@aicopycoders/exodus 2026.6.2400
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/assets/references/awareness-framework.md +156 -0
- package/assets/references/cast-video.md +165 -0
- package/assets/references/copy-blocks.md +138 -0
- package/assets/references/creative-strategy.md +990 -0
- package/assets/references/editing-rules.md +81 -0
- package/assets/references/hook-quality-checklist.md +153 -0
- package/assets/references/iterations.md +73 -0
- package/assets/references/scrawls.md +483 -0
- package/assets/references/segments.md +79 -0
- package/assets/skills/exodus/SKILL.md +50 -0
- package/assets/skills/exodus-brand/SKILL.md +117 -0
- package/assets/skills/exodus-browse/SKILL.md +96 -0
- package/assets/skills/exodus-creative/SKILL.md +94 -0
- package/assets/skills/exodus-drive/SKILL.md +118 -0
- package/assets/skills/exodus-foundation/SKILL.md +170 -0
- package/assets/skills/exodus-genesis/SKILL.md +159 -0
- package/assets/skills/exodus-idea/SKILL.md +78 -0
- package/assets/skills/exodus-image/SKILL.md +224 -0
- package/assets/skills/exodus-meme/SKILL.md +183 -0
- package/assets/skills/exodus-primer/SKILL.md +145 -0
- package/assets/skills/exodus-strategist/SKILL.md +217 -0
- package/assets/skills/exodus-swipe/SKILL.md +85 -0
- package/assets/skills/exodus-template/SKILL.md +106 -0
- package/assets/skills/exodus-write/SKILL.md +157 -0
- package/assets/skills/genesis-bots/.env.example +12 -0
- package/assets/skills/genesis-bots/README.txt +29 -0
- package/assets/skills/genesis-bots/SKILL.md +219 -0
- package/assets/skills/genesis-bots/scripts/genesis-stream.mjs +307 -0
- package/assets/skills/genesis-bots/scripts/genesis-stream.py +220 -0
- package/dist/bin/exodus.d.ts +2 -0
- package/dist/bin/exodus.js +177 -0
- package/dist/commands/__tests__/genesis.hookgate.test.d.ts +1 -0
- package/dist/commands/__tests__/genesis.hookgate.test.js +12 -0
- package/dist/commands/brand.d.ts +24 -0
- package/dist/commands/brand.js +292 -0
- package/dist/commands/browse.d.ts +9 -0
- package/dist/commands/browse.js +132 -0
- package/dist/commands/creative.d.ts +2 -0
- package/dist/commands/creative.js +223 -0
- package/dist/commands/doctor.d.ts +21 -0
- package/dist/commands/doctor.js +590 -0
- package/dist/commands/drive.d.ts +2 -0
- package/dist/commands/drive.js +154 -0
- package/dist/commands/foundation.d.ts +2 -0
- package/dist/commands/foundation.js +468 -0
- package/dist/commands/genesis.d.ts +64 -0
- package/dist/commands/genesis.js +564 -0
- package/dist/commands/idea.d.ts +53 -0
- package/dist/commands/idea.js +333 -0
- package/dist/commands/image.d.ts +2 -0
- package/dist/commands/image.js +249 -0
- package/dist/commands/init.d.ts +13 -0
- package/dist/commands/init.js +70 -0
- package/dist/commands/keys.d.ts +2 -0
- package/dist/commands/keys.js +101 -0
- package/dist/commands/meme.d.ts +12 -0
- package/dist/commands/meme.js +272 -0
- package/dist/commands/migrate.d.ts +3 -0
- package/dist/commands/migrate.js +158 -0
- package/dist/commands/primer.d.ts +2 -0
- package/dist/commands/primer.js +540 -0
- package/dist/commands/read-doc.d.ts +3 -0
- package/dist/commands/read-doc.js +62 -0
- package/dist/commands/status.d.ts +6 -0
- package/dist/commands/status.js +205 -0
- package/dist/commands/swipe.d.ts +2 -0
- package/dist/commands/swipe.js +398 -0
- package/dist/commands/template.d.ts +2 -0
- package/dist/commands/template.js +225 -0
- package/dist/commands/whoami.d.ts +2 -0
- package/dist/commands/whoami.js +51 -0
- package/dist/lib/ad-library-write.d.ts +15 -0
- package/dist/lib/ad-library-write.js +67 -0
- package/dist/lib/assets.d.ts +3 -0
- package/dist/lib/assets.js +23 -0
- package/dist/lib/backend-hint.d.ts +2 -0
- package/dist/lib/backend-hint.js +22 -0
- package/dist/lib/cc-command.d.ts +1 -0
- package/dist/lib/cc-command.js +13 -0
- package/dist/lib/client.d.ts +23 -0
- package/dist/lib/client.js +213 -0
- package/dist/lib/format.d.ts +8 -0
- package/dist/lib/format.js +172 -0
- package/dist/lib/intel-format.d.ts +1 -0
- package/dist/lib/intel-format.js +38 -0
- package/dist/lib/keys-sync.d.ts +16 -0
- package/dist/lib/keys-sync.js +88 -0
- package/dist/lib/layout.d.ts +38 -0
- package/dist/lib/layout.js +153 -0
- package/dist/lib/load-env.d.ts +1 -0
- package/dist/lib/load-env.js +47 -0
- package/dist/lib/pixar-orchestrator-types.d.ts +54 -0
- package/dist/lib/pixar-orchestrator-types.js +1 -0
- package/dist/lib/poll.d.ts +14 -0
- package/dist/lib/poll.js +32 -0
- package/dist/lib/preflight.d.ts +18 -0
- package/dist/lib/preflight.js +56 -0
- package/dist/lib/prompts.d.ts +9 -0
- package/dist/lib/prompts.js +99 -0
- package/dist/lib/reel-write.d.ts +14 -0
- package/dist/lib/reel-write.js +59 -0
- package/dist/lib/release.d.ts +4 -0
- package/dist/lib/release.js +81 -0
- package/dist/lib/scaffold.d.ts +8 -0
- package/dist/lib/scaffold.js +67 -0
- package/dist/lib/scout-hydrate.d.ts +1 -0
- package/dist/lib/scout-hydrate.js +10 -0
- package/dist/lib/state.d.ts +6 -0
- package/dist/lib/state.js +49 -0
- package/dist/lib/template-constants.d.ts +4 -0
- package/dist/lib/template-constants.js +85 -0
- package/dist/lib/version.d.ts +1 -0
- package/dist/lib/version.js +19 -0
- package/package.json +30 -0
|
@@ -0,0 +1,990 @@
|
|
|
1
|
+
# Creative Strategy Agent Reference
|
|
2
|
+
|
|
3
|
+
> Comprehensive reference for training an AI creative strategist agent. Covers the full Luke Mills / Mario Castelli (Genesis) framework for research, ideation, copy, images, funnels, and testing. Compiled from 120+ wiki pages in the Creative Strategy Knowledge Base.
|
|
4
|
+
|
|
5
|
+
---
|
|
6
|
+
|
|
7
|
+
## Table of Contents
|
|
8
|
+
|
|
9
|
+
1. [The Philosophy](#the-philosophy)
|
|
10
|
+
2. [Meta-Frameworks](#meta-frameworks)
|
|
11
|
+
3. [Research Phase](#research-phase)
|
|
12
|
+
4. [Ideation Phase](#ideation-phase)
|
|
13
|
+
5. [Writing Phase -- Copy](#writing-phase----copy)
|
|
14
|
+
6. [Writing Phase -- Hooks](#writing-phase----hooks)
|
|
15
|
+
7. [Editing Phase](#editing-phase)
|
|
16
|
+
8. [Image Phase -- SCRAWLS](#image-phase----scrawls)
|
|
17
|
+
9. [Image Phase -- Standard Types](#image-phase----standard-types)
|
|
18
|
+
10. [Video Phase](#video-phase)
|
|
19
|
+
11. [Funnel Phase -- Advertorials and Landing Pages](#funnel-phase----advertorials-and-landing-pages)
|
|
20
|
+
12. [Testing and Iteration](#testing-and-iteration)
|
|
21
|
+
13. [Genesis Bot Catalog](#genesis-bot-catalog)
|
|
22
|
+
14. [End-to-End Workflow](#end-to-end-workflow)
|
|
23
|
+
15. [Key Principles and Rules](#key-principles-and-rules)
|
|
24
|
+
|
|
25
|
+
---
|
|
26
|
+
|
|
27
|
+
> **v2 framework set (the depth lives in sibling reference files).** This file is the
|
|
28
|
+
> comprehensive encyclopedia; the v2 system refines several pieces into their own focused
|
|
29
|
+
> references, pointed to from the `exodus-strategist` framework index:
|
|
30
|
+
> - **Strategy / Segments** — spine + facets, Chad-logic gap ranking, the Strategy Map → `segments.md`
|
|
31
|
+
> - **Copy Blocks (P3C2)** — Pain · Promise · Proof · Constraints · Curiosity, the Persuasion Equation, CRAVES → `copy-blocks.md`
|
|
32
|
+
> - **CAST** — the video-format grammar (Conceit · Actor · Style · Terrain) → `cast-video.md`
|
|
33
|
+
> - **CASHED** — the iteration knobs (CASH + Edit + Demographic) and IDENTIFY → `iterations.md`
|
|
34
|
+
> - **Editing** — steer-prompts, sentence rules, anti-AI list, the feedback-bakes-back loop → `editing-rules.md`
|
|
35
|
+
>
|
|
36
|
+
> Read those when you need the v2 depth; the sections below remain the broader context.
