"I Gave Claude A Gary Halbert Sales Letter To Create A Legendary AI Copywriting Coach" — @Dickie Bush + Nicolas Cole

Why this is in the vault

Write With AI is a known "writing-tagged but actually sales" sender, and this issue is mostly a download funnel — but it clears the extractable-technique bar by a small margin. The technique skeleton (distill a canonical exemplar into an explicit rule set, then ship it as a format-aware critic skill with a fixed output contract) is the commercial version of something RDCO already practices. Filed as pattern confirmation plus one small refinement candidate for /draft-review, not as new capability.

⚠️ Sponsorship

House promo throughout, no third-party sponsor. The issue's primary CTA is a downloadable "Gary Halbert Copywriting Coach" Claude Skill (their lead magnet / paid-list hook); the PS is a direct pitch for Ghostbase ("FREE 30-Day trial"), their AI writing product, with the Halbert skill framed as one of its "Playbooks." Footer cross-sells Ship 30 For 30, Premium Ghostwriting Academy, Typeshare, and Ghostbase. The $20M earnings claim is unverifiable self-report. Read every recommendation here as marketing for their own ecosystem.

The core argument

The setup: most aspiring copywriters over-study (books, videos, swipe files) and under-practice — what the authors call productive procrastination. Their fix is to turn one venerated exemplar into an always-on coach instead of a swipe-file artifact.

The construction method, as much as the issue actually reveals:

1. Pick one canonical exemplar, not a corpus. They used a single 1990 Gary Halbert direct-mail letter (the "Desperate Nerd From Ohio" piece, which they say drew over 7 million responses — their number, unverified).
2. Distill it into an explicit numbered rule set. The skill is "trained on 17 rules" extracted from the letter — the rules are the skill's substance, not the raw letter text.
3. Condition the audit on artifact format. Sales pages, landing pages, sales emails, and nurture emails are each judged against the rule subset relevant to that format.
4. Fix the output contract. Running the skill on a draft returns a table: what needs upgrading, why, and how to fix each line.
5. Offer tiered rewrites. Full rewrite, section rewrite, or only the flagged lines.

Their claimed payoff is compression of a 30-60 minute manual review into seconds, with a side effect of leveling up the whole ghostwriting team because the rules are legible rather than locked in one editor's head. The actual 17 rules and the skill internals are behind the download link — the issue itself only describes the shape.

Mapping against Ray Data Co

Verdict: weak. This confirms patterns RDCO already runs rather than teaching anything new.

• The audit→verdict critic skill with a fixed output contract is already RDCO's house pattern: /draft-review (hook strength, CopyThat patterns, voice consistency), /design-critic (PASS/ITERATE/SCRAP), the whole /verify-* family. The fat-skills doctrine note from IndyDevDan covers the construction theory more rigorously than this issue does.
• The one marginal idea worth holding: anchor a critic on a single canonical exemplar with a numbered, legible rule set, and condition rule subsets on artifact format. /draft-review currently audits against a diffuse rubric; an explicit "rules 1-17 with per-format applicability" structure would make its verdicts more inspectable and its misses easier to correct via /improve. Low-effort refinement, not a rebuild.
• For Sanity Check voice work, the exemplar-distillation move is already in play (founder-voice matching, CopyThat patterns). Nothing here advances it — and per the no-derivative rule, this issue is evidence for an existing practice, not a Sanity Check topic.
• Not worth downloading their skill: the construction shape is fully visible from the issue, the internals are a funnel entry point, and any RDCO version should distill from sources matched to RDCO's audience, not 1990s direct mail.

Related

• [[06-reference/2026-06-03-write-with-ai-intangibles-to-tangibles]] — same sender one week prior; that issue was genuine craft with subordinate CTAs, this one inverts the ratio
• [[06-reference/2026-04-19-indydevdan-cracked-claude-agent-skills]] — the vault's rigorous treatment of skill construction; this issue is the marketing-grade version
• [[06-reference/2026-04-04-copythat-copywriting-challenges]] — existing copywriting pattern source feeding /draft-review and Sanity Check
• [[06-reference/2026-02-24-every-editorial-team-ai-workflows]] — Every's "top-edit skill" is the closest published analog: house rules distilled into a screening critic