Field notes  ·  1 August 2026

A lapse is not a deletion.

This was a week with no new feature in it. Almost all of it went into what Meridian says when something has stopped: an expired connection, a question you already answered, an account you want gone. Getting that right matters more than it sounds.

At some point in the last month you probably stopped doing something you meant to keep doing. The ring sat uncharged for a week. The app went unopened while a work thing swallowed you. Nothing dramatic happened. Life just moved.

And then you came back, and the thing that was supposed to be helping you had quietly decided you were a stranger. The streak was gone. The chart started at zero. Somewhere in the interface a small red word implied you had failed.

That is a design choice, not a fact of nature. This week we spent almost all of our time on the other side of it: what Meridian says when it has lost sight of something, and what it remembers when the thing comes back.

What we shipped this week

What the research says about stopping

The early models of self-tracking described a staircase: you collect, you reflect, you act, you improve, and each step leads neatly to the next. Anyone who has actually worn a tracker knows it does not go like that. The model that has held up better is the Lived Informatics Model, from Epstein and colleagues at UbiComp in 2015. It describes tracking as something people move in and out of: they decide to track, they track, they act on it, they lapse, and very often they return. Lapsing is not an error in that picture. It is part of the shape of the thing.

A gap in your data is a fact about your week. It is not a verdict on you.

Take that seriously and it changes what you build. Most tools treat a lapse as a failure state: the streak breaks, the nudge sharpens its tone. Meridian is built to hold the gap instead. When a source of your data goes quiet and then returns, the engine is built to pick that source’s own history back up rather than start from blank, so the weeks you did live are not erased by the week you missed. When there is not enough there to say something honest, it says nothing, which is the harder engineering choice and the right one.

And it should never be wrong about why. Not connected and your permission expired, please reconnect describe two different weeks. One says you never started. The other says something ran out while you were busy living. A mirror that tells you the wrong story about your own effort is worse than a mirror that admits it cannot see.

Where this is going

The beta is small on purpose, and the next stretch is about time rather than features: whether the patterns Meridian finds hold up across a real month, with its skipped days and its uncharged rings, and not just across a tidy good week.

There is no launch in this post. It was a week of small corrections, most of which you would only notice by their absence. But every one of them was the same correction underneath: do not let the product say something about a person’s life that is not true, not even in a status label. That standard is easy to write down and slow to meet, and I would rather spend the weeks meeting it.

Still building deliberately, still showing the seams. If a mirror that can admit what it cannot see sounds like something you would use, requesting access is the way in.

Request beta access Read the manifesto

Written by Davide Tempesta, Founder of Meridian.

Tracking as a cycle, not a staircase: Epstein et al., “A Lived Informatics Model of Personal Informatics,” UbiComp (2015). The model describes people moving into, out of, and back into self-tracking, including lapsing and resuming, rather than progressing in a straight line.

On what self-tracking actually does for self-understanding: van Dijk et al., “Personal Informatics, Self-Insight, and Behavior Change: A Critical Review,” Human-Computer Interaction (2017).

A note on scope: Meridian reflects patterns in your own recorded data. It does not diagnose, treat, or predict anything about your health.