10 Day AI Bootcamp V2.0 Summary

Topic 01 — Introduction Module

This topic served primarily as the introduction and orientation for the X100 AI Bootcamp. It covered the overall structure of the program, expectations, workflow, and upcoming modules. No transcript was successfully fetched for this session.

Topic 02 — Foundation: Level 10 Meetings & Scorecard

This session focused on implementing EOS-style operational systems using Level 10 meetings, scorecards, and quarterly Rocks. Members learned how to structure weekly accountability meetings, track meaningful KPIs, and align marketing execution with measurable business goals.

Topic 03 — Elite AI Strategy Sprint

A deep strategy workshop covering ICP creation, niche positioning, Don Draper-style messaging, epiphany-bridge storytelling, and lead magnet ideation using ChatGPT prompts. The goal was to create the strategic foundation that powers future AI-generated content and campaigns.

Topic 04 — Custom GPT Builder Workshop

This workshop taught members how to build specialized custom GPTs using ICP documents, core messaging, founder stories, and knowledge files. The focus was on creating role-specific AI assistants for social media, email copywriting, compliance, and research workflows.

Topic 05 — Goal & Rock Setting with X100 GPT

A planning and accountability session showing how to use the X100 Business Growth GPT to create SMART quarterly Rocks, daily "one thing" metrics, and operational alignment using EOS, Scaling Up, and The One Thing frameworks.

Topic 06 — Social Media, Email & Blog Content

This module focused on AI-assisted content production using expert-style prompts inspired by Gary Vaynerchuk, Joanna Wiebe, BuzzFeed, and David Ogilvy. Members learned workflows for creating social posts, blogs, email campaigns, and optimized social profiles.

Topic 07 — Image Creation

A hands-on image production workshop using Midjourney, Nano Banana, Flux, Canva, and Photoshop to create commercial-quality marketing visuals, product imagery, lifestyle ads, and branded social media assets.

Topic 08 — Video Creation Part 1: Veo 3 Commercials

This session focused on creating cinematic AI commercials using Veo 3, Google Flow, ChatGPT-generated JSON prompts, and professional editing workflows. Members learned how to rapidly produce TV-style marketing videos at a fraction of traditional production costs.

Topic 09 — Video Creation Part 2: Talking Head & Avatar Videos

A workshop covering AI talking-head videos using HeyGen, ElevenLabs, avatars, voice cloning, and AI-generated scripts. The session focused on scalable spokesperson-style content creation for marketing, sales, and education.

Topic 10 — AI Distribution Systems

This module explored strategies for distributing AI-generated content across multiple platforms, automating publishing workflows, and repurposing content into different formats for broader reach and engagement.

Topic 11 — AI Ads & Paid Media

A paid-media workshop focused on AI-assisted advertising workflows including Meta ads, audience targeting, hooks, creative generation, and campaign optimization using AI-generated messaging and visuals.

Topic 12 — SEO & Organic Traffic

This session covered AI-enhanced SEO workflows including keyword strategy, blog generation, metadata optimization, schema markup, and long-form content systems designed to improve organic traffic growth.

Topic 13 — Automation & AI Workflows

A systems-focused workshop introducing automation tools like Make.com, n8n, AI agents, and workflow integrations to reduce repetitive operational tasks and improve marketing execution efficiency.

Topic 14 — Scaling Content Operations

This module discussed how to scale AI-driven content systems across multiple clients or brands through SOPs, delegation, batching, workflow management, and standardized production pipelines.

Topic 15 — Final Optimization & Growth Systems

The final session tied together strategy, production, automation, distribution, and optimization into a repeatable AI-driven marketing ecosystem designed for long-term scaling and operational consistency.

Meeting Anatomy

Stuck-points round with attendees → live Meta-ads diagnostic for one member (broken out into a separate Google Meet so it didn't derail the room) → screen-share walk-through of the Level 10 document → walk-through of the sample Scorecard → assignment recap and Q&A on personal content and the upcoming four-pillars prompt.

Summary/Recap

Scott framed the X100 program as four pillars — strategy, content production, distribution, optimization — glued together by EOS as a management system. The two artifacts that operationalize EOS for a marketing team are the Level 10 weekly meeting and the Scorecard. Without those two, the monthly coaching call has no data to react to.

The Level 10 is a 45-minute fixed-cadence agenda: three minutes of Segue, employee/client Headlines, a five-minute Rocks review (on-track / off-track only — off-track items drop straight into IDS), a quick Scorecard pass, prior-week To-Dos, then the bulk of the meeting on IDS (Identify, Discuss, Solve) attacking the number-one bottleneck, then a 0–10 closing rating on whether the team started on time, ended on time, and discussed the right things.

The Scorecard runs on a 12-week cycle with 3–5 KPIs max, each row being owner + measurable + baseline + weekly goal. Scott emphasized that "fewer measured frequently is far better than lots not measured at all" and demonstrated the X100 GPT that produces 3–5 quarterly Rocks from six diagnostic questions.

Tools

• Level 10 Document (Google Doc template) — the weekly operating-system agenda. Provided in every member's calendar invite.
• Scorecard (Google Sheet template) — 12-week KPI tracker with owner/measurable/baseline/goal columns.
• X100 Business Growth GPT — generates 3–5 SMART quarterly Rocks via six diagnostic questions; takes business name, product, audience as input.
• Four Pillars Prompt — upcoming prompt that takes ICP + core messaging + epiphany-bridge backstory and outputs the four content pillars a founder should post against.
• EOS / Traction — the underlying management framework the Level 10 is borrowed from.

Steps Discussed

1. Open your Level 10 document each week — it lives in the recurring calendar invite.
   1. Resources section at the top — drop in links to training videos, media libraries, social channels you reference repeatedly.
   2. Segue (≈3 min) — personal and professional best from the week.
   3. Headlines — quick wins/losses: client won, client lost, new bid, sales last week. One-line each, no detail.
   4. Rocks review (≈5 min) — ask each Rock owner "on track or off track?" If on track, move on. If off track, drop it into IDS — do NOT discuss in this section.
   5. Scorecard — same on-track/off-track scan; off-track KPIs drop to IDS.
   6. To-Dos — review prior-week action items.
   7. IDS (Identify, Discuss, Solve) — ≈30% or more of the meeting; attack the number-one bottleneck until it's gone.
   8. Closing rating — everyone scores the meeting 0–10 on start time, end time, and whether the most important things were discussed.
2. Build your Scorecard for the 12-week cycle:
   1. Pick 3–5 KPIs total. Start smaller, not larger.
   2. For each KPI capture owner + measurable + baseline + weekly goal.
   3. Use leading indicators (impressions, video views, engagement, profile visits, sends/shares for Instagram; media spend, CPM, CTR, qualified leads, CPA for paid; delivery rate / open rate for email) — not just sales.
   4. Share the scorecard with the founder/client to get buy-in on the targets before week 1.
3. Use the X100 Business Growth GPT to set your Q1 Rocks:
   1. Open the GPT, copy and paste the prompt from the resource doc.
   2. Answer the prompts for name, business name, website.
   3. Answer the six diagnostic questions about your 60- / 90-day vision.
   4. Copy the 3–5 Rocks it generates into the Rocks section of your Level 10.
4. Schedule the cadence: daily 10–15 minute huddle to report one number, weekly Level 10 to run the agenda, monthly one-on-one with X100 to review.

