Review notes — Gildpress sales page
This is a draft preview built by an unattended autopilot session on 2026-08-24/25. Tim has not reviewed it. Publish is blocked until the items below are cleared. The product itself (`~/Programming/resume-polisher/`) has NOT launched; this page is staged for launch day.
What this page is
The 11-stage single-page-sales pipeline run against Gildpress, the AI-assisted resume and cover letter service. Goal: direct sales ($79/$129/$249 credit tiers). All copy claims trace to the product repo (launch-readiness assessment, launch plan, canonical landing copy, competitor research); the prohibited-claims list from the legal review (B-12) was enforced as a hard constraint: no testimonials, no "passes AI detection" promises, no human-writer claims, PDF-only, tiered guarantee only, Premium framed as credits.
Pipeline results
Adversarial gauntlet (stage 08): 3 cycles, 30 hostile persona reads + 2 red-team assaults. Overall persona average 3.18 → 3.67 → 3.95; would-buy 2/10 → 10/10 (7 Yes, 3 Probably); skim audit 7/7; zero unresolved critical vulnerabilities. The 3.95 sits 0.05 under the 4.0 gauntlet bar; the residue lives in polish surfaces and is documented in `08-test-readers/output/feedback-synthesis.md`. Voice metrics: all PASS on the brand-calibrated scorer (sentence CV 0.72, zero em dashes, zero contractions, zero exclamations, zero AI-blacklist hits at all four tiers, MATTR 0.75). Quality scorecard composite: 4.3/5 (all five dimensions pass).
Publish blockers (Tim sign-offs required)
- Strategy checkpoint (stage 03) was autopilot self-approved: confirm the Big Idea ("a resume service that proves its work"), the PAS-WHY-TRY-BUY framework, and the headline "Your Resume Has Never Done You Justice" (a grammatical repair of your original option B).
- Draft checkpoint (stage 06) was autopilot self-approved: read the full page for voice match and factual accuracy.
- THE PRODUCT IS NOT LIVE. Do not publish until the human-only launch actions are done (LLC, gildpress.com registration, Stripe live mode, legal sign-off on Privacy/Terms, production deploy, one live paid smoke test). CTAs point at `/get-started`, which is a stub on this preview.
- Footer identity: support email and business identity block are placeholders (no domain email or registered entity yet). The executive persona flagged this as a purchase-killer if it ships.
- Premium tier decision (HUMAN-ACTIONS #9): this page sells Premium as 3+3 credits. If you decide Premium includes human review, the pricing section, who-it's-for, and the "who writes it" FAQ all need rework.
- Highest-leverage fix (all 10 test personas): run the pipeline on a demo resume and publish the real artifacts: an annotated sample Voice Authenticity Report, a before/after excerpt, and benchmark aggregates ("across N test resumes: median ATS delta +X"). The hero and evidence-section mocks are CSS illustrations labeled "sample layout" until then; three persona verdicts revert if real artifacts do not replace them at launch. Comment slots are reserved in the evidence section.
- New policy embedded by the gauntlet, needs your sign-off: the FAQ now promises "factual corrections are always free, no credit spent." Mirror it in the Terms or strike it from the page.
- Guarantee wording follows the canonical two-step (re-generate, then 7-day refund from most recent delivery, window never shortened). Two terms questions for the ToS: is the refund available without requesting re-generation first, and is the number of re-generations capped?
- The evidence-section claim "We could find no other service that sends your finished document to independent detectors…" was true per the July competitor research; re-verify before publish.
- Turnaround promise "delivered within hours" is an operational commitment: confirm the deployed pipeline's real-world timing before taking money against it.
- Privacy FAQ now names Anthropic (with API no-training terms) — have legal confirm, and decide whether detector-side (GPTZero/Originality.ai) retention needs disclosure in the Privacy Policy.
- Brand calls made without you, reversible: primary CTA buttons read "Start with my resume" instead of "Begin Your Transformation" (three gauntlet personas read the brand line as hype); the pre-headline dropped the "$100K to $1M+" income band (screenshot-ridicule risk) in favor of your own "professionals who take their careers seriously" line.
Recommendations (not blockers)
- Add a deferred-buyer path: an email-capture line or the free ATS-checker lead magnet (HUMAN-ACTIONS #11). The red team rates it the biggest friction lever: the page requires a job posting in hand at the moment of persuasion.
- Consider a short signed founder note. The emotional-buyer persona's only unresolved critical is that no human being appears anywhere on the page; the honest-automation stance limits copy fixes without you.
- Consider a stated remediation policy for poor detector scores (currently the page only notes that internal remediation passes run pre-delivery).
Where things live
- Pipeline instance: `~/Programming/single-page-sales/pages/gildpress/`
- Final page markdown: `11-polish/output/final-sales-page.md` · scorecard: `11-polish/output/quality-scorecard.md`
- Gauntlet record: `08-test-readers/output/` (3 cycles: feedback, red team, heat maps, synthesis)
- Product repo: `~/Programming/resume-polisher/` (launch actions in HUMAN-ACTIONS.md)
Back to the page