Patterns, not verdicts.

“This pattern looks like…”

Every pattern is a description of your logs, not a conclusion about a person. That’s why Crumbs says “This pattern looks like breadcrumbing”, and never claims anyone is doing it. A pattern isn’t good or bad, and it doesn’t predict what happens next. What it means is up to you.

This pattern looks like breadcrumbing.
Based on 6 logged interactions

An example of how a pattern reads in the app.

Built only from what you log

Crumbs never reads your messages and doesn’t connect to any message app. Patterns use only:

• who texted (they texted or you texted) and when,
• the tone tags you choose,
• how a planned date turned out: Happened, They cancelled, I cancelled or Rescheduled.

The rules are fixed and written down. There’s no AI in the first version. Each person’s patterns are worked out again every time you log something, and each one shows when it was last updated.

Low, Moderate or Strong

Each pattern has a strength label (Low, Moderate or Strong) and says how many logs it’s based on: “Based on N logged interactions”. There are no percentages and no scores out of 100.

Not enough to go on, yet

No pattern shows until a person has enough logs. Until then you’ll see “You’ve logged X of N”, so you know how far along you are. The love bomb pattern needs more logs than the others.

The patterns

• Breadcrumbing pattern

  Uses the timing and direction of what you log, your tags, and how planned dates turned out.

  This pattern looks like breadcrumbing.

• Love bomb pattern

  Uses how often you log and your “Intense” tone tags: a spike early on, then a drop, compared with that person’s own earlier logs. It only becomes an alert at Strong.

  This pattern looks like a love bomb: very intense early on, now cooling.

• Benching pattern

  Uses logged contact that comes and goes, and how planned dates turned out.

  Shown as a benching pattern, with its strength.

• Time to Ghost

  Uses the reply times you log. It shows one of four states and doesn’t predict anything.

  Stable · Slowing · Significantly slowing · Not enough data

• Effort Score

  Counts who started each conversation in your logs. It’s a label on the person’s card, not an alert.

  Balanced · Mostly you · Mostly them

You decide

When a pattern shows up as an alert, you can answer it: Accurate, Not really or Keep watching. “Not really” dismisses it. “Keep watching” means Crumbs holds off on that alert for that person unless the pattern gets stronger. Your answers are feedback, not proof either way.

Crumbs might suggest a status, like “Looks like breadcrumbing. Mark it?”, but you set each person’s status and intent yourself, and your choice wins.

Crumbs is a tracker, not an expert. It doesn’t advise, diagnose or coach.