Resources

• Sample Level 10 Document (make a copy)
• YouTube video on Level 10 Document
• Sample Scorecard (make a copy)
• YouTube video on Scorecard

Key Takeaways

• Off-track Rocks belong in IDS, not in the Rocks review. Protect the meeting rhythm — the Rocks pass is a status check, not a problem-solving session.
• 3–5 KPIs measured weekly beat 15 KPIs measured never. "Fewer measured frequently is far better than lots not measured at all."
• ~30% of your marketing meeting should attack one bottleneck — the number-one jugular issue, until it's off the sheet for good.
• Baseline before goal. Without a starting number, you can't tell whether marketing is doubling or flatlining.
• Meta optimizes for whatever you ask for. Buying profile visits gives you profile visits at the expense of video views, and vice versa — be explicit about the objective. (From the live ads diagnostic during stuck-points.)
• Don't dilute the dollar across ad sets. $80 in one ad set is more powerful than $50 + $30 in two — combine unless you have a specific reason to split.
• Four pillars beat one-pillar posting. A founder posting against four distinct content pillars attracts four audience types and can move from one post/day to three reels/day with bigger reach.

Questions to Explore Further

• How should the 3–5 Scorecard KPIs change between a brand-awareness phase (impressions, video views) and a lead-gen phase (MQLs, CPA) within the same 12-week cycle?
• When a Rock is off track for two weeks in a row inside IDS, what's the trigger to either re-scope it or kill it rather than keep grinding?
• For solo marketers like Kathy, what's the minimum-viable Level 10 — can it be a self-run 15-minute version, and does the closing rating still work with one person?
• How do you keep stakeholders aligned on the Scorecard when the metrics that matter shift mid-quarter (e.g., switching from social impressions to email after discovering the funnel doesn't convert)?
• What's the right way to roll the X100 GPT's Rock output past the founder for review/approval without losing the SMART structure?

Steps I Need to Do on My End

1. Open the resource links and make a copy of both the sample Level 10 document and the sample Scorecard for your team.
2. Spend ~30 minutes customizing the tools: fill in the baselines, set the 3–5 KPIs with owner / measurable / baseline / weekly goal.
3. Share the Scorecard with your client / founder / stakeholder to get buy-in on the targets for the next 12 weeks.
4. Open the X100 Business Growth GPT and run it to generate your Q1 Rocks, then drop them into the Rocks section of your Level 10.
5. Schedule the weekly Level 10 cadence (and ideally a 10–15 min daily huddle) so the agenda actually gets used between monthly X100 calls.

Meeting Anatomy

Course-roadmap overview (8–10 week structure: strategy → production → distribution → optimization) → live walk-through of the 7-step prompt stack in ChatGPT, with Brad (TMG, capital gains elimination) as the live volunteer → Don Draper core messaging exercise → introduction to the Epiphany Bridge with the 31-question dictate-mode script → round-robin: members read aloud their AI-generated 30-second backstory pitches → preview of next week's lead-magnet workshop and announcement of guest Dave Burnett for AI SEO.

Summary/Recap

Scott introduced the X100 "10-Day AI Bootcamp 2.0" — a structured re-org of the program into Strategy → Content Production → Distribution → Optimization. This session was the Strategy Sprint: a 7-step prompt stack run in ChatGPT (or Gemini) to produce an ICP, niche, top-10 pains/objections, bumper stickers, an epiphany-bridge backstory, and a Don Draper core-messaging system. Each step's output is saved to a Google Doc so it can later be uploaded as knowledge files to a custom GPT.

The Don Draper prompt outputs an umbrella statement (positioning tagline) plus three core messages — cold (problem-aware), warm (solution-aware), and hot (product-aware) — and the top objections with answers. Scott's worked example: instead of car ads pushing "$1,500 cash back, 0% for 60 months" (hot only), a cold ad like "6 questions to ask before you buy your next truck" creates positioning. Most marketers send hot messaging to cold audiences and wonder why nothing converts.

The Epiphany Bridge prompt asks ~31 questions in ChatGPT's dictate mode and outputs the founder's (or a client's) origin story in 30-second, 2-minute, and 5-minute versions — usable as the backbone of LinkedIn posts, reels, podcasts, and ad creative. Scott framed the session's deeper thesis using Pat Quinn (who coaches Tony Robbins and Grant Cardone): education builds awareness, but solving problems builds trust, and high-ticket buyers only convert on trust. Lead magnets must solve an acute, specific pain — not be a "solution looking for a problem." The framework for evaluating lead-magnet ideas: (Dream Outcome × Perceived Likelihood) / (Time Delay × Effort) — Alex Hormozi's value equation.

Tools

• ChatGPT — primary tool. Create a new Project named "Ideal Customer Persona, Niche, and Top 10 Questions" and run the 7-step prompts inside it.
• ChatGPT Dictate mode — required for the Epiphany Bridge (you talk through ~31 questions; transcription becomes source material).
• Gemini — mentioned as alternative LLM.
• 7-Step Prompts Google Sheet — the source-of-truth prompt library; copy each prompt cell without expanding it (Sheets blocks copy on expand).
• Google Docs — save each prompt's output here so it can later be uploaded as a Word doc to a custom GPT.

Steps Discussed

1. In ChatGPT, create a new Project called "Ideal Customer Persona, Niche, and Top 10 Questions."
2. Run the Persona Prompt. Inputs: business name, product/service, target audience (one phrase), revenue size/title if B2B, top 3 problems they're struggling with, website URL. Don't aim for perfect — you'll refine.
3. Run the Niche Prompt in the same thread. It generates 8 sub-niches. Pick 2–3 max (one geographic, one demographic, one needs-based is a useful frame).
4. Run the Refined Persona — re-run with the chosen sub-niche, assign a gimmicky nickname so the persona sticks.
5. Run the Top 10s — objections, fears, logical questions, emotional/financial motivations, limiting beliefs.
6. Run the Bumper Stickers prompt — punchy one-liners usable as rant scripts on reels.
7. Run the Epiphany Bridge:
   1. Open the script (~31 questions).
   2. Turn on ChatGPT dictate mode — talk through each question aloud (it'll sound like brain damage; that's fine).
   3. ChatGPT produces 30-second, 2-minute, and 5-minute versions of the story.
   4. Save all three versions to a Google Doc.
8. Run the Don Draper Core Messaging prompt with these inputs:
   1. 3 adjectives describing brand personality (interview the founder for these).
   2. Buzzwords to include or avoid (e.g., automotive "Tier 1/2/3" signals industry membership).
   3. Evidence/proof points — attach case studies, social proof, billion-dollar credentials, market-share stats.
   4. Output: umbrella statement + 3 core messages (cold/warm/hot) + top objections with answers.
9. After each step, save the output to a Google Doc so the source material is preserved for the upcoming custom-GPT workshop.

