One engine. It hunts the articles catching fire on X, reads them, tests their claims, and hands you two weapons: a quote-tweet built to detonate, and a sharper, source-tested variant ready to publish. You just decide when to pull the trigger.
Everything up to the fork is built once. Discover, score, ingest — then the same article payload splits into two outputs. Seventy percent shared code, zero duplicated effort.
Grok + web search sweeps for candidate articles per subject.
xAI · web searchRank each candidate on estimated viral potential (Cheap mode).
xAI · heuristicPull clean full-text of the winners. Fetch + readability.
fetch · readabilitySame payload routes to Output A, Output B, or both.
routerScore is an estimate, not a metric. Grok predicts what will spread from the content itself. True like/repost velocity needs a native X API relationship — add it once the loop proves out. The console labels estimates as estimates, never as measured fact.
xAI pay-as-you-go means no subscription, no minimum — just tokens plus search calls. The whole intelligence layer runs for pocket change.
Discover, score, ingest, quote-tweet, claim-test, and variant — end to end.
Search calls amortize across the batch. Discovery is one call, not one per article.
xAI for intelligence. X API for publishing. Two-in-one everywhere it counts.
One subject. One vertical slice. Draft-first. Prove the loop today — scale the surface tomorrow.