# Same Pond

> A similarity map for a few of your own things. Write three to ten things (mugs, hats, the things on a shelf) with a few traits each, or tick them in a grid. The page measures how alike every pair is, finds the clumps, the loners and the closest pair, and floats everything on a lily pond so look-alikes drift together. A metered run adds a pond keeper's short notes.

Site: https://same-pond.skillsafe.ai/
API: https://same-pond.skillsafe.ai/api.html

## What is free (runs in the browser)
- Reading "name: trait, trait" lines, a tick grid that edits them, and .txt / .csv / .json drop-in.
- Likeness of every pair: shared traits over all traits either thing has (Jaccard).
- Clumps: complete linkage at 50% - every pair inside a clump shares at least half its traits; strongest first, ties to the earlier thing.
- The pond map: a neighbour-weighted, UMAP-style layout from a fixed start, so the same ticks always draw the same pond.
- Downloads: pond .png and .svg, the grid as .csv, a Markdown and JSON report, and a link that reopens the same pond.
- After a change, "Since the notes" lists which things moved between clumps since the keeper wrote.

## What is metered
- One task, "pond": one note per clump, one per loner, a line on the closest pair and a closing line. The page re-checks every note against the ticks (first shared trait quoted, no traits a thing lacks, no added numbers).

## Source
Built on the SkillSafe skill @k-dense-ai/umap-learn (UMAP dimensionality reduction; from k-dense-ai/scientific-agent-skills): https://skillsafe.ai/skill/@k-dense-ai/umap-learn