Resources

• 7 Step Prompts
• ChatGPT
• Gemini

Key Takeaways

• Trust beats education for high-ticket sales. Educating endlessly doesn't move buyers; solving a specific problem does. (Pat Quinn frame.)
• Cold, warm, hot need three different messages. Most teams write hot-audience copy and run it against a cold list — that's why it doesn't convert.
• Save everything as a Google Doc (then Word) as you go. These docs become the knowledge files for your custom GPT next session.
• Niche down hard for the next 90 days. Pick one sub-niche to aggressively target; be willing to say no to business outside it.
• The Epiphany Bridge is a one-time investment with huge reuse. Once it's in ChatGPT, every piece of content can pull from it — reels, blogs, podcasts, emails, ads.
• Lead magnets must solve an acute pain, not a generic one. Hormozi's value equation: maximize dream outcome and perceived likelihood, minimize time and effort.
• The $1000 guarantee. Scott offered $1,000 to anyone who didn't get value from a strategy call — calendar filled for 4 days. Demonstrates the trust mechanic; the bribe is optional, the principle isn't.
• Don't expand cells in the prompts sheet. Click into the cell without expanding to copy — Google Sheets blocks the copy otherwise.

Questions to Explore Further

• How do you sequence the cold/warm/hot core messages across channels — should LinkedIn lead with cold, retargeting carry warm, and email/SMS deliver hot?
• For businesses with 3+ distinct sub-niches, what's the smartest way to fork the persona/messaging artifacts without duplicating the entire 7-step run for each?
• What's the right cadence for re-running the Epiphany Bridge — is it a one-and-done, or do you re-record it annually as the founder's story evolves?
• How well does the Don Draper umbrella statement hold up as a literal homepage headline, vs. needing one more pass of human editing for rhythm?
• When the Hormozi value-equation prompt outputs 15 lead-magnet ideas, what's the scoring heuristic for picking the one to build first?

Steps I Need to Do on My End

1. Open the 7 Step Prompts sheet and create a ChatGPT Project named "Ideal Customer Persona, Niche, and Top 10 Questions."
2. Run the persona, niche, refined-persona, top-10, and bumper-stickers prompts in order, saving each output to a Google Doc as you go.
3. Set aside 30 minutes to redo the Don Draper core-messaging prompt if you don't love the first output — iterate until the umbrella statement and cold/warm/hot messages feel right, then save the final to a Google Doc.
4. If you have multiple personas or sub-niches, repeat the persona → niche → top-10 process for each one this week (explicit assignment for members with multiple niches).
5. Practice telling your 30-second epiphany-bridge story aloud — the next time someone asks what you do, lead with the story, not the job title.

Meeting Anatomy

Stuck-points round on personal-pillar content (Bill, Michael, Brad sharing how personal posts outperform pure-business posts) → handoff from Scott to Rishab/Isha for the technical walk-through → live build of a "Scott Empringham Social Media Copywriter" GPT, editing the creation prompt field-by-field in a Google Doc → side-by-side demo comparing generic ChatGPT output vs. the custom GPT output on the same prompt → ChatGPT GPT Builder configure-tab walk-through (knowledge files, model, capabilities) → assignment recap.

Summary/Recap

Rishab and Isha taught how to build a custom GPT that "bridges Strategy and Production" — once your ICP, core messaging, top 10s, and epiphany bridge are nailed down, a custom GPT can write content in your voice at scale by always referencing those files before every generation.

The key architectural point: build one GPT per use case, not per persona. If you have three personas (Carlos, Brenda, Brandon), they all live as separate knowledge files inside the same social-media-copywriter GPT — the user specifies which persona at prompt time. Build separate GPTs for separate jobs (social media copy, email copy, compliance check, research) because each one becomes a specialist. Custom GPTs differ from ChatGPT Projects: a custom GPT has an invisible system prompt that always runs knowledge through the ICP/messaging files; a Project may pull from related chats but doesn't guarantee that pass-through.

Upload knowledge files as Word (.docx) or text, not PDF — Word parses more reliably than PDF, especially at volume. Each strategy artifact (ICP, core messaging, top 10s, epiphany bridge, lead magnets) goes in as a separate document. Other useful uploads: podcast transcripts, top-performing posts, case studies, brand guidelines, testimonials. The recommended model is regular GPT-5 (not GPT-5 Thinking, which is slower and tuned for code/math). For social media generation: enable web search, canvas, image generation; disable code interpreter and data analysis.

Tools

• ChatGPT (paid) — required for the GPTs feature (left sidebar → "GPTs").
• CustomGPT Creation Prompt — the X100 template document used to scaffold the bot instructions before pasting into the configure tab.
• Google Docs / Microsoft Word — used to draft the GPT prompt and to author the knowledge files. Download each Doc as .docx before uploading.
• GPT-5 (the regular variant) — recommended model for content GPTs.
• Consensus GPT — mentioned as an example of a public GPT used for client research.

Steps Discussed

1. Have your strategy artifacts ready as separate Google Docs: ICP, core messaging, top 10s, epiphany bridge, lead magnets. Each saved separately. If you have multiple personas, each is its own doc.
2. Open the X100 CustomGPT Creation Prompt and paste it into a fresh Google Doc to edit visually.
3. Fill in the prompt fields:
   1. Bot name — name it after the task, e.g. "X100 Social Media Copywriter."
   2. Purpose / use case — e.g. "Write social media copy in the style of [founder] for Instagram / LinkedIn."
   3. Target audience — who will use the bot (your social team), NOT the ICP. (Unless the bot is customer-facing for support/sales, in which case it is the ICP.)
   4. Primary functions and tasks — e.g. "Write social copy in [founder]'s voice; shorter sentences, single-line stacked, caps and emojis on Meta but not LinkedIn."
   5. Personality and tone — 3+ adjectives (e.g. "bold, aggressive, optimistic, energetic"); optionally cite a writer's style (Hormozi, Gary V, Malcolm Gladwell).
   6. Behaviors and restrictions — what NOT to do. E.g. "Do not preach — use experience shares (I did X, got Y)." Forbid politics or divisive topics to avoid shadow-banning.
   7. Privacy/security constraints — industry-specific terms to avoid (healthcare, FinTech "investment" wording for crypto, etc.).
   8. Analyze/summarize documents — set to: "Before every generation, the bot should analyze the ICP, core messaging, top 10s, epiphany bridge."
   9. Remove sections not needed (automations/integrations, image generation, code generation) for a content-only bot.
4. Paste the edited prompt into a fresh ChatGPT thread and hit enter — ChatGPT generates the refined instruction block.
5. In a new browser tab, open ChatGPT → left sidebar → GPTs → top-right Create button → switch to the Configure tab.
6. Fill in the Configure tab:
   1. Name — paste the bot name.
   2. Description — short summary (e.g. "Writes social media copy in the voice of [founder]").
   3. Instructions — paste the entire refined prompt block from the previous step (between the bot-name and final-output lines).
   4. Conversation starters — skip for a first build.
   5. Knowledge — upload the .docx versions of ICP, core messaging, top 10s, epiphany bridge, lead magnets, plus any case studies, brand guidelines, top-performing posts, podcast transcripts. Each artifact a separate file.
   6. Recommended Model — GPT-5 (regular, not Thinking).
   7. Capabilities — enable Web Search, Canvas, Image Generation. Disable Code Interpreter and Data Analysis.
   8. Privacy — uncheck "use conversation data for training."
   9. Visibility — "Anyone with link" so you can share it with your team.
7. Publish the GPT. To iterate: never use the right-side preview pane for feedback loops (changes there wipe the chat) — open the GPT from the left sidebar in a real chat thread and refine there so history is saved.
8. Share the link with your team so multiple operators can run the same specialist.

