Rated solutions: sources

Reference ads

• Foreplay #8: Skincu anti-aging body oil, 2:37, 9:16, AI whiteboard explainer (~/research/davidaistar/foreplay/08_authority.gemini.md). Beats, paraphrased: a spokesperson points at a drawn arm on a whiteboard and calls out the age group. A long agitation follows (she hides her arms, picks high necklines). Then three false solutions with rising cost and rising score: drugstore lotion 2/10 (mostly water, evaporates), expensive creams 4/10 (the same water in a nicer jar), spa treatments 6/10 (glow that doesn't last). The product comes in at 100/10, with a mechanism (the skin barrier is oil, so it needs oil) and an ingredient list. Then a quick demo, the offer (50% off, 30-day guarantee, 200k users) and "click below". It is about 100% AI with an uncanny avatar. Gemini guesses its tools from stale training data, so treat those guesses as hints. Keep: the whiteboard device, the rising ladder, a single mechanism that explains every failure, a real agitation beat before the ladder. Change: our presenter is the mom who tried them, not a spokesperson-expert (borrowed person 1.50 in Dawn statics). The band scores 10, not 100. Scores are never drawn as numerals (lane text ban). No lip-sync, so she is staged at the board.
• Smooche "let's rate them": foundation ranking, 96 s, 333 words, 99 days live, 17 duplicates (via output/dawn-beauty-format-2026-10-01/SCRIPTS-v3-1to1-audit.md, S3 section). Beats, paraphrased: a makeup artist sets up the premise ("I've tried all of them, let's rate them"). Then four named products at 6, 5, 7 and 6, each with a grudging credit, a specific failure and a sassy kicker. Then the hero product at 10 with what it does, a borrowed-authority line with an "except for me" joke, and an urgent close with the discount and guarantee. Pace: 3.71 wps speech-only, 0.34 s pauses. Keep: the honest-credit skeleton (credit first, then the failure) and the brisk single-narrator pace. Change: the scores rise instead of zig-zagging. The persona is a mom with the experience, not a pro. Generic methods instead of brand names. The sass, similes and influencer joke are cut (they failed our register and dragged in vo_check).

davidaistar method notes

• REPORT §2 #8, "organic format hijacking": take a viral organic format ("rate these") and graft the product on. Don't clone competitor ads (matches the 10-05 beat-the-leader rule: the ladder's rungs and her rung must fail the competitor-swap test).
• REPORT §3 row 8: authority whiteboard rated "YES animated, medium". The rated structure is strong. An authority person reads as a testimonial in statics, and testimonials lose.
• analysis/diff_copy.md: a "free value" beat that is genuinely useful drives saves. The ladder is that beat (it teaches why sound fails). Scarcity CTAs stay inside our truth gate.
• analysis/diff_animation_styles.md: pick a style per need and fix with an edit pass instead of rerunning. The cartoon-h3 rules already cover both.

Our receipts (10-01 S3 mirror)

• Pack output/concept-packs/dawnbands-s3-rate-them-mirror-2026-10-01.md: a 1:1 Smooche mirror at pixar style, 30 lines, scores 6/5/7/6/10, no stake beat, no root-cause line, "Sixty day guarantee".
• Run output/cartoon-h3/dawnbands-s3-rate-them-mirror-2026-10-01/: VO 86.19 s, 293 words, 3.40 wps, vo_check PASS with 2 warnings (line 28 at 2.31 wps; a 12-word clump at 37.5 s). 32 beats in 9 batches. Assemble passed technically (freezes 0, dark 0, 2 black frames patched). gemini_qa final FAIL 4/10: character inconsistency, "poor lip-sync", a confusing hallway scene, inconsistent product design. Clips parked in gen/_failed-qa for batches 1, 2, 6, 7, 8 and 9. Upscale $0.188.
• Launch (project_launch_2026-10-01_beauty_format_cartoons): ad bf1001-S3 rate them 120249989102510232 in SCALE CBO ad set 120249989091270232, no floor, verdict date 10-08. Spend: $0.40 / 10 impressions through 10-02 (output/dawn-bottleneck-2026-10-03/audit/ad_level_30d.json) and $0.18 lifetime in the 10-05 hub pull. Moved to Testing on 10-03: ad set 120250033290300232, ad 120250033291530232 (output/testing-restructure-2026-10-03/state.json). Unread.
• Open flag at launch: the automated band audit flagged S3 batches 7 and 9 for screen/glow. The pack's own intents 24 and 26 asked for an LED glow, and its product-truth sentence ("never a screen") is obsolete per product-truth.md (display dark when idle).

starpop.ai articles (written versions, read 10-06 from ~/research/davidaistar/starpop/txt/)

• ai-static-ads-that-convert: "3 Reasons Why" is his mid-funnel static; he says it draws a lower CTR but better-qualified clicks that lift landing-page conversion, and splits it by awareness: "3 reasons I love [product]" for product-aware, "3 reasons I ditched [alternative]" for problem-aware. The ditched-alternative version is our ladder card's job in fewer words. Adopted: the CTR expectation and the CPA / click→purchase read (SKILL §11 static test); the reasons frame as a later headline-frame variant in brand voice. Not adopted: first person as the default (mom-first statics 1.53 vs brand voice 1.87); "reasons I love [product]" (product-aware, outside Dawn's problem-aware gate).
• starpop-static-ads-guide: improvise first, then put exact copy in on a second pass. Already ours (dawn-statics --improvise then --recycle).