Strava for Your Work Life Is the Wrong Metaphor. Eight Sleep Is the Right One.

Every few weeks someone describes what we are building as "Strava for your work life," or "Strava for your computer usage." It is a good line. It lands instantly, which is the whole job of an analogy, and it points at something real that the time-tracking category has never delivered.

It is also wrong in a specific and useful way. Working out where it breaks turns out to be the clearest way to say what a tracker for knowledge work should actually be.

What Strava actually did

Strava did not make people run. Running was already the most popular participatory sport on earth and had been for decades.

What Strava did was give an invisible activity a record. Before it, a run happened and then was gone. You could tell someone about it, and they would have to take your word for the distance, the route, the effort. Afterwards, a run left a trace: a line on a map, a number, an artifact that existed after the thing was over.

That is the part people are grasping at when they use the analogy. Knowledge work is the most invisible activity most of us do. You sit down at nine. You get up at six. Somewhere in there, nine hours happened, and you cannot produce a single artifact that shows what they were. Not a commit history, which covers a fraction of it. Not a calendar, which shows what you intended. We have written about the four places those hours actually go, and the honest summary is that most people have never seen their own day.

So yes: knowledge work needs a trace. That much of the analogy holds.

Where it breaks

Strava rests on three assumptions, and knowledge work violates all three.

Assumption one: you chose to do the thing, and you want credit for it.

Nobody accidentally runs ten kilometres. You decided, you got dressed, you went. Strava's entire emotional engine runs on that voluntariness. The kudos, the segments, the leaderboards all assume you are proud of the activity and want it seen.

Almost nothing about a work day fits this. A large share of any given day is busywork, interruption, and things you did not choose. Publishing a leaderboard on top of that does not produce pride. It produces the thing every employee-monitoring vendor has already built, and which knowledge workers correctly hate.

Assumption two: faster is better.

A ten kilometre run in forty minutes beats the same run in fifty. The metric is unambiguous, which is why comparison works at all.

There is no equivalent for a work hour. An hour of hard thinking that produced one paragraph may have been the best hour of the quarter. An hour of fast, fluent, high-volume output may have been the worst. This is the distinction between output and outcome, and it is why every attempt to score knowledge work by throughput has collapsed. Lines of code broke. Story points broke. We have written at length about how hours quietly survived every one of those transitions without ever being a good unit either.

Strava can rank you because running has a pace. Work does not have a pace. Any product that pretends otherwise is measuring the wrong thing confidently.

Assumption three: you press start.

This is the quiet one, and it is the assumption the whole time-tracking industry is built on. Strava works because there is a discrete moment when the activity begins and you mark it.

Work has no such moment. It starts before you notice, gets interrupted, resumes in a different shape, and the boundaries are invented afterwards. Every manual timer ever shipped has run into this, which is why the category's most reliable feature request is "I forgot to start the timer again."

The better analogy is a sleep tracker

Consider what Eight Sleep and Oura actually solved, and notice how different the shape is.

Sleep has three properties. You cannot self-report it. Ask someone how they slept and you get a feeling, not a number, and the feeling is frequently wrong: people routinely believe they slept eight hours when they slept six, and the ones who feel they barely slept often slept fine. It happens without your attention by definition. And the value is in the morning readout, not in watching a live feed at three in the morning.

That is knowledge work, almost exactly.

You cannot self-report your day. This is not a soft claim. In 2025, METR ran a randomised controlled trial on sixteen experienced developers doing real tasks from their own repositories. Afterwards they estimated AI tooling had made them about 20% faster. Measured, they were 19% slower: a roughly forty percentage point error about work they had personally just done, and the estimate barely moved even after they had lived through it. To establish what actually happened, the researchers had to record screens and hand-review 143 hours of video.

That is the sleep problem. Your introspection about your own day is not weak evidence. It is evidence pointing the wrong way, with confidence.

The second property holds too. You do not want to attend to your work being measured. The measurement competing for the attention it is measuring is the central defect of manual time tracking, and it is why the data is worst exactly on the days that were most interesting.

And the third. Nobody wants a live dashboard of their own focus at 2pm. What people want, and what a sleep tracker delivers, is a readout after the fact, when acting on it is still possible but reacting to it is not.

What this changes about the product

The two analogies produce different products, and the differences are concrete.

• Capture is passive, or the data is fiction. Not "easy to start," not "one click." Zero. A sleep tracker that required you to press start would have no users and, worse, would have systematically biased data: you would only capture the nights you remembered, which are not a random sample. The same bias destroys manual time tracking and almost nobody says so out loud.

• The artifact is a retrospective, not a feed. Strava's most-loved feature is not the leaderboard, it is the year in review. Sleep trackers are built almost entirely out of that idea. The unit is the morning summary, the week, the trend, not the live number.

• The comparison is to yourself. A sleep score compares you to your own baseline, not to a leaderboard of strangers. This is the only form of comparison that survives contact with knowledge work, because your baseline is the one context in which "more" and "less" mean anything at all.

• It has to be gentle, or it gets turned off. A sleep tracker that shamed you would be uninstalled within a week. We have argued this for focus tools specifically: nudges outperform blockers, and the reason is retention, not kindness. A tool you resent is a tool with no data.

• Privacy is load-bearing, not a feature bullet. People let a sleep tracker into their bedroom, which is a remarkable amount of trust, and it holds only because the data is theirs. A work tracker is asking for the same trust about the same kind of intimate, unflattering, continuous record. The moment that record has a second audience, the product is a different product.

Why the category never built this

Almost every time tracker on the market descends from billing. Its ancestor is the lawyer's timesheet: you claim an interval, you attach it to a matter, someone else pays for it. That lineage explains the feature set, which is overwhelmingly about invoices, approvals, and defending the number to a third party.

That is a real market and a crowded one. It is also the Strava premise in disguise: you chose the block, you want credit, and there is an audience. Nobody in that lineage was ever trying to answer "what actually happened to my Tuesday," because that question has no invoice attached to it.

Which leaves the more interesting question mostly unbuilt. Not "how many hours do I bill" but "what is the shape of my attention, and is it the shape I think it is." The METR result suggests almost nobody knows the answer, including the people most confident about it.

The honest limit

There is a version of this that goes too far, so it is worth naming.

A sleep score is one number because sleep is one activity. Work is not. Any product that collapses a day into a single productivity score is doing the thing this whole argument objects to: inventing a pace for something that does not have one. The readout should show you the shape of the day and let you judge it. The judgment is not something a tracker is qualified to have, and pretending otherwise is how these tools earn the distrust they get.

Also worth admitting: the model breaks again when the work stops being one person doing one thing. When you have several AI agents running in parallel, the sleep analogy strains along with everything else, and nobody has a good answer yet. Including us.

So: what is Cronus

Not Strava for work. Nobody needs a leaderboard for their Tuesday.

Closer to a sleep tracker for your working hours: it runs without being started, it does not ask you to remember anything, and it gives you a readout of a day you genuinely could not have described yourself. Then it gets out of the way, because the only thing that keeps a passive tracker installed long enough to be useful is that living with it costs nothing.

The trace Strava gave running was valuable because running had never left one. Nine hours a day of the most cognitively demanding work most people will ever do still leaves none. That is the actual gap, and it is a strange one to have gone unfilled for this long.