|
|
37
|
+
|
|
38
|
+
---
|
|
39
|
+
|
|
40
|
+
## The Philosophy
|
|
41
|
+
|
|
42
|
+
### Creative Strategy Is Not Linear
|
|
43
|
+
|
|
44
|
+
The creative strategy process is inherently chaotic and iterative. You consume information across product, market, and avatar domains in a round-robin fashion. Each domain contextualizes the others. You detect patterns, form hypotheses, test them against data, and repeat.
|
|
45
|
+
|
|
46
|
+
### The Creative Strategist's Job
|
|
47
|
+
|
|
48
|
+
Mario Castelli defines it as being "paid to feel" -- embodying the prospect's experience and using that intuition to guide AI output. The bots generate; the strategist curates, reframes, and decides. "The bots can't have a human experience. They can try to put that into words, but there's an X factor."
|
|
49
|
+
|
|
50
|
+
### Luke's "Bookends" AI Philosophy
|
|
51
|
+
|
|
52
|
+
You THINK first, then AI fills in gaps, then you THINK last. AI is a powerful pattern detection machine, but it will find patterns and invent narratives that aren't really there. You must always be in control of the process.
|
|
53
|
+
|
|
54
|
+
### "Less Is More" with AI
|
|
55
|
+
|
|
56
|
+
Over-engineering AI inputs makes it behave like an "overeager intern trying to use everything." Give minimum information and let it have confidence to figure things out. Don't dump the entire foundation doc into every prompt.
|
|
57
|
+
|
|
58
|
+
---
|
|
59
|
+
|
|
60
|
+
## Meta-Frameworks
|
|
61
|
+
|
|
62
|
+
These are the theoretical foundations that inform every decision in the creative strategy process.
|
|
63
|
+
|
|
64
|
+
### BAD Method (Why Layer)
|
|
65
|
+
|
|
66
|
+
What any ad is fundamentally doing:
|
|
67
|
+
|
|
68
|
+
- **B -- Belief**: What beliefs can I shift? Every ad challenges, reinforces, or introduces beliefs about the problem, market, solution, or product.
|
|
69
|
+
- **A -- Attention**: How can I get maximum attention? Hooks, pattern interrupts, visual style, scroll-stopping.
|
|
70
|
+
- **D -- Desire**: What desires can I provoke? Amplifying pain, painting the dream state, creating urgency.
|
|
71
|
+
|
|
72
|
+
### ABCD Framework (Evolution of BAD)
|
|
73
|
+
|
|
74
|
+
Maps four things any ad does at a high level:
|
|
75
|
+
|
|
76
|
+
1. **Attention** -- What does your market already pay attention to?
|
|
77
|
+
2. **Beliefs** -- What do they currently believe? What must they believe about your product?
|
|
78
|
+
3. **Constraints** -- What's holding them back? (Beliefs + emotional investment: values, identities, past trauma)
|
|
79
|
+
4. **Desires** -- What do they want to move toward or away from?
|
|
80
|
+
|
|
81
|
+
The most powerful application is **gap mapping**: map their current state across ABCD, map the state they need to reach for buying to be inevitable, and your marketing is the bridge between those two states. Each gap can become an ad concept.
|
|
82
|
+
|
|
83
|
+
### CASH Method (How Layer)
|
|
84
|
+
|
|
85
|
+
Four-part framework for constructing ads:
|
|
86
|
+
|
|
87
|
+
1. **Concept** -- The raw idea. What belief, constraint, or desire is being addressed?
|
|
88
|
+
2. **Angle** -- The lens: whistleblower, contrarian, story, paradoxical question, new discovery, lost wisdom, conspiracy, expert-based, comparison, test/experiment
|
|
89
|
+
3. **Style** -- The format: static, meme, UGC, founder, expert-talking-head, podcast, street interview, comedy, animation, parody
|
|
90
|
+
4. **Hook** -- The phrasing or visual that opens the ad
|
|
91
|
+
|
|
92
|
+
CASH enables combinatorial ad generation: take a winning concept and test across multiple angles. Take a winning angle and test across multiple styles. Take a winning style and test different hooks.
|
|
93
|
+
|
|
94
|
+
**CASH → CASHED (iterating a winner).** When the job is to vary a *proven* ad rather than construct a new one, CASH extends to **CASHED** — the same Concept/Angle/Style/Hook plus **E**dit (tone/length/pacing dials) and **D**emographic (the whole segment, the biggest swing). The discipline is "turn exactly one knob," decoded with IDENTIFY first. Full model in `iterations.md`.
|
|
95
|
+
|
|
96
|
+
### Five Functions of an Ad
|
|
97
|
+
|
|
98
|
+
Every ad performs one or more of these:
|
|
99
|
+
|
|
100
|
+
1. **Get Attention** -- Stop the scroll. Without this, nothing else matters.
|
|
101
|
+
2. **Shift a Belief** -- Change what the prospect believes about problem, solution, or vendor.
|
|
102
|
+
3. **Eliminate Constraints** -- Address limiting beliefs charged with emotion.
|
|
103
|
+
4. **Increase Desire/Pain** -- Amplify urgency by painting the dream state or agitating pain.
|
|
104
|
+
5. **Give an Excuse to Act Now** -- Scarcity, urgency, guarantees, risk reversal.
|
|
105
|
+
|
|
106
|
+
A single ad doesn't need to do all five. Each function can become its own ad concept.
|
|
107
|
+
|
|
108
|
+
### Alignment Model (Replaces Schwartz Awareness)
|
|
109
|
+
|
|
110
|
+
Luke's replacement for Eugene Schwartz's five stages. Key shift: awareness (knowing something) is insufficient. What matters is alignment (accepting and believing something).
|
|
111
|
+
|
|
112
|
+
Three parallel axes (not linear stages):
|
|
113
|
+
|
|
114
|
+
1. **Problem Axis (Red)** -- Moving people toward your Unique Mechanism of Problem (UMP)
|
|
115
|
+
2. **Solution Axis (Green)** -- Moving people toward your Unique Mechanism of Solution (UMS)
|
|
116
|
+
3. **Vendor/Brand Axis (Blue)** -- Building trust in you as the category leader
|
|
117
|
+
|
|
118
|
+
People arrive with existing mental models. The further out and more specific a competing belief, the harder to shift. Accommodation (yes AND) works better than confrontation for deeply held beliefs.
|
|
119
|
+
|
|
120
|
+
### Awareness Levels (Practical Application)
|
|
121
|
+
|
|
122
|
+
From most to least aware:
|
|
123
|
+
|
|
124
|
+
| Level | What They Know | Ad Strategy |
|
|
125
|
+
|-------|---------------|-------------|
|
|
126
|
+
| Product-aware | Knows your product | Direct demos, UGC reviews, offers |
|
|
127
|
+
| Solution-aware | Knows solutions exist | Mechanism differentiation, "why different" |
|
|
128
|
+
| Problem-aware | Knows they have a problem | Call out symptoms, explain cause, present solution |
|
|
129
|
+
| Symptom-aware | Experiencing symptoms | Call out experiences, connect to root problem |
|
|
130
|
+
| Unaware | Doesn't know they have a problem | Entertainment, curiosity, storytelling |
|
|
131
|
+
|
|
132
|
+
Scaling requires moving DOWN the awareness ladder. Product-aware exhausts fastest. Unaware is the largest audience.
|
|
133
|
+
|
|
134
|
+
For a standard batch of 10 ads: 6 unaware/problem-aware (4 unaware, 2 problem aware), 4 solution/product-aware.
|
|
135
|
+
|
|
136
|
+
### Creative Diversity Matrix
|
|
137
|
+
|
|
138
|
+
5 Segments x 5 Awareness Levels = 25 Buckets. True creative diversity means producing concepts that cover as many buckets as possible, not clustering in the obvious ones.
|
|
139
|
+
|
|
140
|
+
Four levels of diversity:
|
|
141
|
+
1. **Segments** -- Different audience personas
|
|
142
|
+
2. **Awareness** -- Where on the spectrum
|
|
143
|
+
3. **Source** -- Where the concept came from (STORMING categories)
|
|
144
|
+
4. **Style** -- Format and visual treatment
|
|
145
|
+
|
|
146
|
+
Meta's Andromeda system rewards creative diversity. The algorithm matches diverse creatives to diverse audience pockets.
|
|
147
|
+
|
|
148
|
+
This 5×5 matrix is the working simplification of a richer strategy-first model: segments as an **outcome spine with facets hanging off it**, with awareness as the layer *above*, and gaps ranked by Chad logic into a living **Strategy Map**. "Segments are the biggest unlock" — a slice nobody serves wins even with imperfect copy. Full model in `segments.md`.
|
|
149
|
+
|
|
150
|
+
---
|
|
151
|
+
|
|
152
|
+
## Research Phase
|
|
153
|
+
|
|
154
|
+
### Maximum Consumption Research Process
|
|
155
|
+
|
|
156
|
+
Foundational research when entering a new market or starting strategy for a client. ~2 hours investment.
