We had ideas about what we would be able to see
based on other projects, and were working
to make models more robust and scalable.

It was also a bit of experimenting with the platforms to see what else we could see.

In TJ's words, 2026, from memory. TJ thinks the decks of the time say it better; that wording has not been found yet.

Pockets is one person's days,
recorded by what that person carried.

“pockets is about days in the life of a person.”

The person is TJ, and the data is TJ's own, from apps and wearables carried between 29 November 2013 and 8 July 2014.

It is one study in Instrumented People, Places and Things: work from when the Iota team at SapientNitro in Chicago put sensors in rooms, on objects and on people. Its lens is the person.

Quotation: TJ, 2026.

Every study here ends in a model of activity,
and in this one that model is types of day.

“We were thinking about understanding the world and building computational models of behaviors to provide richer context for design.”

“Stories and typologies of people and events, narrated by hard data.”

First: TJ, 2026. Second: SapientNitro, presentation, 23 September 2014, slide 75, set there on two lines. The team also wrote “We tell stories and develop typologies of people and events”: Iota, SapientNitro, presentation, 2013, p. 6.

TJ carried four main streams,
and only Moves runs through the whole period.

Fig. 1 · What each stream covers, 29 November 2013 to 8 July 2014

Stream	What it recorded	Days
Moves	Places and trips, to the second	210
Fitbit	Steps, by the minute	71
RescueTime	Apps and documents on a phone and two computers, in five minutes	14
Basis	A watch: heart rate, skin and air temperature, skin response, steps	3

Blank means not measured, which is not the same as nothing happening. Counted from the exports kept on TJ's drive.

Five more sources sit beside them,
nearly all in February and March.

Fig. 1, continued · The four main streams in gray, with what was added

Stream	What it recorded	Days
The four	Moves, Fitbit, RescueTime, Basis
Netatmo	The home's rooms: temperature, humidity, CO₂, noise	7
SmartThings	Doors, windows and two sets of keys	14
A GPS log	When it had a fix, with no positions used	9
Phone photos	When each was taken, with no photo shown	33
The home in July	Sensors from the Instrumented Homes study	2

The home's sensors record a household and not only TJ. One set of keys is someone else's, labeled Anonymous B.

No day has all four streams, so they can
be read together in only two short stretches.

Fig. 1, continued · The two stretches, marked

Stream
Moves
Fitbit
RescueTime
Basis

Moves, Fitbit and RescueTime: 4 days
Moves, Fitbit and Basis: 3 days
All four: 0 days

Elsewhere 144 days have Moves alone and 57 have Moves and Fitbit. In December, the 11th and 14th lack Fitbit's steps.

So there are three views: every day,
and a close look at each stretch.

All days

222 days

One row a day, with Moves as the backbone.

9–14 December 2013

Place, steps and screen

Six days, each with its streams in lanes on one clock.

18–20 February 2014

Place, steps and the watch

Three days on one scale, so they compare by height.

In each view, a click on a day opens its path, its streams on one clock and its records. TJ chose these three on 3 November 2026.

Every view runs midnight to midnight,
and blank means not measured.

A row or a block is one day, on Chicago's clock, even on trips.
Color shows place and movement, with its key under the rows.
The map is drawn from Moves's own points, with no base map.
The gray streets are TJ's other days.
Point at anything, and it says what it is.

Estimated, not measured: place names, which are Moves's labels and not checked; “the office” and regular places; the day types.

So far the types are set by hand,
and the model is still to come.

The page's day types are thresholds chosen by hand, a first look.
No weekday enters them, yet no Office day falls on a weekend, and 39 of 46 Out days do. That is a check on the rules, and no more.
Next is the model: types drawn from Moves alone, with steps, the watch and screen time kept out as checks.

Its limits: one person, about seven months, no labels. It would describe these days and predict nothing.

We had ideas about what we would be able to see based on other projects, and were working to make models more robust and scalable.

Pockets is one person's days, recorded by what that person carried.

Every study here ends in a model of activity, and in this one that model is types of day.

TJ carried four main streams, and only Moves runs through the whole period.

Five more sources sit beside them, nearly all in February and March.

No day has all four streams, so they can be read together in only two short stretches.

So there are three views: every day, and a close look at each stretch.