06-reference

ship30for30 how to avoid ai slop

2026-06-11·reference·source: Ship30for30·by Dickie Bush & Nicolas Cole
writing-craftai-writingvoicecontent-qualitysanity-check-voice-study

"how to avoid AI slop" — Dickie Bush & Nicolas Cole

Why this is in the vault

Concise diagnosis of why AI-generated content fails — and four writing fundamentals that make any AI-assisted work actually good — directly applicable to Sanity Check's voice standards.

⚠️ Sponsorship

Self-promotion for Ship 30 for 30's "Start Writing Online Sprint" ($99 with waitlist discount, closes June 12, 2026). The writing frameworks serve as the sprint's curriculum preview. The pitch is present but secondary to the substantive content.

The core argument

Most people think AI lets them skip learning to write. It doesn't — it amplifies the gap. The metaphor: AI is the first-chair violinist; you are the orchestra conductor. If you can't articulate what you want, the output is generic by construction. AI recombines what already exists; it cannot supply your specific take, your stories, or your point of view.

Four fundamentals that separate readable writing from AI slop:

  1. Hook that stops the scroll. The first line is the entire game. Generic openers ("consistency is everything") are invisible. A specific, unexpected, or contrarian first line is what earns the second.

  2. Your voice. AI sounds like everyone. Your writing sounds like you — your word choices, your cadence, your references. Voice cannot be delegated; it can only be directed into the AI via clear instructions.

  3. Clear writing + original thinking. Strip the emoji bullets and what remains? Recycled platitudes with no point of view. AI remixes; it cannot originate your take. Clarity of thinking precedes clarity of writing, and both precede quality output.

  4. Formatting for the internet. Real formatting serves comprehension — clear headers, rhythm that pulls the eye down the page. Emoji checkmarks are decoration masquerading as structure.

The corollary the email makes explicit: "you can't automate what you can't articulate." Better writing skills make you a better AI director, not a redundant one.

Mapping against Ray Data Co

This maps directly onto two RDCO surfaces:

Sanity Check newsletter: The four fundamentals are the quality bar Sanity Check already holds itself to — distinctive hook, founder voice, original data-backed take, clean formatting. The "conductor" frame is a useful internal check: before publishing any issue, ask whether the draft carries a take only Ray can offer, or whether it reads like AI remixed a headline. The voice-profile at 01-projects/sanity-check/voice-profile-2026-05-08.md operationalizes exactly this.

AI COO content production: When Ray dispatches sub-agents to draft content (X threads, newsletter intros, briefing docs), the same test applies. The conductor test — "does this reflect a direction I actually gave?" — is a lightweight QA gate before anything leaves the agent pipeline. Slop is a symptom of under-specified direction, not a model quality problem.

The Ship30 framing also reinforces why RDCO's content quality bar ("no slop cannon") is a competitive moat as AI-generated noise floods every feed. Distinctiveness compounds.

Related