most platforms let anything through and hope the comments section sorts it out. FFT Social does the opposite. before a menu is shown to strangers, it has to clear a published bar for what counts as a healthy debate. that bar is the product. this page is the plain-language version.
a good menu is coherent, balanced, and worth arguing about. those three words do a lot of work — every check we run is a way of asking one of them more precisely.
every menu created with our help — by a person, by AI, or by the two together — is reviewed against a checklist before it can be published to the public feed. drafts and private menus skip the check. the checklist is opinionated on purpose. we are not trying to be neutral about what makes a good debate. we're trying to enforce one.
every dish is labeled — supports the menu, opposes it, or it depends. mislabeled takes break the entire vote tally, so this is the first thing we verify and the most important.
a dish that drifts onto a different topic looks fine on its own but ruins the menu. we flag drift and let the author either rewrite or remove the off-topic dish.
a menu where 9 of 10 takes lean the same direction is a loaded question wearing a costume. real debates have at least some opposition. we don't force fake balance — some questions are genuinely lopsided — but we do flag the imbalance and offer to draft missing takes from the source material.
most online arguments are just two sides yelling. we draw on a well-known structured-thinking method (Edward de Bono's Six Thinking Hats) to make sure a menu reaches across multiple modes of thinking, not just pro vs con. concretely, a healthy menu should have takes that hit:
a menu that covers four or more of these reads like a real conversation. a menu that only hits two is a yelling match.
a debate that only argues next year is reactive. a debate that only argues the next century is detached. honest menus do both.
"water is wet" is not a take. "we should care about people" is not a take. a real take is one a reasonable person could vote against without being a monster. we filter out non-takes — slogans, tautologies, virtue statements that nobody would oppose.
bored readers scroll past. menus that don't strike a real nerve don't make it to the public feed. softball questions get refused with a note from us suggesting how to sharpen the menu's premise.
if a take needs a glossary, it's a research paper, not a debate. we flag jargon and overly technical phrasing and offer simpler rewrites the author can accept or reject.
we check that menus and takes don't quietly assume a default gender, race, religion, age, ability, sexuality, nationality, or class. this isn't about whether a take is correct — it's about whether the framing treats some groups as default and others as deviation. when we spot it, we flag it; the author decides how to rewrite. we never auto-edit.
because the bar is the product. anyone can build a forum. anyone can build an AI menu drafter. what's hard to copy is a published, defended, evolving standard for what makes a good debate — enforced at the moment a menu is put in front of strangers, and improved every time an author overrides one of our flags. every menu shipped through the check makes the next check smarter.
we don't share the exact checklist weights, prompts, or thresholds — those are part of the engine. but the criteria above are the whole story of what we're optimizing for. if you can hold a menu up against this list and feel good about it, it's a Food for Thought menu.
