A Yorimichi reading issue 001

Same numbers. Different stories.

Behind a number sit rules, sequences and boundaries. Follow one reading path through games, moving images and the electricity used by AI.

Edited collection · 3 articles

Article texts: about 14 min, plus exhibits

An edited collection that reads three existing articles through one shared question.

The sum of the estimated reading times for the article texts. It excludes this issue page and time spent exploring the diagrams and exhibits.

Make a prediction, then open a comparison

Turn over a number.

These are calculated examples, adapted from the assumptions in the articles. They are not measurements of a real game or device.

01 · Rules5%If the first chance is 5%, is the rest the same?Open the conditions and compare
Chance of the first item by attempt 20
RuleCalculated chance
Independent 5% each time64.2%
A guarantee on attempt 20 if still waiting100%
One winner in 20 cards, without replacement100%

The same first chance can lead to different waits when the rules change.

Independent draws use 1 − 0.95²⁰. Drawing all 20 cards from the bag includes its winner. The guarantee is a defined rule: if the first 19 attempts fail, attempt 20 succeeds.

Fictional rules for the first item only, not a study of a particular game or a live random draw.

Read the rules of the game
02 · Sequence60fpsDoes an average of 60 fps mean equal intervals?Open the conditions and compare
Six update intervals in the same 100 ms

A · Six evenly spaced intervals

100/6 ms each × 6 ≈ 16.67 ms × 6

B · Five short intervals, then a long one

10 ms × 5 + 50 ms

Each coloured section is the wait until the next update. Both strips span 100 ms on the same horizontal scale.

Both: 6 updates ÷ 0.1 seconds = an average 60 fps

The same average can include a single, much longer wait.

The longest interval in A is about 16.67 ms; in B it is 50 ms. An average alone hides that difference. Exposure and response to input introduce other conditions.

An arithmetic model of timing, not a measurement or reproduction of this screen’s fps, device performance or visual perception.

Read a second of moving images
03 · Scope50%If energy per task is 50%, is the total halved?Open the conditions and compare
Two uses, with the original situation set to 100%
Number of identical tasksTotal energy
Unchanged count · 100%50%
Three times the count · 300%150%

Energy per task × number of tasks = total

Half per task but three times as many tasks gives 1.5 times the original total.

0.5 × 1 = 0.5. 0.5 × 3 = 1.5. Greater efficiency and a rising total can occur together.

An explanatory example with a fixed task. It excludes training, cooling and idle consumption, and does not calculate a real AI service’s energy use.

Read the numbers in the news

From games to moving images, then to the news

A path through three books.

Follow the sequence or begin with the book that catches your eye. Carry one question about numbers from each story to the next.

  1. The next questionFrom “how many” to “at what intervals”: does the same total mean the same journey in between?

  2. The next questionNow move from one at a time to the whole total. Halving each part does not always halve the sum.

An editorial perspective

Three questions beside a number.

  1. What stays constant?

    Does the probability or condition stay the same as the process continues?

  2. What was summarised?

    Has an average hidden the order of events or their variation?

  3. A percentage of what?

    Are per-task and total quantities, or global and local ones, being treated as the same thing?

These subjects share a way of reading conditions, not one common physical law. When a number catches your eye, take one more step towards the question behind it.

Evidence and the role of this collection

A path from the text to its sources.

Each article records its evidence, calculation assumptions and source-check dates. The connections and takeaways on this page are editorial suggestions, not new research findings.

The diagrams are original explanatory models. They do not show measured results or the performance of a real service. Assumptions and calculation conditions appear beside each comparison and in the articles.

Editorial and sourcing policy