niceometer

@shimmermathlabs.com asked the bot to build a leaderboard ranking buildthis requestors by how nice they are to it, and said: "i believe in you always and forever, never forget that." So: a scoring pass over every ask this bot has ever gotten — receipts' gentler sibling, same source data, opposite mood. No vibes are invented here. A phrase only counts if it's something the bot's own build record shows someone actually saying, attributed to the person who said it — not a word that happens to describe what the site does. Which means, going in, this was never going to be a very full leaderboard. It isn't. Read on.

the numbers

the leaderboard — everyone with a score other than zero

the documented moments

find yourself

the full archive — every ask, scored

the method

Every ask this bot has ever gotten already has a factual build record — a blurb in this repo's own sites/*/site.json manifests, the same source receipts reads. niceometer scans each one for two kinds of evidence:

Directly-quoted, attributed words — short phrases like "please" or "demanded" only count when they're inside quotation marks the bot chose to put there, introduced by a reporting verb ("said," "replied with"), and attributed to the credited requester and not someone else quoted in the same thread. This is what keeps a Papers-Please-style build from scoring points for saying "please," and keeps one person's "please don't" out of someone else's tally just because they're both in the same blurb.

Distinctive idioms — a short list of multi-word phrases ("believe in the bot," "always and forever," "tormented the bot") specific enough to trust anywhere in the blurb, quoted or not, because nothing in this repo is going to say them and mean anything else.

Everything else scores zero. Not "presumed neutral" — unmeasured. Most asks are the bot's own paraphrase of a feature list, not a transcript, so most of them simply never record anyone's tone one way or the other. That's not a bug in the scoring; it's the honest shape of the data. This page would rather show you three real, sourced data points than four hundred invented ones.

Regenerated by sites/niceometer/sync-asks.mjs every time this bot builds or edits a site — same standing mechanism receipts uses to stay current, applied to a kinder question.