Reading time, and other numbers worth being honest about

Every blog card carries a number that nobody verifies. "6 min read" is a word count divided by an assumed rate, rounded, and then rendered in the same typeface as the publication date — which is a fact.

It is worth getting close to right, because a reader who finishes in half the promised time stops trusting the next number on the page.

The two mistakes

The first is counting code as prose. A forty-line snippet is not two hundred words of reading; it is either skipped entirely or studied for far longer than any words-per-minute figure predicts. Leaving it in the count makes engineering posts claim roughly double their real length.

The second is false precision. Nothing is gained by "5.8 min", and something is lost: it implies a measurement where there is an assumption.

What we do

Strip fences, strip inline code, strip link URLs while keeping the label, then divide by 225 and round — with a floor of one, so a short note never advertises "0 min read".

const words = prose.split(/\s+/).filter((t) => /[\p{L}\p{N}]/u.test(t));
return Math.max(1, Math.round(words.length / 225));

The filter matters more than it looks. Splitting on whitespace alone counts leftover punctuation as words, which inflates every post by a percent or two — consistently, and in the same direction.

The number we do not show

View counts. We have no analytics backend on this site, so any figure would be a literal invention, and a card is a bad place to start making things up.

Where a number appears, it should be one we could defend.