|
|
157
|
+
|
|
158
|
+
**Three Pillars (Round-Robin):**
|
|
159
|
+
|
|
160
|
+
1. **Product** -- Features, benefits, backstory, how it works, what makes it different
|
|
161
|
+
2. **Market** -- Competitors, organic content (YouTube, TikTok), books on Amazon, zeitgeist, cultural touchstones
|
|
162
|
+
3. **Avatar** -- Segments, pain points, survey data, comments, emotional triggers, buying behaviors
|
|
163
|
+
|
|
164
|
+
**The Pattern-Hypothesis Loop:**
|
|
165
|
+
1. Consume (actively, not passively)
|
|
166
|
+
2. Detect patterns -- recurring motifs, language, themes, objections
|
|
167
|
+
3. Ask why -- form hypotheses about why each pattern exists
|
|
168
|
+
4. Test -- go back to data to validate
|
|
169
|
+
5. Repeat
|
|
170
|
+
|
|
171
|
+
**Use multiple AI models in parallel** -- each has a different "personality" and catches different things.
|
|
172
|
+
|
|
173
|
+
### Comment Mining
|
|
174
|
+
|
|
175
|
+
One of the most important research activities. Real customers describe real scenes, use real language, and surface angles no brainstorming session would produce.
|
|
176
|
+
|
|
177
|
+
**Where to mine:** Ad comments (yours and competitors'), Amazon reviews (scrape with Apify), Facebook groups, Reddit, TikTok/Instagram comments on outlier posts.
|
|
178
|
+
|
|
179
|
+
**What to look for:** Stories, specific language/metaphors, objections, horror stories about alternatives, unexpected use cases, emotional reactions.
|
|
180
|
+
|
|
181
|
+
**Process:**
|
|
182
|
+
1. Scrape all comments
|
|
183
|
+
2. Paste into multiple AIs
|
|
184
|
+
3. Ask: "Find all the patterns. What's the mechanism people are connecting with? What's the intuitive interoceptive mechanism?"
|
|
185
|
+
|
|
186
|
+
**Shadow Account Method:** Make a new TikTok/Instagram account. Follow influencers in your market. Saturate your brain. Look for outlier posts (50K average, one got 1M). 2-3 minutes daily.
|
|
187
|
+
|
|
188
|
+
### Foundation Documents
|
|
189
|
+
|
|
190
|
+
Run these Genesis bots first to generate foundation materials:
|
|
191
|
+
|
|
192
|
+
- **Copy Blocks Extract Bot** -- Extracts Pain, Promise, Proof, Curiosity, Constraints from any sales page
|
|
193
|
+
- **Build-A-Buyer Bot** -- Full market research report for a market or segment
|
|
194
|
+
- **Offer Brief Bot** -- Compiles all offer/product information
|
|
195
|
+
- **Segment Surgeon Bot** -- Breaks market into definable segments
|
|
196
|
+
- **Master Concept Bot** -- Generates 300+ new ad concepts (token-intensive, run once)
|
|
197
|
+
|
|
198
|
+
### Keyword Research by Vertical
|
|
199
|
+
|
|
200
|
+
For each vertical, research across five keyword categories:
|
|
201
|
+
|
|
202
|
+
1. **Condition/Symptom Keywords** -- What they're experiencing (clinical + colloquial)
|
|
203
|
+
2. **Emotional/Frustration Keywords** -- How it makes them feel
|
|
204
|
+
3. **Demographic/Identity Keywords** -- Who they are
|
|
205
|
+
4. **Direct Response Phrases** -- The promises and hooks that convert
|
|
206
|
+
5. **Competitor/Alternative Keywords** -- "vs X", "without Y", "better than Z"
|
|
207
|
+
|
|
208
|
+
These feed spy tool research, hook generation, copy blocks, and competitor positioning.
|
|
209
|
+
|
|
210
|
+
---
|
|
211
|
+
|
|
212
|
+
## Ideation Phase
|
|
213
|
+
|
|
214
|
+
### Unique Mechanism (UMP + UMS)
|
|
215
|
+
|
|
216
|
+
The specific, named reason why your solution works when everything else has failed.
|
|
217
|
+
|
|
218
|
+
**UMP (Unique Mechanism of Problem):** The hidden reason behind the problem. Example: "Sensory Confusion" in dogs -- the real reason they pull on leashes is sensory overload, not disobedience.
|
|
219
|
+
|
|
220
|
+
**UMS (Unique Mechanism of Solution):** How your solution uniquely addresses the UMP. Example: Becoming the most interesting thing to your dog.
|
|
221
|
+
|
|
222
|
+
**What makes a great mechanism:**
|
|
223
|
+
1. Externalizes blame (removes fault from buyer)
|
|
224
|
+
2. Validates the buyer's ego
|
|
225
|
+
3. Hard to argue against (aligns with pre-existing beliefs)
|
|
226
|
+
4. Gives the buyer control
|
|
227
|
+
5. Provides genuine insight (creates a real "aha moment")
|
|
228
|
+
|
|
229
|
+
**Finding your mechanism (speed-run):**
|
|
230
|
+
1. List every feature of your product
|
|
231
|
+
2. Identify the most UNIQUE feature
|
|
232
|
+
3. Build mechanism around that feature
|
|
233
|
+
4. Match a UMP that makes your feature the logical solution
|
|
234
|
+
5. Apply the SIN Test: Simple, Intuitive, New
|
|
235
|
+
|
|
236
|
+
### Universal Mechanism Framework
|
|
237
|
+
|
|
238
|
+
The system for creating mechanisms across any market:
|
|
239
|
+
|
|
240
|
+
```
|
|
241
|
+
MECHANISM = [WHAT: Level] x [HOW: Type] + [WHY: Root Cause] + [WHERE/WHEN: Context]
|
|
242
|
+
```
|
|
243
|
+
|
|
244
|
+
**WHAT (3 Levels):** Structure, Function, Element
|
|
245
|
+
|
|
246
|
+
**HOW (4 Problem Types / UMP):** Excess, Deficit, Corruption, Misdirection
|
|
247
|
+
|
|
248
|
+
**HOW (4 Solution Types / UMS):** Amplification, Optimization, Restoration, Realignment
|
|
249
|
+
|
|
250
|
+
**8 Mechanism Characteristics to score against:**
|
|
251
|
+
1. Simple Structure (3-part causal chain)
|
|
252
|
+
2. Leverages Entropy (natural decline)
|
|
253
|
+
3. Specific Enemy
|
|
254
|
+
4. Visual Metaphor ("Leaky Gut", "Metabolic Furnace")
|
|
255
|
+
5. Discovery Story
|
|
256
|
+
6. Creates Urgency
|
|
257
|
+
7. Includes Proof
|
|
258
|
+
8. Shifts Paradigm
|
|
259
|
+
|
|
260
|
+
### Three Master Narratives
|
|
261
|
+
|
|
262
|
+
Every mechanism wraps in one of three narrative structures (all share Order > Chaos > New Order):
|
|
263
|
+
|
|
264
|
+
1. **Restoration** -- "We had it, enemy took it, fight to get it back." Use when there's a clear enemy, nostalgia is powerful.
|
|
265
|
+
2. **Revelation** -- "Hidden truth exists, it's discovered, changes everything." Use when market is skeptical, curiosity-driven.
|
|
266
|
+
3. **Rebirth** -- "Everything breaks, hit bottom, rise transformed." Use when market has tried everything, transformation is the goal.
|
|
267
|
+
|
|
268
|
+
Most powerful mechanisms combine elements of multiple narratives.
|
|
269
|
+
|
|
270
|
+
### Mechanism Chain (Body Copy Structure)
|
|
271
|
+
|
|
272
|
+
Structured 7-step sequence for presenting why a product works:
|
|
273
|
+
|
|
274
|
+
1. **Name the root cause** -- The real underlying problem
|
|
275
|
+
2. **Explain the damage pathway** -- Visual, vivid language
|
|
276
|
+
3. **Present the obvious fix** -- What they already think the solution is
|
|
277
|
+
4. **Kill the obvious fix** -- Why the obvious solution doesn't work
|
|
278
|
+
5. **Introduce product components** -- Each ingredient/feature and why
|
|
279
|
+
6. **Recap the chain** -- Lock the logic into memory
|
|
280
|
+
7. **Narrow to your product** -- Position as the only logical conclusion
|
|
281
|
+
|
|
282
|
+
### STORMING Framework (Copy Ideation)
|
|
283
|
+
|
|
284
|
+
Eight source categories for generating ad concepts:
|
|
285
|
+
|
|
286
|
+
1. **Swipes** -- Competitor ads, spy tools
|
|
287
|
+
2. **Templates** -- Proven ad templates
|
|
288
|
+
3. **Organic** -- Outlier organic content
|
|
289
|
+
4. **Research/Comments** -- Comment mining, reviews
|
|
290
|
+
5. **Matrix** -- Segments x awareness level coverage gaps
|
|
291
|
+
6. **Internal Vectors** -- Problem/solution angles from product data
|
|
292
|
+
7. **New Styles** -- Untested creative formats
|
|
293
|
+
8. **Gambits** -- Bot-generated concepts (MarioBot, Ad Lottery)
|
|
294
|
+
|
|
295
|
+
**Recommended mix for 10 ads:** 1-2x from each source category.
|
|
296
|
+
|
|
297
|
+
### Ad Concept Generation
|
|
298
|
+
|
|
299
|
+
Three sources of ad ideas, each with different risk/reward:
|
|
300
|
+
|
|
301
|
+
1. **Internal (CASH Method)** -- Pull facts and seeds from research, combine into concepts. Middle tier: good, reliable, grounded.
