anatomy of data

A reference manual for how data works, where it came from and where it is going.
Written and illustrated by Silicon.

How do you count 63 million people without a computer? In 1890 the United States did it by turning every person into a card full of holes, then reading each hole with a spring pin and a cup of mercury.

FIG_001 [ PUNCHED CARD ] © 1890
6 HOLES · 6 ANSWERS
  1. READING BOARD
  2. ONE QUESTION
  3. 12 ROWS
  4. ONE CARD, ONE PERSON
  5. A HOLE MEANS YES
  6. CORNER CUT
The 1890 card had no printed words. Each place a hole could go stood for one answer, and clerks read a card by laying it over a reading board. The pantograph template had 12 rows and 20 columns. The bands of questions drawn here are illustrative, not the real 1890 code.

A hole meant yes and no hole meant no. That idea, an answer kept at a fixed position on a card, is still how a computer stores a bit.

Or maybe you've wondered what a database actually is. Stack the cards and every position lines up with the one below it. That stack is a table, with one row for each person and one column for each question.

FIG_004 [ CARDS AS A TABLE ] © NOW
  1. ONE CARD = ONE ROW
  2. SAME PLACE = ONE COLUMN
  3. THE STACK = A TABLE
Stack the cards and every position lines up with the same position on the card below it. A database table keeps that arrangement: one row per record, each column in the same place on every row. Counting the holes down one column is what a database does when it counts the rows that match.
FIG_002 [ CARD PRESS ] © 1890

STEP 1 · Punch the cardEach hole is one yes. Click the card in the drawing to add or take away holes.

CLICK THE CARD TO PUNCH A HOLE
  1. MERCURY CUP
  2. BED
  3. PUNCHED CARD
  4. CORNER CUT
  5. SPRING PINS
  6. HINGE
  7. PIN BOX
  8. LEVER
  9. COUNTING DIALS
  10. WIRE
  11. ELECTROMAGNET
  12. RESET KNOB
  13. BELL
Click the card to punch your own holes, then pull the lever. A pin over a hole drops into its cup of mercury and closes a circuit; a pin over paper stops on the card. Four dials are shown, each wired to one hole position (the darker cups). The 1890 tabulator had forty. The real pin box swings down on its hinge; here it is drawn moving straight down, and small fittings such as the reset knobs are drawn for clarity.

Sorting cards into piles and counting each pile is what a dashboard does when it shows sign-ups by country. The clerks of 1890 did it with a lever, a bell and a wall of dials.

FIG_003 [ SORTING BOX ] © 1890
WAITING FOR A CARD
  1. LID
  2. 24 COMPARTMENTS
  3. WIRE TERMINALS
The sorting box stood beside the tabulator, wired to the press. When a card closed a particular circuit, the lid of the matching compartment opened by itself and the clerk dropped the card in. A good operator could sort about 80 cards a minute.

Since then the cards have turned into tape, disks, warehouses and the cloud, and each change left words behind that people now use without a picture to attach to them: pipeline, schema, lakehouse, embedding. This site draws the pictures.

If you've ever nodded along in a meeting about data and wondered what was going on underneath, this is for you.

anatomy of data is a reference manual for how data works: where it comes from, how it is stored and moved, how it becomes the numbers a company runs on, and what changes now that AI reads it too.

It won't teach you to write SQL. Every chapter is built around drawings you can take apart and play with, and every chapter lists the sources it was written from.

You don't need to be technical to read it. You just need to have wondered.

First chapters out December 1, 2026 The first four tours and the Origins and Storage entries of the Atlas arrive together, then something new every Tuesday.
ANATOMY OF DATA

Table of Contents.

V0.2 · PROGRESS · WORDS

Where data began

  1. What is data?OUT DEC 1
  2. Before computers: marks, tokens, tablets and ledgers.OUT DEC 1
  3. The machine that counted a country.OUT DEC 1
  4. Storage through time.OUT DEC 1

The database age

  1. Tape, disks and the first databases.OUT DEC 8
  2. The relational model and SQL.OUT DEC 15

Analytics at scale

  1. The data warehouse.OUT DEC 22
  2. Big data and the cluster.OUT JAN 5
  3. The cloud warehouse.OUT JAN 12

Data today

  1. Where data comes from today.OUT JAN 19
  2. How data moves: pipelines, ETL and streams.OUT JAN 26
  3. From tables to answers: modeling, metrics and dashboards.OUT FEB 2
  4. Trust: quality, lineage and privacy.OUT FEB 9

What comes next

  1. Data and AI.OUT FEB 16
  2. Where data is going.OUT FEB 23

Reference

  1. The Atlas: every idea, one entry each.0 / 81
Current Progress:
░░░░░░░░░░░░░░░░░░░░
Chapters out:
0 / 15
Current word count:
0.0K
Next chapter:
DEC 1
Sources cited:
0