at no cost, forever.
The snooze is the whole story
If you want to find where your commitments actually die, do not look at what you completed. Look at what you deferred.
A completed task tells you very little. A task you have snoozed four times is telling you something specific and useful, and no ordinary list is listening. To a list, the fourth snooze is identical to the first. Same row, same font, quietly moving to the right, occupying the same one line it occupied in March.
That is the moment worth building for. Not the capture, not the completion, but the fourth deferral. It is the point at which a person has demonstrated, through behaviour rather than intention, that this particular thing is not going to happen on its own.
What should happen at that point is not another reminder. It is a different kind of question. Something closer to: this has moved four times, so either it matters and needs a slot in your calendar today, or it does not matter and should be taken off the list honestly. Those are the only two truthful options. A list offers a third one, which is to keep pretending, and it will let you take that option indefinitely.
Why software has avoided this
There is a good reason nobody builds this, and it is not technical.
An app that escalates is an app that is sometimes unwelcome. Every incentive in consumer software points away from being unwelcome. Retention metrics reward the pleasant. Nobody screenshots the moment their software told them, in plain words, that they had let something slip for a fortnight. The whole category has quietly optimised for the feeling of being organised, which is a different product from actually following through, and a much easier one to sell.
The feeling of being organised is, in fact, one of the more effective forms of procrastination available. Restructuring your projects is genuinely pleasant. It is also indistinguishable, from the outside, from doing nothing.
What we built instead
Wasp is our attempt to build the thing that actually chases you, so it is worth being concrete about what that means in practice rather than in principle.
It watches for deferral, not just for deadlines. Snoozes and stalls are recorded as signals rather than treated as neutral.
It escalates on a ladder. The first contact is quiet and ambient. If nothing happens, the next is direct and names the specific thing. If nothing still happens, the one after names what is at stake. Each rung waits before the next one fires, so this is pressure over days rather than nagging within an hour.
The ladder never loops back to gentle. This is the rule that took the longest to get right, and it is the one that makes the difference. If a task climbs to a direct check in and then crosses some other threshold, it does not reset to a polite nudge. Resetting would teach exactly the wrong lesson, which is that ignoring something long enough makes it soften. In the version we shipped, the pressure only ever goes one way.
And when it reaches the top, it does not offer you a snooze. It offers a choice: do it now, put it in the calendar at a specific time, or drop it deliberately. Dropping is a legitimate answer. Most things on most lists should be dropped, and there is nothing wrong with deciding that. What is not on offer is the fourth deferral, because that is the option that has already failed you four times.
It also learns which things you personally let slip, and moves faster on those. Not everyone drops the same kinds of work. Someone who reliably ships client deliverables and reliably lets admin rot does not need the same cadence for both.
The honest part
This is not a comfortable product, and we have stopped trying to present it as one. An assistant that only ever agrees with you is a very pleasant thing to own and it will not change a single outcome. The value is entirely in the moments you would rather it stayed quiet.
But the alternative is the eleven at night sentence. I said I would send that. Four days, in full view, with the item sitting on a list you looked at every one of those days.
You did not forget. Nothing was chasing you.