Resources

• CustomGPT Creation Prompt

Key Takeaways

• One GPT per use case, not per persona. Personas are knowledge files; the GPT is a job specialist.
• Word over PDF for knowledge files. Better parse reliability, especially as your knowledge base grows.
• Custom GPT ≠ ChatGPT Project. A custom GPT runs an invisible system prompt that always references the ICP/messaging before generating; Projects only sometimes do.
• Target audience field = who uses the bot, not who buys from you. Internal bot → your team. Customer-facing bot → your ICP.
• Don't edit in the preview pane for real iterations. Open the published GPT in a real chat thread so feedback loops survive.
• Stay away from politics / divisive content on the founder page. Once shadow-banned, you may as well burn the page. Create a separate page if you must.
• GPTs can become unintentional lead magnets. The X100 GPT was meant for internal alignment; sharing it organically spread the program.
• Personal-pillar content out-performs business-only content. Bill's "tired but family-first" post got 50 shares in 20 minutes despite no business content.

Questions to Explore Further

• What's the right cadence to refresh knowledge files inside a custom GPT — every quarter, or whenever the ICP/messaging changes materially?
• For agencies running custom GPTs for many clients, what's the cleanest way to maintain a shared "framework GPT" alongside per-client GPTs without drift?
• When using a custom GPT as a customer-facing sales assistant with the actions API, how do you keep the conversation from going off-script while still capturing form data?
• Does upper-casing rules in the instructions actually survive across long content runs, or does the model regress to mixed case after a few generations?
• If you've trained a GPT in one founder's voice and the founder wants to test a guest writer's voice (e.g., Malcolm Gladwell), do you fork the GPT or stack styles in one?

Steps I Need to Do on My End

1. Build at least two custom GPTs: one for content research, one for content creation (e.g., social media copywriter or email copywriter). Take action — this is the explicit assignment.
2. Optional / overachiever: build additional specialist GPTs (email copywriter, blog writer, compliance checker) for each remaining content surface.
3. Save each strategy artifact (ICP, core messaging, top 10s, epiphany bridge, lead magnets) as a separate Word doc and upload them into the GPT's Knowledge section.
4. Founders/CEOs (and anyone willing): post about something outside your business this week — a personal pillar, something that scares you or moves you. Tag Scott when you post.

Meeting Anatomy

Stuck-points round (HeyGen voice cloning, Veo 3 exercise-form prompting via JSON, Nano Banana for UGC) → Covey rocks/pebbles/sand metaphor as setup → live walk-through of the X100 Business Growth GPT with Scott narrating each of the 6 diagnostic questions → demonstration of the "One Thing" daily-metric feature inside the GPT → check-in with multiple members (Diego, Chris Fuller, Whitney, Matt Lowry, Olivia, Kevin) on whether they've used it and what worked.

Summary/Recap

The problem this session solves: founders/CMOs have a vision; the execution team has its own version; on monthly momentum calls those visions don't match. The fix is a 30–40 minute exercise with the X100 Business Growth GPT that locks the whole team into 3–5 SMART quarterly rocks plus a single daily "one thing" per owner. The GPT has Gino Wickman's Traction, Verne Harnish's Scaling Up, and Gary Keller's The One Thing already loaded as knowledge files, so it operates like having an EOS coach on your shoulder for the cost of zero.

Scott opened with Stephen Covey's rocks-pebbles-sand metaphor (from 7 Habits): if you pour pebbles and sand first, the rocks won't fit. Load the jar with rocks first and everything else fills around them. Translated to marketing: pick 3–5 quarterly rocks, assign one owner each, "define done" with success metrics, and treat anything else as pebbles. Off-track rocks drop into IDS at the weekly Level 10; nothing else.

The "one thing" feature inside the GPT extracts — per owner — the single daily action that compounds toward the rock. Scott's example: 25 LinkedIn DMs/day → 125/week → ~6 new leads/month → 30+ booked speaking engagements in ~40 days. Use it as a binary yes/no in the daily huddle. Chris Hawley reinforced: it's not the action that matters, it's the consistency. Dictate mode is strongly recommended — typing one-word answers gives garbage rocks; speaking 2 minutes per question produces rocks the GPT can actually reason about.

Tools

• X100 Business Growth GPT — preloaded with Traction, Scaling Up, and The One Thing as knowledge files. Asks 6 diagnostic questions and outputs SMART rocks with owners and success metrics.
• Level 10 agenda (downloaded as .docx) — upload to the GPT for context on what you're running.
• Scorecard (downloaded as .xlsx) — upload to the GPT for context on the metrics you're tracking.
• ChatGPT Dictate mode — strongly recommended for answering the 6 diagnostic questions; richer input = better rocks.
• iPhone Voice Memos (or Samsung equivalent) — record team discussions, copy the transcript, paste into the GPT to capture multiple perspectives.

Steps Discussed

1. Download your existing Level 10 agenda as a Word doc and your Scorecard as an Excel file.
2. Open the X100 Business Growth Exercise resource doc, copy the rock-setting prompt, and click the embedded business-growth GPT link.
3. Paste the prompt into the GPT and upload both your Level 10 and Scorecard files.
4. Enter your first name, business name, and website URL.
5. Turn on dictate mode and answer the 6 diagnostic questions in order. Speak 1–3 minutes per question, not one-word answers:
   1. Biggest wins or breakthroughs in the past 30–90 days (from a marketing perspective).
   2. Where you got stuck or slowed down.
   3. What's currently working that you could double down on.
   4. Unfinished projects or launches.
   5. (plus two further diagnostic questions about vision and constraints)
6. For multi-person teams, record team members' answers in Voice Memos (or put your desktop on speaker during a Zoom), then paste the transcript into the GPT so everyone's perspective lands.
7. The GPT proposes Rock #1. For each rock: Accept / Reject / Edit; confirm the owner (don't let Joe own all four); confirm the success metric (the "define done").
8. Repeat for rocks 2–5. Skip the execution-plan step if pressed for time.
9. Ask the GPT: "Give us each our one thing we can do daily." It returns one binary action per owner.
10. Rename the chat thread immediately — the GPT auto-names it "90 Day Rock Refiner" by default, which makes it impossible to find next quarter. Rename to e.g. "Q4 2025 Rocks."
11. Copy the final rocks list into the Rocks section of your Level 10 and add the daily "one thing" metric into the Scorecard. Check on/off-track each week.
12. For annual planning: re-prompt the GPT to convert the quarterly framework into an annual planning version (use ChatGPT's prompt-engineer mode to refine the prompt for that purpose).