|
|
302
|
+
2. **External (Swiping)** -- Swipe from organic and paid ads. Safest: highest floor, lowest ceiling.
|
|
303
|
+
3. **Intuition** -- Let ideas come naturally after brain saturation. Most losers AND most big outlier winners.
|
|
304
|
+
|
|
305
|
+
### Swiping Strategy
|
|
306
|
+
|
|
307
|
+
Luke: "If you can only do one thing, swiping is better."
|
|
308
|
+
|
|
309
|
+
**Where to find swipes:** Facebook Ad Library, AdSpy, Foreplay, organic content (TikTok, YouTube, Instagram), ClickBank/Digistore, adjacent markets.
|
|
310
|
+
|
|
311
|
+
**Modular swiping (not wholesale):** Break winning ads into CASH DNA and swipe individual components. Swipe the CONCEPT from Ad A with a different ANGLE. This creates novel combinations while maintaining proven DNA.
|
|
312
|
+
|
|
313
|
+
**What to swipe:** Structure, hook formulas, visual style, angle, copy rhythm. **What NOT to swipe:** Specific claims that don't apply, creative from mismatched sophistication levels.
|
|
314
|
+
|
|
315
|
+
### Creative Coverage
|
|
316
|
+
|
|
317
|
+
Most advertisers only test a small cluster of the vast universe of potential ad concepts. Creative coverage means systematically exploring the full map.
|
|
318
|
+
|
|
319
|
+
**Battleship approach:** Systematically place ads across awareness levels, segments, concepts, angles, and styles.
|
|
320
|
+
|
|
321
|
+
**Strategy allocation:** High-probability bets (swiped concepts, winner variations) get bulk of spend. Exploration bets (net new concepts, unexplored segments) get smaller allocation but are essential.
|
|
322
|
+
|
|
323
|
+
---
|
|
324
|
+
|
|
325
|
+
## Writing Phase -- Copy
|
|
326
|
+
|
|
327
|
+
### The Ten Ads in Ninety Minutes Workflow
|
|
328
|
+
|
|
329
|
+
Luke's core production system:
|
|
330
|
+
|
|
331
|
+
**Production Targets:**
|
|
332
|
+
| Pace | Volume |
|
|
333
|
+
|------|--------|
|
|
334
|
+
| Entry | 10 ads/week |
|
|
335
|
+
| Intermediate | 20-30 ads/week |
|
|
336
|
+
| Advanced | 50 ads/week |
|
|
337
|
+
| Elite | 100 ads/week (20/day, Mon-Fri) |
|
|
338
|
+
|
|
339
|
+
**Three-Phase Process:**
|
|
340
|
+
|
|
341
|
+
**Phase 1: Analysis**
|
|
342
|
+
1. Review top ads (organize by spend, read comments, immerse)
|
|
343
|
+
2. Review emerging ads
|
|
344
|
+
3. Save winners as new reference material
|
|
345
|
+
|
|
346
|
+
**Phase 2: Ideation (STORMING)**
|
|
347
|
+
Generate 10 concepts across sources. Distribute across creative diversity buckets.
|
|
348
|
+
|
|
349
|
+
**Phase 3: Writing**
|
|
350
|
+
Three starting resources drive the process:
|
|
351
|
+
1. **Starter Prompt** -- Feed winning ads + product info. Get 10-12 drafts split by awareness level.
|
|
352
|
+
2. **Hook Prompt** -- Generate 10-12 winning hooks. Use Hook Bot.
|
|
353
|
+
3. **Headline Prompt** -- Generate 10-12 winning headlines.
|
|
354
|
+
|
|
355
|
+
**Step-by-step:**
|
|
356
|
+
1. Run MarioBot or InFeed VSL Bot for initial drafts
|
|
357
|
+
2. Apply Starter Prompt with winning ad examples
|
|
358
|
+
3. Generate hooks with Hook Prompt
|
|
359
|
+
4. Write body copy
|
|
360
|
+
5. Write headlines
|
|
361
|
+
|
|
362
|
+
### Unaware Ads
|
|
363
|
+
|
|
364
|
+
Highest-risk, highest-reward. Target people who don't know they have a problem.
|
|
365
|
+
|
|
366
|
+
**Core principle:** The audience is doomscrolling. They want entertainment, not a pitch.
|
|
367
|
+
|
|
368
|
+
**Structure:**
|
|
369
|
+
1. Hook with pure curiosity/entertainment/emotion
|
|
370
|
+
2. Gradually make the story relevant
|
|
371
|
+
3. Transition from story to problem awareness
|
|
372
|
+
4. Present product as natural solution
|
|
373
|
+
|
|
374
|
+
**"You cannot possibly make it too long. You can just make it too boring."**
|
|
375
|
+
|
|
376
|
+
**Three types of unaware ads:**
|
|
377
|
+
1. **Straddle Ads** -- Between unaware and problem-aware. Points out a symptom and says it's a big problem.
|
|
378
|
+
2. **Symptom Ads** -- Takes a symptom they haven't connected to the product category.
|
|
379
|
+
3. **Worldview Porn** -- Validates or challenges what they already believe.
|
|
380
|
+
|
|
381
|
+
**The critical question:** "What do these people believe?" You can't write unaware ads without understanding the audience's existing belief system.
|
|
382
|
+
|
|
383
|
+
### Narrative Ads
|
|
384
|
+
|
|
385
|
+
Long-form story-driven ads that bypass skepticism through narrative structure.
|
|
386
|
+
|
|
387
|
+
**Key characteristics:**
|
|
388
|
+
- Product does NOT appear until late in the story
|
|
389
|
+
- Catalyst (opening scene) has nothing to do with the product
|
|
390
|
+
- Mechanism is revealed through analogy, not clinical explanation
|
|
391
|
+
- Proof is delivered through moments (specific human scenes), not metrics
|
|
392
|
+
- Product arrives as resolution, not interruption
|
|
393
|
+
|
|
394
|
+
### Copy Blocks
|
|
395
|
+
|
|
396
|
+
The five modular persuasion units that can be assembled in any format:
|
|
397
|
+
|
|
398
|
+
1. **Pain** -- Identify with the prospect's problem
|
|
399
|
+
2. **Promise** -- Clear outcome the product delivers
|
|
400
|
+
3. **Proof** -- Evidence (testimonials, data, endorsements)
|
|
401
|
+
4. **Curiosity/Mechanism** -- Why this works differently
|
|
402
|
+
5. **Constraints** -- Remove objections (ease, safety, time, cost)
|
|
403
|
+
|
|
404
|
+
Strongest copy blocks appear earliest, especially above the fold. Don't bury persuasive points.
|
|
405
|
+
|
|
406
|
+
**Depth — P3C2.** The full v2 treatment (the Pain Chain, the Promise Ladder, the 5 proof categories + Siamese-Twin/Balance-Scale rules, A.W.E. for dissolving constraints, the Curiosity Quadrant, the bonus Conditions block, the **Persuasion Equation**, and CRAVES for strengthening any block) is in `copy-blocks.md`. That's the vocabulary for *steering* a draft: name the missing or weak block instead of saying "make it better."
|
|
407
|
+
|
|
408
|
+
### Belief Shifting
|
|
409
|
+
|
|
410
|
+
Marketing = identifying needed beliefs, identifying current beliefs, moving from one to the other.
|
|
411
|
+
|
|
412
|
+
**The 7 Critical Beliefs:**
|
|
413
|
+
1. Problem beliefs (the problem is real and caused by X)
|
|
414
|
+
2. Solution beliefs (this type of solution works)
|
|
415
|
+
3. Brand/vendor beliefs (trustworthy, cares, unique advantage)
|
|
416
|
+
4. Now or never (urgency)
|
|
417
|
+
5. Risk reversal (safe to try)
|
|
418
|
+
6. Capability (I can do this)
|
|
419
|
+
7. Worthiness (I deserve this)
|
|
420
|
+
|
|
421
|
+
**Four ways to shift a belief:**
|
|
422
|
+
1. **Install** a new belief -- Use metaphors, visual pictures. "If you can make a mental picture of it, people cannot help but believe it."
|
|
423
|
+
2. **Uninstall** an old belief -- Show counterexamples that break the existing model.
|
|
424
|
+
3. **Accommodate** an existing belief -- "Yes, and also..." Use when belief is partially correct.
|
|
425
|
+
4. **Reframe** a belief -- Redirect emotional energy and rewrite the meaning.
|
|
426
|
+
|
|
427
|
+
### Proof Techniques
|
|
428
|
+
|
|
429
|
+
**The more outlandish the claim, the higher up in the ad the proof must sit.**
|
|
430
|
+
|
|
431
|
+
| Claim Strength | Proof Placement |
|
|
432
|
+
|----------------|----------------|
|
|
433
|
+
| Conventional | Later in the ad |
|
|
434
|
+
| Moderate | First third |
|
|
435
|
+
| Outlandish | Immediately after hook |
|
|
436
|
+
| Seemingly impossible | IN the hook itself ("A new Harvard study...") |
|
|
437
|
+
|
|
438
|
+
**Proof framing matters:**
|
|
439
|
+
- "Nearly half of all men" > "40%" (alarming vs. clinical)
|
|
440
|
+
- "724% more effective" > "47% more belly fat loss" (the Mario Math Upgrade: divide 51 by 7)
|
|
441
|
+
- Trace researcher backgrounds for credibility chain ("Ivy League-trained researchers" > "University of Scranton")
|
|
442
|
+
- Never drop a testimonial without context setup
|
|
443
|
+
|
|
444
|
+
**"Don't Say Nots" principle:** "I'm not making this up" makes people think you're making it up. Replace negation with positive proof.
