How Friction finds what people struggle with

Ten thousand public reviews a week, filtered in code, grouped by what repeats, and quoted. Every number here is calculated; almost nothing is written.

Check the arithmetic

WHERE IT COMES FROM

Two public review feeds

Both stores publish reviews without credentials or an agreement. Friction read 9,994 of them in its last scan.

WHAT GETS THROWN AWAY

Ten thousand in, a hundred out

Screened for age, for length and by the crisis check, then for a complaint, then capped at 150. All of it code, before any model sees anything.

WHAT THE MODEL DOES

It groups, and it quotes

It puts reviews describing the same struggle together and copies a short quote from each. It never produces a score, a rank or a priority.

THE ONE WRITTEN FIELD

One sentence is written

Every value is copied from a review or calculated in TypeScript, except the four lenses and 3 sentences. Those carry a badge and a date.

What the model does, and never does

The model does
The model never
  • group reviews describing the same struggle
  • copy a short verbatim quote from each
  • produces a score
  • produces a rank
  • produces a priority

Every quote is checked character by character against the review it came from.

See the full method

Esc

43 results