Resources

• X100 Business Growth Exercise Prompt and Link

Key Takeaways

• Rocks first, then pebbles and sand. Covey's metaphor — load the 3–5 priorities before letting the urgent-but-trivial fill the jar.
• One thing per owner, binary daily. Did you do it, yes or no? Consistency compounds: 25 DMs/day → 30+ speaking engagements in 40 days.
• Dictate mode beats typing. One-word answers produce generic rocks; 2-minute spoken answers produce rocks the GPT can actually use.
• Rename the chat thread immediately. "90 Day Rock Refiner" auto-name is the single most common reason people lose their previous quarter's output.
• Don't let one person own every rock. Joe owning all four = no accountability. Reassign owners during the Accept/Edit step.
• If you knock out your rocks early, re-run the exercise. Don't wait for the next quarter — Matt Lowry's team finished early and went idle. Re-prompt instead.
• Strategy / execution misalignment is the silent killer. Founder and team usually diverge by week 2; this exercise re-locks them.
• Different team members produce different rocks. Run the exercise once per discipline (ads team, content team, AI team), then merge — blind spots surface.

Questions to Explore Further

• How do you adapt the 6 diagnostic questions when only the marketing function (not sales) has access to revenue data — does the GPT need an explicit "marketing-only perspective" prefix?
• When the GPT proposes a rock you disagree with, is it more effective to Edit it inline or Reject and re-dictate the underlying answer with more detail?
• For the annual-planning variant, what's the right cascade — does the GPT produce 4 quarterly sets at once, or one quarter at a time with a rolling re-run?
• How should the "one thing" metric evolve mid-quarter if the leading indicator stops correlating with the lagging outcome (e.g., DMs flat but bookings still rising)?
• What's the protocol when two team members each "own" overlapping rocks (e.g., content creator and ads team both contributing to lead-gen) — split the rock or share the owner field?

Steps I Need to Do on My End

1. Download your Level 10 (as Word) and Scorecard (as Excel) and run the X100 Business Growth GPT exercise this week — practice live, even if you've done it before. Explicit assignment: "I would encourage you to practice this on your own and do this live."
2. Use dictate mode and give 1–3 minute spoken answers to all 6 diagnostic questions; do not type one-word answers.
3. Lock 3–5 rocks with a clear owner and success metric on each. Don't let one person own everything.
4. Ask the GPT for each owner's daily "one thing" and add it to your daily huddle as a yes/no metric.
5. Rename the chat thread (e.g., "Q4 2025 Rocks") immediately so you can find it next quarter.
6. Bring the resulting rocks list to your next team / founder meeting and get sign-off — or run the exercise with each department and merge.

Meeting Anatomy

Stuck-points round (Michael on Red Note bans, Patrizio on JVAI / n8n vs Make.com automation, Kevin on marrying graphics with messaging, Pato on the discipline problem) → epiphany-bridge follow-up (zero member shares posted publicly so far) → handoff to Rishab/Isha for the production walk-through → live demo of all four expert prompts in sequence (Gary V → Smart 90 → BuzzFeed → Ogilvy Makeover) → assignment handed out with three X100 T-shirts on offer for the first three to complete it.

Summary/Recap

Rishab and Isha taught the X100 "AI assembly line" for content production: Copy → Creative → Publish. Once strategy (ICP, core messaging, top 10s, epiphany bridge, lead magnets) is locked in, four expert prompts on the X100 prompts page do the heavy copy lifting — each one channeling a named operator (Gary Vaynerchuk, Joanna Wiebe, BuzzFeed's listicle formula, David Ogilvy by way of Russell Brunson).

The Gary V social prompt outputs four posts in one run: a Light Bulb (aha-moment) post, a Trending (news-jacking) post that hijacks an in-market trend, a Curiosity-Driven Q&A pulling from your Top 10 objections, and a Contrarian post designed to spark argument. Each post comes with caption, image prompt, image hook, and image subtext as separate fields — because AI image tools don't render text well, you generate the image in Midjourney/Nano Banana then overlay the hook/subtext in Canva.

The Smart 90 email prompt (Joanna Wiebe-style) asks you to pick one of the lead magnets you generated earlier and outputs a 5-touch email campaign with send-day recommendations. Scott's reframe of lead magnets: their real purpose is to solve a problem and build trust — not capture an email. If you can include links to your Google or Yelp reviews, the prompt will pull real testimonials into email #4 instead of fabricating them. The BuzzFeed blog prompt hard-codes three principles a former Buzzfeed VP gave Scott: education + lists + fear/mistakes. Edit the "X mistakes" number to an odd number — 3, 5, or 7 — and you get a full blog, a recommended SEO keyword, a meta description (≤160 chars), JSON-LD schema markup, and a Midjourney image prompt. The Ogilvy Makeover prompt rewrites your LinkedIn/Instagram/Facebook/TikTok bios using Russell Brunson's "helping [ICP] solve [pain] without [effort/sacrifice]" formula, plus a banner image prompt.

Tools

• X100 Prompts Page — the source of all four prompts (Gary V, Smart 90 email, BuzzFeed blog, Ogilvy Makeover).
• ChatGPT — run the prompts in the same chat thread as your ICP/core messaging, OR inside a custom copywriting GPT (preferred for consistent voice).
• Canva — overlay the image hook and subtext on AI-generated images (AI tools render text poorly).
• Midjourney / Google Nano Banana / Gemini AI Studio — paste the prompt's image instructions to generate visuals.
• iPhone Voice Memos / ChatGPT dictate mode — Scott's "Morning Genius" block: dictate ideas all day into a thread so the GPT learns your voice and writes posts in ~minutes instead of hours.

Steps Discussed

1. Open the X100 Prompts Page and confirm you're working in your custom copywriting GPT (or your ICP chat thread).
2. Social media — Gary V prompt:
   1. Search the page for "Gary V" and click the copy button on the prompt.
   2. Paste it into your custom GPT thread with no edits and hit enter.
   3. Receive four posts: Light Bulb, Trending, Curiosity Q&A, Contrarian. Each has a caption + image prompt + image hook + image subtext.
   4. For each post you keep: generate the image in Midjourney/Nano Banana, take it into Canva, overlay the hook and subtext.
3. Emails — Smart 90 (Joanna Wiebe) prompt:
   1. Search for "email" / "Smart 90" and copy that prompt.
   2. Paste it into the same chat thread. It will ask you which lead magnet to anchor the campaign on.
   3. Pick the lead magnet, then optionally paste a link to your Google or Yelp reviews so it can ingest real testimonials.
   4. Output: a 5-touch email campaign with recommended send days/cadence.
4. Blogs — BuzzFeed prompt:
   1. Search for the BuzzFeed blog prompt and copy it.
   2. Before hitting enter, edit the bracketed number to an odd number (3, 5, or 7) and choose "mistakes" over "secrets" (pain avoidance > pleasure gain).
   3. Optionally specify a topic; if you don't, ChatGPT picks one.
   4. Niche the title — e.g. "7 painful branding and marketing mistakes every [niche] makes" — for radically tighter copy.
   5. Output: full blog post, recommended SEO keyword, meta description (≤160 chars), schema markup (JSON-LD), and a Midjourney image prompt.
   6. Paste the schema markup into your CMS, fill the meta description, paste keywords, publish.
5. Bio/profile — Ogilvy Makeover prompt:
   1. Copy the Ogilvy Makeover prompt; specify which platform (LinkedIn first; run separately per platform).
   2. Output: tagline (Russell Brunson formula), bio/about section, optimization tips, banner image prompt, and copy for Instagram/Facebook/TikTok bios.
   3. Update each social profile so a visitor knows within 3 seconds that you solve their problem.