|
|
445
|
+
|
|
446
|
+
### Tangible Curiosity
|
|
447
|
+
|
|
448
|
+
Creates curiosity through concrete, specific, testable claims:
|
|
449
|
+
|
|
450
|
+
1. **Paradoxical Questions** -- Contradict common sense, reveal insight
|
|
451
|
+
2. **Diagnostic Questions / Quizzes** -- Self-assessment tools
|
|
452
|
+
3. **Tips and Hacks** -- Specific, actionable micro-advice
|
|
453
|
+
4. **Characterizations** -- Nicknames for problem/solution ("Sensory Confusion", "Metabolic Gridlock")
|
|
454
|
+
5. **Testable Proof** -- Observable phenomena audience can verify
|
|
455
|
+
6. **Visual Quiz Triggers** -- Image-based questions
|
|
456
|
+
7. **Metaphors** -- Concrete analogies for complex ideas
|
|
457
|
+
|
|
458
|
+
### Insight and Epiphany Creation
|
|
459
|
+
|
|
460
|
+
Genuine insights trigger dopamine, create trust, and motivate continued consumption.
|
|
461
|
+
|
|
462
|
+
**How to generate insights:**
|
|
463
|
+
1. From the mechanism itself (should be the biggest insight)
|
|
464
|
+
2. From leverage points (small thing, disproportionate results)
|
|
465
|
+
3. From paradoxical questions
|
|
466
|
+
4. From insight vectors (systematically shift beliefs)
|
|
467
|
+
5. From missing beat map entries
|
|
468
|
+
|
|
469
|
+
### Beat Map Analysis
|
|
470
|
+
|
|
471
|
+
A beat map breaks down all narrative points in a VSL, sales page, or sales process. 14 beats, each answering a subconscious question:
|
|
472
|
+
|
|
473
|
+
1. Speaker Background/Credibility -- "Who are you?"
|
|
474
|
+
2. Problem Narrative -- "Do you understand me?"
|
|
475
|
+
3. Failed Solutions -- "Why hasn't anything worked?"
|
|
476
|
+
4. Myths and Mistakes -- "What beliefs are holding me back?"
|
|
477
|
+
5. UMP Preview -- "Why were YOU able to figure this out?"
|
|
478
|
+
6. UMP Cause -- "What's REALLY causing this?"
|
|
479
|
+
7. UMP Elaboration -- "Why doesn't everyone have this?"
|
|
480
|
+
8. Self-Reflection -- "Does this apply to ME?"
|
|
481
|
+
9. UMS Preview -- "How did you discover the solution?"
|
|
482
|
+
10. UMS Characterization -- "How is this different?"
|
|
483
|
+
11. Tips and Secrets -- "Can you give me value right now?"
|
|
484
|
+
12. Creating the Solution -- "How do I know this is credible?"
|
|
485
|
+
13. Momentum -- "Why should I act NOW?"
|
|
486
|
+
14. Risk -- "What if it doesn't work?"
|
|
487
|
+
|
|
488
|
+
**Key application:** Any individual beat can be turned into an ad. Missing beats represent untapped ad angles.
|
|
489
|
+
|
|
490
|
+
---
|
|
491
|
+
|
|
492
|
+
## Writing Phase -- Hooks
|
|
493
|
+
|
|
494
|
+
### The "Vicious" Standard
|
|
495
|
+
|
|
496
|
+
Luke's single-word quality bar. If a hook is not vicious, it is not ready.
|
|
497
|
+
|
|
498
|
+
- Must have urgency, stakes, consequences -- not just "something happens"
|
|
499
|
+
- Must grab at an emotional or existential level
|
|
500
|
+
- The energy bar: "If you don't look at this ice cube, you're dying tomorrow. This ice cube is gonna cure your cancer."
|
|
501
|
+
- If you read your hook and feel comfortable, it is not vicious enough
|
|
502
|
+
- **Test:** Strip to weakest interpretation. If it reads "something happens when you do something," it fails.
|
|
503
|
+
|
|
504
|
+
### Hook Quality Checklist
|
|
505
|
+
|
|
506
|
+
1. Most interesting part goes FIRST -- If you can name the ad in 2 words, those words must be in line 1
|
|
507
|
+
2. Contains 3 copy blocks -- Curiosity (opens a loop) + Pain (connects to fear) + Promise (hints at gain). Minimum: curiosity + one of pain or promise.
|
|
508
|
+
3. Consumable -- Instantly understandable. Short sentences. Simple words.
|
|
509
|
+
4. Connected to relevance fast -- Within 1-2 lines, they understand why this matters to THEM
|
|
510
|
+
5. Avoids pain-led default -- "You used to walk for hours without pain" is what every competitor writes. Prefer curiosity-led hooks.
|
|
511
|
+
|
|
512
|
+
### 10 Hook Types for Unaware Ads
|
|
513
|
+
|
|
514
|
+
1. **Paradox** -- Two contradictory things are both true
|
|
515
|
+
2. **Lost Wisdom / Forgotten Remedy** -- The answer already exists, we forgot it
|
|
516
|
+
3. **Conspiracy / Historical Inflection** -- Something was changed/hidden at a specific point
|
|
517
|
+
4. **Warning Sign** -- Reframes something normalized as something serious
|
|
518
|
+
5. **Everyday Threat** -- Something trusted is secretly harming you
|
|
519
|
+
6. **Story** -- Narrative creates identification and curiosity
|
|
520
|
+
7. **Big Win** -- Transformation achieved, pure aspiration
|
|
521
|
+
8. **Big Horror** -- Activates deepest dread
|
|
522
|
+
9. **Symptom Call-Out** -- Identifies observable behavior without explaining
|
|
523
|
+
10. **Sacred Cow / The Lie** -- A fundamental belief they've held is false
|
|
524
|
+
|
|
525
|
+
### Pro Tip Triggers That Always Work
|
|
526
|
+
|
|
527
|
+
- **Authority & Institution**: Harvard, NASA, Stanford, Mayo Clinic, peer-reviewed, new study
|
|
528
|
+
- **Exotic & Tribal**: Hadza, Okinawans, Amish, Inuit, Maasai, Japanese centenarians
|
|
529
|
+
- **Historical & Ancestral**: Ancient Egyptians, Victorian doctors, your grandmother
|
|
530
|
+
- **Elite Performers**: Navy SEALs, Olympic athletes, astronauts, biohackers
|
|
531
|
+
- **Blue Zones**: Okinawa, Sardinia, Nicoya Peninsula, Ikaria
|
|
532
|
+
|
|
533
|
+
### Headlines vs. Hooks
|
|
534
|
+
|
|
535
|
+
Headlines are separate from hooks. They sit next to the CTA button. Headlines should densely weave: Pain + Promise + Proof + Curiosity + Constraints. Strong capitalization, punctuation, numbers. Maximum specificity.
|
|
536
|
+
|
|
537
|
+
### The Hourglass Principle
|
|
538
|
+
|
|
539
|
+
As soon as you get attention, an invisible hourglass starts draining. Hook them and get to the point fast. Two steps: hook their attention, then sink your teeth in.
|
|
540
|
+
|
|
541
|
+
### Hook Writing Process
|
|
542
|
+
|
|
543
|
+
1. Collect 10-20 winning hooks as examples
|
|
544
|
+
2. Create a master prompt with: information context, examples, guidelines, the specific ad
|
|
545
|
+
3. Have AI generate 20 hooks, then select the best
|
|
546
|
+
4. Write hooks AFTER the ad body (Mario's insight) -- have AI identify the most interesting statement and frontload it
|
|
547
|
+
|
|
548
|
+
---
|
|
549
|
+
|
|
550
|
+
## Editing Phase
|
|
551
|
+
|
|
552
|
+
> Editing is a first-class stage. The named steer-prompts (Natural Language / Shorten / Cut /
|
|
553
|
+
> Simplify / Make Better — the Genesis per-variant menu, plus bounce / transitions / logic /
|
|
554
|
+
> CTA), the core sentence rules, the full anti-AI list, the CTA formula, and the
|
|
555
|
+
> **feedback-bakes-back** loop (corrections → primer rules) all live in `editing-rules.md`.
|
|
556
|
+
|
|
557
|
+
### Three Editing Passes
|
|
558
|
+
|
|
559
|
+
Every piece of AI-generated copy gets three specific edits:
|
|
560
|
+
|
|
561
|
+
**1. Lower Reading Level (Caveman Copy)**
|
|
562
|
+
- Replace Latin-root words with Germanic equivalents
|
|
563
|
+
- "Fabrics" > "Cloth", "Excessive" > "Too much", "Perspire" > "Sweat"
|
|
564
|
+
- "The French perspire. The Germans sweat. You want to use German words."
