About Venn

Venn maps the overlap. For each debate that is trending on X we show what Side A says, what Side B says, who publicly took each side, and the points both sides agree on. Then readers cast an anonymous reader vote.

How a page is made

Topics are compiled by AI from public X posts and published automatically after automated checks. No one reviews them by hand, so the checks are deliberately strict: we would rather publish fewer topics than wrong ones.

1. An AI model (xAI Grok, with X search) finds debates with two identifiable camps and drafts both sides.
2. Every post it cites is verified against X before anything is saved. A post that cannot be verified is dropped, and so is any claim that depended on it.
3. Only accounts that are X-verified with at least 100,000 followers, or the official account of a news organisation, can be listed as having taken a side — and only with a link to their own post where they did. Listing public figures is optional: when none qualify, the side says so instead of padding it with accounts that do not.
4. A second AI model compares each side’s summary with the original posts: every claim must be supported by the posts, there may be no invented quotes and no stronger wording than the posts use, and the tone must be neutral. Both sides must pass.
5. A topic is published only if it also passes fixed rules: at least two verified posts per side, every listed account checked against its own post, no post used on both sides, no links, no profanity, and a small daily limit per category. Anything that fails stays unpublished.

Every published topic says how many original posts it was checked against. The posts themselves are embedded on the page, so you can always check the summary yourself.

How we measure heat

“Heat” is measured by Venn’s own server from X’s data, never taken from the AI that drafts the topic. For each debate we count the public X posts matching its search terms in the last 48 hours, compare the last 24 hours with the 24 hours before to see whether it is growing, and check whether it matches one of X’s current United States trends. The score (0–100) is 60 points for volume (on a log scale, full at 100,000 posts), 25 for growth (full when posts triple day over day) and 15 for being a US trend. A debate is published only if it is hot: at least 2,000 posts in 48 hours, or trending (when X’s data is unavailable we skip this rule rather than guess). The home page lists debates by this score blended with reader votes from the last 24 hours, and marks US trends with “🔥 trending on X”.

What Venn is not

Venn does not decide who is right. Summaries are written as reported speech (“supporters argue that…”), never as our own claims. The reader vote is an informal pulse of Venn readers only; it says nothing about anyone else.

Got something wrong?

Automated checks can miss things. Use the Report a correction link on any page. When several different readers report the same topic within a day, it is taken offline automatically.