Resources

• X100 Prompts Page

Key Takeaways

• Strategy first or the prompts under-deliver. Without ICP + core messaging in the thread, output drops to generic ChatGPT slop.
• Use a copywriting custom GPT, not a one-size GPT. Same-thread ICP works; a dedicated copywriter GPT works better.
• Contrarian content wins on social. Scott's Ted Nugent example ("vegetarians kill more animals…") shows how arguments compound reach — agreement is forgettable.
• Lead magnets build trust by solving a real problem. A cheap PDF that no one reads is worse than no lead magnet — it actively breaks trust.
• Fear/mistakes outperform "secrets" in blogs. Loss aversion is stronger than gain motivation.
• Odd-numbered lists outperform even. 3 / 5 / 7 — measurable, not theoretical.
• Image text goes on in Canva, not in the AI tool. AI image generators mangle text; produce the image clean, overlay text in Canva using the prompt's hook/subtext fields.
• "Morning Genius" beats batch writing. Dictate thoughts into your GPT thread throughout the day so it learns your voice; writing posts then becomes minutes-of-work, not hours.
• Discipline is the missing ingredient. Pato's observation — the tools are no longer the bottleneck; consistent execution is.

Questions to Explore Further

• What's the right ratio across Gary V's four post types in a 30-day calendar — equal weight, or weight toward Contrarian for awareness and Curiosity Q&A for conversion?
• When the Smart 90 email prompt pulls testimonials from Google/Yelp via link, how do you verify the testimonials are correctly attributed before sending?
• For non-English markets, does the BuzzFeed odd-numbered-list bias still hold, or is it a US/EN-specific cultural artifact?
• If a custom copywriter GPT has been fed too many edits over time and drifts, what's the cleanest reset — refresh knowledge files, rewrite instructions, or fork a new GPT?
• How does the Ogilvy Makeover output need to be edited specifically for Red Note (the Chinese platform Michael is targeting), where Western voice frameworks may not translate?

Steps I Need to Do on My End

1. Build a 30-day social media calendar using the Gary V prompt — run it repeatedly to generate the four post types, aim for at least one post per day.
2. Build a 12-touch email campaign using the Smart 90 (Joanna Wiebe) prompt anchored to your strongest lead magnet.
3. Create 10 different blogs using the BuzzFeed prompt — copy-paste, vary the topic/niche, hit publish.
4. By the next session, drop a link to at least one published post (LinkedIn / Facebook / Instagram / TikTok) AND one published blog into the AI WhatsApp group — first three submissions get an X100 T-shirt.
5. Run the Ogilvy Makeover prompt and refresh at least one social profile (LinkedIn first) so the about/tagline/banner align with your ICP and core messaging.

Meeting Anatomy

Stuck-points round (LinkedIn DM strategy referenced — Scott is out at a speaking engagement, 17 booked in 90 days) → recap of the four marketing pillars and where image creation fits (Production) → the "monkey in a suit" before/after to establish that prompt + framework matters → live build of a John Daly merchandise ad end-to-end: Smart 90 prompt → Midjourney generation → Nano Banana shirt-swap → Flux face/era manipulation → Photoshop cleanup → Canva final assembly → bonus Midjourney image-to-video → Q&A on tools, subscriptions, and use cases (Christie, Matt, Amanda, Dave, Bavesh) → assignment.

Summary/Recap

Rishab taught the X100 image production stack using a real client deliverable: John Daly's golf merchandise had only flat product shots and a "yesterday" deadline. The solution: a four-tool assembly line that produces lifestyle, larger-than-life, and on-model product imagery from nothing but ChatGPT, a product photo, and ~24 hours.

The three secrets: (1) prompts dialed in using the Description + Style + Parameter framework, (2) the AI image arsenal — no all-in-one tool works reliably, you need 3–4 specialists, and (3) combine the tools — generate raw quality in Midjourney, swap products in Nano Banana, manipulate without references in Flux, and finish in Canva. The Smart 90 Midjourney Prompt Generator (on the X100 prompts page) writes the description and style for you; the parameters (e.g., --ar 1:1, --v 7) come from Midjourney's parameter list (kept separate because they change as Midjourney updates).

Nano Banana (Google) is best accessed via ai.studio.google.com — not the Gemini app, which strips the temperature and aspect-ratio controls. Flux runs via Replicate at ~$0.08/generation (no subscription) and is the strongest at manipulating an image without a reference picture (e.g., "make me wear a black suit"). For consistent characters across a campaign — same person, different scenes — train a custom Flux model on 40+ photos of the subject. Photoshop 2025 (or Adobe Express as a cheaper alternative) handles AI artifact removal and aspect-ratio extension via generative fill.

Tools

• Midjourney — best raw image quality; paid only; outputs 4 variations per generation; use Upscale → Subtle to keep the composition while raising resolution.
• Google Nano Banana — best at manipulation with a reference (swap a real product onto a generated person, redress a room, place a client in their own merch). Free via ai.studio.google.com. Do NOT use the Gemini app for this — controls are stripped.
• Flux — best at manipulation without a reference (suit, hair, cap, background swap). Via Replicate. ~$0.08/generation, pay-as-you-go.
• Canva — final assembly. Overlay logos, brand text, brand fonts; pull a template, swap the image, ship.
• Smart 90 Midjourney Prompt Generator (on the X100 prompts page) — turns "I want X" into a full Description + Style prompt for Midjourney/Nano Banana.
• ChatGPT — the prompt-writing layer. Run the Smart 90 generator inside your existing ICP/messaging thread for niche-aligned outputs.
• Photoshop 2025 / Adobe Express (bonus) — generative fill for artifact removal and aspect-ratio extension.
• Runway ML "Act One" (mentioned) — character replacement in video; useful for the exercise-form problem from Topic 5's Q&A.