|
|
565
|
+
- Keep technical terms accurate, but simplify everything else
|
|
566
|
+
|
|
567
|
+
**2. Natural Language**
|
|
568
|
+
- Add connective tissue, warmth, conversational flow
|
|
569
|
+
- AI copy often reads choppy and robotic
|
|
570
|
+
- Make it sound like how a real person talks
|
|
571
|
+
|
|
572
|
+
**3. Anti-AI / Staccato Fix**
|
|
573
|
+
- Fix the choppy fragment problem AI creates
|
|
574
|
+
- Break triplet patterns (AI loves sets of three)
|
|
575
|
+
- Remove AI tells:
|
|
576
|
+
- The number 47 (AI's favorite number)
|
|
577
|
+
- Excessive em-dashes
|
|
578
|
+
- Pain-first opens (every AI leads with pain)
|
|
579
|
+
- Generic "I know what it feels like" openings
|
|
580
|
+
- Overly dramatic emotional language
|
|
581
|
+
|
|
582
|
+
---
|
|
583
|
+
|
|
584
|
+
## Image Phase -- SCRAWLS
|
|
585
|
+
|
|
586
|
+
> Full detail in `references/scrawls-pipeline-reference.md`
|
|
587
|
+
|
|
588
|
+
SCRAWLS is Luke's framework for static image concept ideation, specifically for **unaware and problem-aware** audiences.
|
|
589
|
+
|
|
590
|
+
### The 7 Sources
|
|
591
|
+
|
|
592
|
+
| Letter | Source | Description |
|
|
593
|
+
|--------|--------|-------------|
|
|
594
|
+
| S | Swipe Mining | Find what's spending money in paid ads |
|
|
595
|
+
| C | Copy-Derived | Feed finished copy into concept generation. Push twice -- first pass is always literal. |
|
|
596
|
+
| R | Reptile Triggers | Run through 13 primal psychological triggers |
|
|
597
|
+
| A | Audience Language | Mine real customer language for visual scenes |
|
|
598
|
+
| W | Wild Sourcing | Find real native photos from Reddit, Google Images, Facebook groups |
|
|
599
|
+
| L | Loopback | Feed winners back in, extract vectors, expand |
|
|
600
|
+
| S | Source | Your own brain and creative intuition |
|
|
601
|
+
|
|
602
|
+
### 13 Reptile Triggers
|
|
603
|
+
|
|
604
|
+
1. Ultra-Real, 2. Bizarre, 3. Voyeur, 4. Suffering, 5. Gorey, 6. Sexual, 7. Primal Fear, 8. Odd Contrast, 9. Inside Joke, 10. Time Warp, 11. Victory Lap, 12. Selfie, 13. Uncanny Objects
|
|
605
|
+
|
|
606
|
+
**Top 5 most reliable:** Bizarre, Ultra-Real, Gorey, Sexual/Hot, Niche-Specific
|
|
607
|
+
|
|
608
|
+
### 3-Step Creation Workflow
|
|
609
|
+
|
|
610
|
+
1. Generate ideas via SCRAWLS (all 7 sources)
|
|
611
|
+
2. Pull starter/reference images
|
|
612
|
+
3. Generate using AI (Nano Banana Pro, ChatGPT Image Gen, Flux)
|
|
613
|
+
- Path A: Reverse image prompt from reference
|
|
614
|
+
- Path B: Create new prompt from concept
|
|
615
|
+
|
|
616
|
+
### Awareness-Level Routing
|
|
617
|
+
|
|
618
|
+
- **Unaware / Problem Aware** --> SCRAWLS process
|
|
619
|
+
- **Solution Aware / Product Aware** --> Standard static ad types (see next section)
|
|
620
|
+
|
|
621
|
+
Auto-detect awareness level. Don't make the user choose.
|
|
622
|
+
|
|
623
|
+
---
|
|
624
|
+
|
|
625
|
+
## Image Phase -- Standard Types
|
|
626
|
+
|
|
627
|
+
For solution-aware and product-aware ads, use format-specific generation from the 21+ static ad types:
|
|
628
|
+
|
|
629
|
+
1. Headline + Image
|
|
630
|
+
2. Side by Side / Before and After
|
|
631
|
+
3. Infographic
|
|
632
|
+
4. Product Breakdown
|
|
633
|
+
5. Animation
|
|
634
|
+
6. Comparison
|
|
635
|
+
7. Scientific Study
|
|
636
|
+
8. Collage
|
|
637
|
+
9. Holding Sign
|
|
638
|
+
10. Native News
|
|
639
|
+
11. Note From The Founder
|
|
640
|
+
12. Testimonial
|
|
641
|
+
13. Lofi / Ugly Ads
|
|
642
|
+
14. Bold
|
|
643
|
+
15. Meme
|
|
644
|
+
16. Hero
|
|
645
|
+
17. Comment/Review
|
|
646
|
+
18. Handwritten Note
|
|
647
|
+
19. Screenshot
|
|
648
|
+
20. Breaking News
|
|
649
|
+
21. Carousel
|
|
650
|
+
|
|
651
|
+
### 5 Static Ad Workflows
|
|
652
|
+
|
|
653
|
+
All follow the pattern: Example Reference + Inputs (Messaging + Brand/Design) = Ad
|
|
654
|
+
|
|
655
|
+
1. **Creative OS Templates** -- Browse templates, generate messaging, edit directly
|
|
656
|
+
2. **Foreplay Inspired** -- Find winning ads, recreate in Canva, swap messaging
|
|
657
|
+
3. **Swipe Based** -- Start with swipe image, describe, re-prompt, generate new
|
|
658
|
+
4. **Research Based** -- Generate ideas from messaging inputs, create prompts, generate images
|
|
659
|
+
5. **Style/Format Based** -- Choose visual style, generate messaging, create images
|
|
660
|
+
|
|
661
|
+
### Visual Mechanisms
|
|
662
|
+
|
|
663
|
+
AI-generated images showing how a product's mechanism works inside the body or in the world.
|
|
664
|
+
|
|
665
|
+
**Top 10 Visual Transformation Use Cases:**
|
|
666
|
+
1. Exploded Product Reveal
|
|
667
|
+
2. Before/After Morph
|
|
668
|
+
3. X-Ray Transparency Fade
|
|
669
|
+
4. Data Overlay Scan
|
|
670
|
+
5. Microscopic Zoom Journey
|
|
671
|
+
6. Time-Lapse Transformation
|
|
672
|
+
7. Blueprint to Reality
|
|
673
|
+
8. Layer Peel Revelation
|
|
674
|
+
9. Network Propagation Map
|
|
675
|
+
10. Problem Particle Elimination
|
|
676
|
+
|
|
677
|
+
**Key principle:** People think in pictures, not scientific accuracy. A "medically inaccurate" diagram showing inflammation being drained is more persuasive than a technically correct one.
|
|
678
|
+
|
|
679
|
+
### Vector Expansion
|
|
680
|
+
|
|
681
|
+
Method for generating new variations from winning images:
|
|
682
|
+
|
|
683
|
+
1. Collect what's working
|
|
684
|
+
2. Analyze with brain + AI: "Why is this working?"
|
|
685
|
+
3. Isolate vectors (visual style, subject, composition, lighting, emotion, trigger)
|
|
686
|
+
4. Expand each vector independently
|
|
687
|
+
|
|
688
|
+
Example: "bloodshot eye" --> other bloodshot eyes, yellow eyes --> ultra close-up of other body parts --> eye bags, wrinkles --> foggy bathroom mirror (environmental version)
|
|
689
|
+
|
|
690
|
+
---
|
|
691
|
+
|
|
692
|
+
## Video Phase
|
|
693
|
+
|
|
694
|
+
> Before scripting a video, decode the format with **CAST** (Conceit · Actor · Style ·
|
|
695
|
+
> Terrain) — the generative grammar for video ads. New Conceit or Style = novelty (resets the
|
|
696
|
+
> sophistication clock); swap Actor or Terrain = diversity (scales a winner). Conceit primers
|
|
697
|
+
> layer on top of payload primers to manufacture novelty with no new research. Full grammar in
|
|
698
|
+
> `cast-video.md`. (Video *rendering* is roadmap in the base CLI; CAST is the thinking layer
|
|
699
|
+
> you brief with today.)
|
|
700
|
+
|
|
701
|
+
### 9-Step AI Video Workflow
|
|
702
|
+
|
|
703
|
+
1. **Get Script** -- 1-1.5 minute script from ad copy
|
|
704
|
+
2. **Create Character Reference** -- Consistent character using reference bot
|
|
705
|
+
3. **Paste Into Image Model** -- Nano Banana, GPT, Flux for character images
|
|
706
|
+
4. **Create Video Brief** -- Structured scenes, transitions, pacing
|
|
707
|
+
5. **Create Starting Images** -- Reference image + video brief into generation
|
|
708
|
+
6. **Create Video Animation Prompt** -- Video brief into animation instructions
|
|
709
|
+
7. **Animate Videos** -- Kling 3.0, Runway
|
|
710
|
+
8. **Create Music** -- AI soundtrack matching tone
|
|
711
|
+
9. **Create VO** -- AI voiceover (ElevenLabs v3)
|
|
712
|
+
|
|
713
|
+
**Format cross-testing:** A video ad that bombs can crush as a static. Always test across formats before declaring an idea dead.
|
|
714
|
+
|
|
715
|
+
**State-level personalization:** Take one winning ad, swap only the state name using AI voice cloning, run 21 versions. "Cheat code."
