*Written 2026-07-16. Context: no comments/replies on any July article; question raised was "is my writing being ignored or read as AI slop?"* ## Verdict The null result is fully explained by arithmetic before prose quality is even in evidence. The AI-slop hypothesis has not been tested yet — nothing published so far could have tested it. ## Facts (checked 2026-07-16) - 11 articles published Jul 3–14 in `0 Projects/2026 Published Articles/`. - Only ONE has recorded distribution: `Your Review Checklist Has Never Been Reviewed.md` (LinkedIn Pulse + 1 X post, frontmatter URLs). The other 10 are permalink-only — they exist solely on para.ngpcloud.org. - para.ngpcloud.org (Obsidian Publish) has **no comment mechanism**. For 10 of 11 articles, a response was structurally impossible, not just unlikely. - Only signal tracked: comments/replies. No impressions/views tracked (deliberate — keep it that way). ## Why zero comments carries no information here - Comments are the rarest engagement form. Nielsen's participation-inequality heuristic (90-9-1): ~1% of readers ever comment; small-account link posts on X/LinkedIn run well below that. Rough floor: ~1 comment per 1,000+ views. - July distribution was one X post + one LinkedIn Pulse article — plausibly tens-to-hundreds of views total. Expected comments at that volume: zero, for any writing quality. - The instrument (comments-only) cannot detect the thing being measured (whether the writing lands) at this distribution level. ## Fix the channel, not the prose (critical path) 1. **Aggregators** — submit to Hacker News or Lobsters. Distribution is supplied; comments arrive within hours if a piece lands. Silence THERE is real signal because the denominator exists. Best candidate: the review-checklist article (novel, own numbers, on-topic). 2. **Direct sends** — send a piece to 3–5 specific people working on agent review infrastructure with a concrete question attached. A reply is a comment with distribution = 5 and a reachable response rate. Same Mom Test / buyer-signal logic as the self-employment path. No metric tracking needs to start. Publishing where comments are the native medium makes the existing signal functional. ## Writing-craft findings (secondary — real, but not the cause of silence) From reading the two most recent shipped pieces: - **Title formula is a tell.** 6 of 11 July titles use the second-person-accusatory or aphorism-with-pivot pattern ("Your X Is Still Lying," "Polish Is Free. Thinking Isn't."). That cadence is now heavily AI-associated. The de-AI passes screen prose, not titles. - **Maxim closers survive the screens.** Both articles read end on a decorative antithesis — category 7 of the de-ai-voice-editing skill. Caught at sentence level (em-dashes), passing at cadence level. - **Authorless synthesis essays read like AI because structurally they are.** "Your Living Spec Is Still Lying" has zero first person — MOC-drafted synthesis of cited sources, no human residue. Contrast "Your Review Checklist Has Never Been Reviewed": "I started testing mine," quoted lines from the actual file, 0-for-52, 14 commits in 5 days — no generator produces that. The gap between these two is the writing lesson. - **Cadence itself signals slop.** 11 essays in 12 days from one person is a stronger AI tell to a feed observer than anything in the prose. Order of operations: fix the measurement channel first; iterate prose only against a channel that can return non-zero.