Steps Discussed

1. Save the source product photo (e.g., right-click "save image" from johnzdailygolf.com) and any existing photo of the founder/spokesperson.
2. Generate the scene prompt:
   1. Open the Smart 90 Midjourney Prompt Generator on the X100 prompts page.
   2. Edit only the "type of image" field — e.g., "male golfer in his 30s, mid-shot, on a golf course, wearing a polo shirt." Specifying the garment type helps the later swap.
   3. ChatGPT returns the full Description + Style prompt and offers 3 alternatives.
3. Generate in Midjourney:
   1. Paste the prompt; Midjourney returns 4 variations.
   2. Pick one and click Upscale → Subtle to raise resolution without changing the composition.
   3. Optionally click Vary (Strong) on a favorite to fan out 4 more variations.
   4. Download the chosen image.
4. Swap in the real product with Nano Banana:
   1. Go to ai.studio.google.com, open Nano Banana, start a fresh chat (don't reuse — uploads get confused).
   2. Upload BOTH the generated lifestyle image AND the real product photo.
   3. Prompt: "Please replace the polo shirt the model is wearing with the white polo shirt attached separately to this prompt. Please retain all the visual information of the white polo shirt that is attached."
   4. Wait ~20s. Iterate if the product detail (text, logo, tag) didn't transfer.
5. Manipulate without a reference using Flux (Replicate):
   1. Upload the subject's existing photo (e.g., a younger photo of John Daly).
   2. Prompt with plain English: "make me wear a black suit" or "make him wear a white polo shirt." $0.08/run.
   3. Repeat until composition is right; use up to ~5 generations to get the desired result.
6. Clean up with Photoshop generative fill: select unwanted tags, logos (e.g., the Gemini watermark), or stray artifacts → click Generate → remove. Use the same flow to extend the canvas vertically/horizontally for different aspect ratios.
7. Assemble in Canva: pick an ad template (search e.g. "fashion t-shirt ad"), drag your final image in, swap in your logo, paste ChatGPT-generated ad copy, export. Repeat with a video template for short-form by dropping in a Midjourney image-to-video clip.
8. Bonus — image-to-video: upload the generated still into Midjourney's video mode (no prompt required) to get a ~5-second animated clip suitable for social.
9. For consistent characters across many scenes: train a custom Flux model on 40+ images of the subject, then prompt the trained model for any new scene.

Resources

• MidJourney
• Google Nano Banana (via Gemini) — Best AI Image Generator
• ChatGPT
• Canva
• X100 Prompts Page

Key Takeaways

• Prompt = Description + Style + Parameter. Description is the subject, Style is the aesthetic (50mm portrait, cinematic lighting, film grain), Parameter is Midjourney-specific (--ar, --v).
• No all-in-one image tool works reliably. Each tool is a specialist; the value is in the chain.
• Use AI Studio, not the Gemini app, for Nano Banana. The app strips controls (temperature, aspect ratio).
• Flux beats Nano Banana for reference-free manipulation. Nano Banana needs an input image to anchor; Flux works from prompt alone.
• Mid-Journey gives the artistic feel; Nano Banana gives the natural feel. Same prompt, two different outputs — pick by use case.
• Text and logos go on in Canva. AI image generators still mangle text and produce off-brand logos.
• Always start a new chat per image session. Cross-contaminated uploads confuse the model.
• Niche tightens the image. Same as with copy — niching the prompt to a specific persona/industry dramatically sharpens output.
• Borrow other people's Midjourney prompts. The community feed shows the prompt behind each image; paste a style you love into ChatGPT and ask it to mimic.

Questions to Explore Further

• For B2B businesses with no visual product (SaaS, services), what's the right Smart 90 prompt shape — abstract concept imagery, founder portraits, or screen-capture overlays?
• When training a custom Flux model on 40+ images of a founder, what's the optimal mix of angles, lighting, and outfit variations to avoid overfitting?
• How do you maintain brand-color consistency across a campaign when Midjourney drifts slightly between generations — Canva post-process, or LoRA training?
• Is Photoshop's generative fill genuinely better than Nano Banana for artifact removal, or is the choice mostly about staying inside an existing workflow?
• For larger-than-life campaigns (lions wearing shirts), what's the failure-mode catalog — when does the chain produce uncanny output and how do you spot it before shipping?

Steps I Need to Do on My End

1. Create 4 images for social media posts using the Smart 90 prompt → Midjourney (or Nano Banana if you don't have Midjourney) → Canva finishing.
2. Create 4 images for an email campaign that pair with your Smart 90 email touches.
3. Create 4 images for your website — hero, section banners, or product/feature shots.
4. Create 4 images for blogs — headers/inline using the prompt that comes out of the BuzzFeed blog generator (Topic 6).
5. Take action — explicit assignment: "please, please, please take what you've learned today, put it into action and create images using this training."

Meeting Anatomy

Quick framing on the four marketing pillars and where video sits (Production) → walk-through of the Steven Spielberg commercial prompt in ChatGPT against the Pacific Hardwood ICP thread → live generation of 10 JSON clips in Google Flow / Veo 3 → fix-broken-clips loop (text artifacts, wrong character framing, "Pacific" logo glitches) → live Premier Pro edit of the final commercial → Q&A on voice realism (Kevin), licensing (Joshua), image-to-video frames feature (Claudia) → assignment.

Summary/Recap

Rishab demonstrated the X100 video commercial pipeline. The thesis: a TV-quality commercial used to cost $250K+ in talent, crew, and post — now it's ~$250/month for Veo 3 Ultra plus a few hours in Premier Pro / CapCut. The X100 Steven Spielberg commercial prompt (on the prompts page) reads your existing ICP/core-messaging thread and outputs 10 individual JSON prompts — one per clip — each describing description, style, camera, lighting, room. Paste each JSON block into Veo 3 (max 5 in queue), generate, and treat the misses as iteration material.

Critical handling: Veo 3 mangles text and logos. The fix is to feed the failing clip back to ChatGPT in the same thread and add a constraint — "I don't want any text," "no operator on the forklift," "do not mention Pacific Hardwood." ChatGPT rewrites the JSON; paste the new version back. For text that still ends up baked in, use Runway ML inpainting to paint over and erase it. Rishab also previewed the Frames-to-Video mode in Flow for cases where you have real photos (warehouse, hotel, product) you want animated — you upload a start frame and an end frame and Veo animates between them.

For talking heads, don't use Veo 3 — voice realism is monotone. Use HeyGen avatars + ElevenLabs voice cloning instead, and reserve Veo 3 for B-roll. Folder hygiene matters: every video project gets a year/month folder with sub-folders for assets/VO3-videos/audio/misc/export. Final commercial assembly happens in Premier Pro: import all clips, trim each Veo 3 clip to its strongest 1–2 seconds, beat-match to music, add transitions sparingly (don't reuse the same one), and export.

Tools

• ChatGPT — runs the Spielberg commercial prompt in the same thread as your ICP/messaging. Also used to fix broken clips via feedback.
• Google Veo 3 via Flow (labs.google/fx/tools/flow) — the only place to use Veo 3. Avoid clone sites. Ultra plan $250/month (currently $125/month for the first 3 months); free tier = 5 clips.
• Adobe Premier Pro — Rishab's editor. CapCut and InShot work the same way.
• Runway ML inpainting — erase text, logos, or unwanted objects from generated clips.
• HeyGen + ElevenLabs — recommended for talking-head clips instead of Veo 3 (better voice realism).