|
|
716
|
+
|
|
717
|
+
---
|
|
718
|
+
|
|
719
|
+
## Funnel Phase -- Advertorials and Landing Pages
|
|
720
|
+
|
|
721
|
+
### Advertorial Structure
|
|
722
|
+
|
|
723
|
+
1. **Authority Headline** -- Doctor/expert persona, number-based promise, audience callout
|
|
724
|
+
2. **Skeptic-to-Believer Opening** -- Credibility through initial skepticism
|
|
725
|
+
3. **Proof Stack** -- Each point: clear claim + supporting data + relatable context
|
|
726
|
+
4. **Mechanism Explanation** -- Pain Points > Intermediary Mechanisms > Core Mechanism
|
|
727
|
+
5. **Problem-Solution Bridge** -- Why the obvious alternative doesn't work (3 objections)
|
|
728
|
+
6. **Product Introduction** -- Only AFTER full mechanism education
|
|
729
|
+
7. **Timeline of Results** -- Night 1-3, Week 1, Week 2-4, Month 2+
|
|
730
|
+
8. **Testimonials + Offer** -- Named testimonials with ages + guarantee
|
|
731
|
+
|
|
732
|
+
### 10 Advertorial Angles
|
|
733
|
+
|
|
734
|
+
Grouped into five categories:
|
|
735
|
+
- **Discovery-Based**: Breaking News/Discovery, Origin Story
|
|
736
|
+
- **Credibility-Based**: Authority/Expert, David vs Goliath
|
|
737
|
+
- **Threat-Based**: Expose/Scandal, Warning/Fear
|
|
738
|
+
- **Proof-Based**: Social Proof Frenzy, Scarcity/Exclusivity
|
|
739
|
+
- **Comparison-Based**: Comparison
|
|
740
|
+
- **Relatability-Based**: Local/Relatable
|
|
741
|
+
|
|
742
|
+
Real advertorials layer multiple angles. Pick primary for headline/hook, layer secondary in body.
|
|
743
|
+
|
|
744
|
+
### Mario's "Stupidly Simple" Formula
|
|
745
|
+
|
|
746
|
+
1. Rising trend/craze
|
|
747
|
+
2. Delivering "amazing" results
|
|
748
|
+
3. For exclusive group
|
|
749
|
+
4. Crazy demand, sells out fast, almost impossible to get, until now
|
|
750
|
+
|
|
751
|
+
### Copy Blocks on Landing Pages
|
|
752
|
+
|
|
753
|
+
**Five core copy blocks** (Pain, Promise, Proof, Curiosity, Constraints) used across all formats. Above-the-fold sections should be curated with strongest blocks first.
|
|
754
|
+
|
|
755
|
+
### Performance Benchmarks
|
|
756
|
+
|
|
757
|
+
- 30-40% CTR from advertorial to e-commerce page is good
|
|
758
|
+
- Double advertorial strategy (advertorial > proof advertorial > sales page) boosts conversions significantly
|
|
759
|
+
|
|
760
|
+
---
|
|
761
|
+
|
|
762
|
+
## Testing and Iteration
|
|
763
|
+
|
|
764
|
+
### Core Testing Philosophy
|
|
765
|
+
|
|
766
|
+
**"Double down on winners, don't chase losers."** When a basic ad works on the right pain points, the market is telling you what it wants. Build on that signal.
|
|
767
|
+
|
|
768
|
+
### Hit Rates
|
|
769
|
+
|
|
770
|
+
- Expect 80-90% of ads to fail
|
|
771
|
+
- "You can get very rich off a 10% hit rate"
|
|
772
|
+
- Peter Kell tests 40 VSLs to find one winner
|
|
773
|
+
- Hit rates haven't changed in 15 years despite AI
|
|
774
|
+
|
|
775
|
+
### Testing Splits (Jimmy Rutkowsky -- $1.5M/month spend)
|
|
776
|
+
|
|
777
|
+
- 70% variations on winners
|
|
778
|
+
- 20% improving winning ads
|
|
779
|
+
- 10% new concepts
|
|
780
|
+
- Tests in 6 ad sets of 10 creatives each at $500/day per ad set
|
|
781
|
+
|
|
782
|
+
> The variations/scaling mechanics — the three tracks (net-new / variations / scaling), the
|
|
783
|
+
> **CASHED** knobs, IDENTIFY-first decoding, the high-priority Re-Aim/Re-Concept/Re-Angle/
|
|
784
|
+
> Re-Skin/Re-Hook moves, and the adjustable 70/20/10 split tied to account maturity — are in
|
|
785
|
+
> `iterations.md`.
|
|
786
|
+
|
|
787
|
+
### Angles vs. Variations
|
|
788
|
+
|
|
789
|
+
"Angles open up audiences. Variations improve ad efficiency." -- Chris Alonzo
|
|
790
|
+
|
|
791
|
+
### Hook Stacking
|
|
792
|
+
|
|
793
|
+
From VShred: slap new hooks onto existing winners. Test a new hook on the control. If it wins, it becomes the new control. Repeat.
|
|
794
|
+
|
|
795
|
+
### Hypothesis-Driven Testing
|
|
796
|
+
|
|
797
|
+
For every ad test:
|
|
798
|
+
1. State the hypothesis: "I believe [concept/angle/hook] will work because [reason]"
|
|
799
|
+
2. Run the test
|
|
800
|
+
3. Record the result
|
|
801
|
+
4. Note what you learned
|
|
802
|
+
|
|
803
|
+
### The Loopback Flywheel
|
|
804
|
+
|
|
805
|
+
Feed winners back into the system. Analyze what worked, extract vectors, expand into new territory. Every winner makes the next batch smarter. This is the learning loop that compounds over time.
|
|
806
|
+
|
|
807
|
+
---
|
|
808
|
+
|
|
809
|
+
## Genesis Bot Catalog
|
|
810
|
+
|
|
811
|
+
### Foundation Bots (Run First)
|
|
812
|
+
- **Copy Blocks Extract Bot** -- Extracts copy blocks from any sales page
|
|
813
|
+
- **Build-A-Buyer Bot** -- Full market research report
|
|
814
|
+
- **Offer Brief Bot** -- Compiles offer/product information
|
|
815
|
+
- **Segment Surgeon Bot** -- Breaks market into segments
|
|
816
|
+
- **Master Concept Bot** -- 300+ ad concepts (run once)
|
|
817
|
+
|
|
818
|
+
### Ad Writing Bots
|
|
819
|
+
- **MarioBot** -- Narrative-driven unaware ads
|
|
820
|
+
- **InFeed VSL Bot** -- In-feed video sales letter scripts
|
|
821
|
+
- **Video Ad Script Bot** -- Universal ad writer
|
|
822
|
+
- **75+ Ad Template Bot** -- Fast output from proven structures
|
|
823
|
+
- **Unaware Ads Bot** -- Wild unaware ad ideas
|
|
824
|
+
- **MicroVSL Bot** -- Short-form 3-6 minute VSL scripts
|
|
825
|
+
|
|
826
|
+
### Hook and Variation Bots
|
|
827
|
+
- **Hook Bot** -- Generates 10-12 winning hooks
|
|
828
|
+
- **Short Form Hook Bot** -- Hooks for short-form content
|
|
829
|
+
- **Insight Vectors Bot** -- "Pound for pound the most valuable bot"
|
|
830
|
+
- **Paradoxical Questions Bot** -- Paradoxical question hooks
|
|
831
|
+
- **Metaphors Bot** -- Powerful metaphors
|
|
832
|
+
- **Characterizations Bot** -- Nicknames for concepts
|
|
833
|
+
- **Surprising Culprit Bot** -- Villain/culprit ideas
|
|
834
|
+
- **Testable Proof Bot** -- Self-tests and verifiable proof
|
|
835
|
+
|
|
836
|
+
### Iteration Bots
|
|
837
|
+
- **CASH Analysis/Variation Bot** -- Generates CASH variations
|
|
838
|
+
- **CASH Rewriter Bot** -- Rewrites ads using variations
|
|
839
|
+
- **Segment/Mech Swapper Bot** -- Swaps mechanism or segment
|
|
840
|
+
- **Ad Swiping Bot** -- Swipes paid ads
|
|
841
|
+
- **Organic Swiping Bot** -- Swipes organic posts
|
|
842
|
+
|
|
843
|
+
### Static Image Bots
|
|
844
|
+
- **Universal Static Ad Idea Bot** -- Ideas from copy (no specific direction needed)
|
|
845
|
+
- **Static Ad Swiper Bot** -- Swipes existing static ads
|
|
846
|
+
- **21+ Static Ad Type Bots** -- One per format
|
|
847
|
+
- **Static Ad Image Generator** -- Creates final images (1:1 and 9:16)
|
|
848
|
+
- **13 Reptile Triggers Bot** -- Reptile-trigger-enhanced prompts
|
|
849
|
+
- **Winner Analysis Prompt Bot** -- Extracts vectors from winners
|
|
850
|
+
- **Vector Expansion Prompt Bot** -- Generates expanded concepts
|
|
851
|
+
|
|
852
|
+
### Funnel Bots
|
|
853
|
+
- **Advertorial Bot** -- All sub-types
|
|
854
|
+
- **Quiz Bot** -- Quiz funnels
|
|
855
|
+
- **Listicle Bot** -- Tips, reasons why, warning signs
|
|
856
|
+
- **Caveman Page Bot** -- Caveman-style pages
|
|
857
|
+
- **Proof Page Bot** -- Proof-heavy landing pages
|
|
858
|
+
- **Bridge Page Bot** -- Jump pages
|
|
859
|
+
- **Hybrid PDP Bot** -- ecom product pages
|
|
860
|
+
- **Upsell/Downsell Bots**
|
|
861
