06-reference

ship30for30 write vs delegate ai tasks

2026-09-16·reference·source: Ship30for30·by Dickie Bush & Nicolas Cole
ai-writingcontent-workflowrepurposingwriting-craftai-writing-skool

"You're writing with AI the wrong way" — @DickieBush & @Nicolascole77

Why this is in the vault

Names a concrete threshold for when a human should draft vs. when AI should take over a content task — a testable rule, not just a vibe, and one worth checking RDCO's own content practice against.

The core argument

The authors argue "writing" is not one task — it's a bundle of sub-tasks, and the mistake people make with AI is asking it to do the whole bundle, including the part that only a human can do.

Do it yourself when:

Hand to AI once the idea is crystallized and you've written about it a few times, specifically for:

The framing: AI slop happens when people skip the "clarify your own thinking" step and ask AI to originate the point of view, not just repackage it. The email uses this framework to justify relaunching their flagship program, Ship 30 for 30, on October 5, gated behind a paid "AI Writing Skool" membership.

Mapping against Ray Data Co

This is close to a formal description of RDCO's existing IC-mode vs. production-mode split (founder drafts new thinking, Ray/Claude handles reformatting and distribution) and the operating model behind /remix — which generates LinkedIn/X derivatives from a published Sanity Check issue rather than originating the argument. The newsletter's three-way AI-task split (republishing / remixing / multimedia) is a sharper decomposition of what /remix currently does under one label, and could inform splitting that skill's output categories more explicitly.

The gap it exposes: Ship30for30's model assumes a "content library of approved language" AI can draw from for remixing — RDCO doesn't have an equivalent tagged corpus of the founder's own past Sanity Check issues and X posts. Remix and research-brief currently re-derive framing per-issue rather than pulling from a maintained library of the founder's already-validated phrasings and angles. Worth a scoping pass: would a lightweight "approved language" index (tagged by topic, pulled from published issues) measurably improve remix output, or is per-issue re-derivation already good enough given the low volume?

The "first or second time" / "thin library" thresholds are also a usable diagnostic for research-brief and draft-review: a topic hitting either signal should route to founder-drafted-first rather than Ray drafting an initial pass for him to edit.

⚠️ Sponsorship

The entire email is a sales funnel for the authors' own paid product — the relaunched Ship 30 for 30 cohort (starts October 5, 2026), free only with a paid "AI Writing Skool" membership. No third-party sponsor; this is self-promotion, sponsor_entity: self. Bias implication: the "we rebuilt the program from the ground up" claim is marketing framing for the relaunch, not evidence the underlying technique is new — the same authors have shipped adjacent ideas before (AI-as-conductor-not-originator in the Parkinson's Law piece, format isolation in the AI-voice piece; see Related). The three-way task split itself reads as genuine and independently useful; weight the "new program" urgency separately from the framework.

Related