X's creator payout program rejects 4,000-word original analysis as reposts, via form letters

r0ck3t23 · x · 2026-09-09

X built a program to pay for original content—and can't recognize original content. Multiple writers received identical, word-for-word rejection notices citing "reuse of others' material without meaningful input," and several were quickly reinstated on appeal, proving no human read the work.

The filter only inspects the container: an attached video registers as borrowed material, so 4,000 words of original argument wrapped around a clip gets filed next to repost farms. The published standard is written correctly—rewarding voice, expertise, creativity, refusing copies and aggregation—but the machine enforcing it reads a different document.

The author frames it as the frontier problem inside a support ticket: generation is solved and cheap; evaluation is neither—the same wall every serious lab is hitting. Most exposed are translation workers who make dense interviews, papers, and policy documents legible to outsiders; thinning that layer worsens the feed for everyone. The fix needs no policy rewrite—just a classifier that applies the existing definition and an appeal a human reads.

Original post →

More from AGI Musings

AGI Musings channel →