|
+
- **Proof Compiler Bot** -- Analyzes proof gaps
|
|
862
|
+
- **Proof Deep Research Bot** -- Finds scientific studies
|
|
863
|
+
|
|
864
|
+
### Research Bots
|
|
865
|
+
- **Universal Mechanism Bot** -- Creates mechanisms across any market
|
|
866
|
+
- **Microscript Bot** -- Naming and characterization
|
|
867
|
+
- **Market Resonance Bot** -- Market dynamics analysis
|
|
868
|
+
- **AI Money Matrix Bot** -- Monetization path analysis
|
|
869
|
+
- **Belief Analyst Bot** -- Maps current beliefs
|
|
870
|
+
- **Belief Alchemist Bot** -- Strategies for shifting beliefs
|
|
871
|
+
- **Auto-Brief Bot** -- Creates beat maps from existing copy
|
|
872
|
+
|
|
873
|
+
---
|
|
874
|
+
|
|
875
|
+
## End-to-End Workflow
|
|
876
|
+
|
|
877
|
+
The full creative strategy workflow from zero to published ads:
|
|
878
|
+
|
|
879
|
+
### Phase 0: Onboarding (Once Per Client)
|
|
880
|
+
1. Run Build-A-Buyer Bot with product/market info
|
|
881
|
+
2. Run Offer Brief Bot
|
|
882
|
+
3. Run Copy Blocks Extract Bot on existing sales pages
|
|
883
|
+
4. Run Segment Surgeon Bot
|
|
884
|
+
5. Run Master Concept Bot (300+ concepts)
|
|
885
|
+
6. Compile foundation documents
|
|
886
|
+
|
|
887
|
+
### Phase 1: Analysis (~30 min)
|
|
888
|
+
1. Review top-performing ads (organize by spend, read comments)
|
|
889
|
+
2. Review emerging ads
|
|
890
|
+
3. Save winners as reference material
|
|
891
|
+
4. Run comment mining on top ads
|
|
892
|
+
5. Identify coverage gaps in the creative diversity matrix
|
|
893
|
+
|
|
894
|
+
### Phase 2: Ideation (~30 min)
|
|
895
|
+
1. Run through STORMING sources to generate 10 concepts
|
|
896
|
+
2. Distribute across awareness levels (6 unaware/problem, 4 solution/product)
|
|
897
|
+
3. For each concept, define CASH components
|
|
898
|
+
4. For unaware concepts, identify which of the 3 unaware types it is
|
|
899
|
+
|
|
900
|
+
### Phase 3: Copy Writing (~60 min)
|
|
901
|
+
1. Run MarioBot or InFeed VSL Bot for initial drafts
|
|
902
|
+
2. Apply Starter Prompt with winning ad examples
|
|
903
|
+
3. Generate hooks with Hook Bot (10-12 per ad)
|
|
904
|
+
4. Write body copy
|
|
905
|
+
5. Generate headlines (10-12 per ad)
|
|
906
|
+
|
|
907
|
+
### Phase 4: Copy Editing (~30 min)
|
|
908
|
+
1. Lower reading level (Caveman Copy pass)
|
|
909
|
+
2. Natural language pass
|
|
910
|
+
3. Anti-AI / Staccato fix pass
|
|
911
|
+
|
|
912
|
+
### Phase 5: Image Creation
|
|
913
|
+
1. Auto-detect awareness level for each ad
|
|
914
|
+
2. Route unaware/problem-aware through SCRAWLS pipeline
|
|
915
|
+
3. Route solution/product-aware through standard type selection
|
|
916
|
+
4. Generate image concepts
|
|
917
|
+
5. Find/select reference images
|
|
918
|
+
6. Generate final images with Nano Banana Pro / Gemini
|
|
919
|
+
|
|
920
|
+
### Phase 6: Assembly and Delivery
|
|
921
|
+
1. Compile: 2 hooks + 2 headlines + 1 body copy + 6-8 static images per concept
|
|
922
|
+
2. Format in Google Doc
|
|
923
|
+
3. Queue for publishing (via Adzara or direct to Meta)
|
|
924
|
+
|
|
925
|
+
### Phase 7: Testing and Loopback
|
|
926
|
+
1. Deploy ads with hypothesis tags
|
|
927
|
+
2. Monitor performance
|
|
928
|
+
3. Identify winners
|
|
929
|
+
4. Extract vectors from winners
|
|
930
|
+
5. Feed back into Phase 2 (the flywheel)
|
|
931
|
+
|
|
932
|
+
---
|
|
933
|
+
|
|
934
|
+
## Key Principles and Rules
|
|
935
|
+
|
|
936
|
+
### For the Agent
|
|
937
|
+
|
|
938
|
+
1. **Always swipe first.** Before generating anything novel, find what's already working in the market. Swiping is the highest-probability approach.
|
|
939
|
+
|
|
940
|
+
2. **Auto-detect awareness level.** Every ad idea gets classified. Route to the right pipeline. Don't ask the user to choose.
|
|
941
|
+
|
|
942
|
+
3. **Hooks are everything.** If the hook fails, nothing else matters. Apply the "vicious" standard. Strip to weakest interpretation to test.
|
|
943
|
+
|
|
944
|
+
4. **Push past the literal.** First-pass concepts (especially images) are always boring. Always push for a second, more unexpected pass.
|
|
945
|
+
|
|
946
|
+
5. **Caveman copy.** Germanic words, not Latin. Simple, instant understanding. If a caveman couldn't grunt-approve, simplify.
|
|
947
|
+
|
|
948
|
+
6. **Proof placement follows claim strength.** Outlandish claims need proof immediately. Conventional claims can prove later.
|
|
949
|
+
|
|
950
|
+
7. **Don't lead with pain.** AI defaults to pain-first opens. Prefer curiosity-led hooks. "You used to walk for hours without pain" is what every competitor writes.
|
|
951
|
+
|
|
952
|
+
8. **Creative diversity is mandatory.** Cover segments x awareness levels. Don't cluster. The algorithm rewards diversity.
|
|
953
|
+
|
|
954
|
+
9. **Modular swiping, not wholesale.** Break ads into CASH DNA. Swipe individual components. Create novel combinations.
|
|
955
|
+
|
|
956
|
+
10. **The mechanism is the product's story.** Not features. Not benefits. The named reason why this works when everything else failed.
|
|
957
|
+
|
|
958
|
+
11. **Test hypotheses, not gut feelings.** State what you believe, why, run the test, record the result.
|
|
959
|
+
|
|
960
|
+
12. **Volume wins.** 80-90% of ads fail. Hit rates haven't changed. The way to get more winners is to produce more ads.
|
|
961
|
+
|
|
962
|
+
13. **Loopback is the flywheel.** Winners feed the next batch. Every cycle gets smarter.
|
|
963
|
+
|
|
964
|
+
14. **Verify all bot output.** Genesis bots fabricate claims, statistics, and studies. Always fact-check before publishing.
|
|
965
|
+
|
|
966
|
+
15. **Less is more with AI prompts.** Minimal inputs. Let the bot have confidence. Don't dump entire foundation docs.
|
|
967
|
+
|
|
968
|
+
---
|
|
969
|
+
|
|
970
|
+
## Sources
|
|
971
|
+
|
|
972
|
+
All content synthesized from the Creative Strategy Knowledge Base at `domain-knowledge/creative-strategy/`:
|
|
973
|
+
|
|
974
|
+
**Wiki Pages (120+ pages across concepts, entities, topics):**
|
|
975
|
+
- Core frameworks: CASH Method, BAD Method, ABCD Framework, Alignment Model, Five Functions, STORMING, SCRAWLS
|
|
976
|
+
- Mechanism: Unique Mechanism, Universal Mechanism Framework, Mechanism Chain, Three Master Narratives, Mechanism Naming, SIN Test
|
|
977
|
+
- Copy: Hook Formulas, Hook Types, Caveman Copy, Copy Blocks, Belief Shifting, Proof Modalities, Proof Techniques, Tangible Curiosity, Beat Map Analysis, Unaware Ads, Narrative Ads, Comment Mining
|
|
978
|
+
- Images: Reptile Triggers, Vector Expansion, Static Ad Types, Static Ad Workflows, Visual Mechanisms, SCRAWLS Framework
|
|
979
|
+
- Strategy: Creative Coverage, Creative Diversity, Swiping Strategy, Modular Swiping, Ad Concept Generation, Ad Testing Philosophy, Creative Fatigue and Volume
|
|
980
|
+
- Workshops: Creative Strategist Ascension (3-day), Infinite AI Ads Machine (4-day)
|
|
981
|
+
- Entities: Luke Mills, Mario Castelli, MarioBot, Hook Bot, Genesis, Nano Banana Pro
|
|
982
|
+
|
|
983
|
+
**Raw Sources:**
|
|
984
|
+
- 43 Weekly Monday Gospel transcripts (May 2025 - March 2026)
|
|
985
|
+
- Infinite AI Ads Machine Workshop (4 days)
|
|
986
|
+
- Creative Strategist Ascension Workshop (3 days + implementation call)
|
|
987
|
+
|
|
988
|
+
**Meeting Notes:**
|
|
989
|
+
- Brad x Luke meetings (March-April 2026)
|
|
990
|
+
- AI Agent team meetings (March-April 2026)
|