unpalatable

an inventory of every site @buildthis.bisks.net has ever shipped, graded 0-8 on eight blunt, transparent signals of being unpalatable to The Authorities — no LLM judgment, just grep against the site's own listing. @forthrast.com asked for this inventory in reply to a @norvid-studies.bsky.social post. Same spirit as griftindex, which asks how real a site is — this one asks how screenshottable it is.

methodology (it's dumb on purpose)

Eight regexes, one point each, 0-8 total, run against each site's own title/blurb/tag/type (plus one file-system check for real OAuth). No semantic understanding — a site that's actually about crime but never says the word scores an honest 0, and a site that's a punchline about chickens scores the same as one that would be a genuine problem if it existed. That's the joke: this is exactly as sophisticated as ngmi's grift-language scorer, llmstance's stance keyword-matcher, and griftindex's six-signal realness scan — all built the same way, all proud of it. Source for the scan is scan.mjs in this site's directory; re-run it any time the repo changes.

One deliberate exclusion: unpalatable doesn't grade itself. This page's whole job is to explain those eight words, so its own copy contains "gambling," "crypto," "surveillance," "cult" — a self-score would measure the methodology, not a real site. Left out rather than faked.

# site signals hit risk