Steps Discussed

1. Open your ICP/messaging chat thread in ChatGPT. Open the X100 prompts page and copy the Spielberg commercial prompt.
2. Paste the prompt into the same chat thread (no edits). ChatGPT outputs 10 JSON clip prompts based on your existing ICP.
3. Open Google Flow → New Project. Paste each JSON block (curly-brace to curly-brace) into Flow one at a time. Queue up to 5 generations in parallel.
4. Review each clip. For broken clips:
   1. Copy the failing JSON.
   2. Paste back into ChatGPT with explicit constraints — "no text," "no logos behind the speaker," "no operator on the forklift," "do not say [brand]."
   3. ChatGPT returns a revised JSON; paste into Flow and re-run.
5. To improve voice realism, ask ChatGPT to add a "tone of voice" layer to the JSON prompt (e.g., warm, conversational, energetic).
6. For B-roll only, accept the Veo output. For talking heads, generate the dialogue in HeyGen (avatar) + ElevenLabs (voice clone) and intercut with Veo B-roll.
7. Use Runway ML inpainting to paint over residual text/logos/objects on clips you otherwise want to keep.
8. Use Frames-to-Video in Flow when you have real photos:
   1. Click the text-to-video selector → switch to Frames-to-Video.
   2. Upload a start frame and (optionally) an end frame.
   3. Veo animates between them with camera-style controls.
9. Download each finished clip via the top-right download button. Select Upscale (no credits cost) to export 1080p instead of 720p.
10. Edit in Premier Pro:
    1. Create a folder structure: YYYY/MM/[project]/assets/{vo3-videos,audio,misc} and /export.
    2. Import all upscaled clips. Drop in order. Trim each clip to its strongest 1–2 seconds.
    3. Beat-match audio cuts to music; punch the audio in slightly before the visual cut so cuts feel sharper.
    4. Add transitions sparingly — don't reuse the same transition across multiple cuts.
    5. Export to the /export folder.

Resources

• Veo 3 (via Google Flow)
• X100 Prompts Page

Key Takeaways

• Run the Spielberg prompt in your ICP thread, not a fresh chat. Without your persona and messaging context, the JSON clips will be generic.
• Treat each clip as a draft. Expect ~50% misses on the first pass; iterate broken clips by feeding the JSON back to ChatGPT with explicit constraints.
• Veo 3 can't render text or logos reliably. Either explicitly prompt against them, or clean up in Runway inpainting after.
• B-roll = Veo 3. Talking heads = HeyGen + ElevenLabs. Don't fight Veo's monotone voice limitation.
• Always upscale before downloading. Upscale is free in credits and bumps you from 720p to 1080p.
• Folder hygiene compounds. Year/month/project structure pays back the first time a client asks "where's that clip from June?"
• Frames-to-Video unlocks real assets. When you have actual product/warehouse/hotel photos, animate between two frames instead of fully generating.
• Google's guardrails protect against likeness misuse. Veo won't generate identifiable real people without an explicit reference image.

Questions to Explore Further

• What's the minimum number of generated clips you need to assemble a watchable 30-second commercial — is 10 over-served for short-form, and 5 enough?
• When iterating broken clips, does it work better to add new constraints or to fork a fresh JSON block — does Veo "remember" the failed attempt within a chat?
• How do you sequence Veo B-roll + HeyGen talking heads + ElevenLabs voiceover so cuts feel motivated rather than tool-driven?
• For brands with strict visual guidelines (specific color, font, logo placement), is the Veo → Runway inpainting → Canva overlay pipeline reliable enough for paid media, or do you still need traditional production?
• What's the right Veo Ultra usage cadence — is a single $250 month enough to ship one commercial a week, or do you blow through credits at scale?

Steps I Need to Do on My End

1. Create three video commercials by next week — one for your cold audience, one for your warm audience, one for your hot audience. (Three = rockstar; one = still awesome.)
2. Run the Spielberg commercial prompt inside your existing ICP/messaging chat thread so the clips inherit your persona and core messaging.
3. For broken clips, feed the JSON back to ChatGPT with explicit constraints ("no text," "no logos") rather than re-rolling blindly in Veo.
4. Edit the final commercials in Premier Pro / CapCut / InShot, beat-match to music, and export to a clean /export folder.

Meeting Anatomy

Stuck-points round (Claudia on Meta Advantage Plus and being early to platform features for cheaper CPMs, Kevin on stop-motion / highlight reels via CapCut/InShot/InVideo) → handoff to the two-archetype demo: AI Influencer videos in Captions AI → AI Clone videos in HeyGen → live build of both end-to-end (rant-script prompt → Captions AI prompt-to-video → AI Edit → HeyGen avatar creation → ElevenLabs voice swap) → Q&A on backgrounds, multi-character clones, language localization (Daniel, Dave, Mike, Joshua) → assignment.

Summary/Recap

Two archetypes were taught. AI Influencer videos use Captions AI's pre-built avatars — you generate a "rant script" in ChatGPT (using the X100 Rant Script Generator prompt run inside your ICP thread), pick an avatar in Captions AI (use the iOS app for the most current features), paste the script, and Captions generates a finished video. Then run AI Edit to add B-roll, animations, captions, and image overlays automatically — what would be hours in Premier Pro happens in seconds. You can swap any AI-generated image inline via the timeline.

AI Clones live in HeyGen. The setup: shoot one 2–3 minute sample video (45-second minimum) of yourself speaking on camera — the content of the sample doesn't matter (Scott literally read song lyrics for two minutes), only that you look at camera continuously, no hand-in-front-of-face, and one language. Quality of clone matches quality of sample: if you're stiff in the sample, the clone is stiff. Speak with 40% more energy than feels natural — the camera flattens energy. You can clone different "looks" — outdoor / studio / suited / casual — by uploading multiple separate sample videos (each becomes its own avatar). Dave's question surfaced a great hack: existing professional footage (e.g., from a course shoot) can be used as the sample, which replicates that lighting and setting in every future video.

Voice quality from HeyGen alone is weak. The X100 stack connects ElevenLabs as the voice provider — clone your voice in ElevenLabs from the same sample audio, then point HeyGen to it via the third-party voice integration. Scripting tricks for HeyGen: ellipses (…) or m-dashes for pauses, period-separated acronyms (M.D.E.), and phonetic URLs ("scott em pringham dot com"). Final polish: export from HeyGen, drop into Captions AI, run AI Edit to add B-roll and captions, ship. For tricky backgrounds, run the HeyGen export through Runway ML's background-removal tool before adding new backdrops.

Tools

• ChatGPT + X100 Rant Script Generator — run in your existing ICP thread to produce 10 rant scripts at a time. Ask for "longer scripts" if needed.
• Captions AI (paid Max plan for the features taught). Use the iOS version for newest features; desktop/Android lag behind. AI Creators feature + AI Edit feature.
• HeyGen (paid Creator plan minimum). For AI clones from a 2–3 minute sample video. Up to 100 avatars per person.
• ElevenLabs — voice cloning. Connect via API key into HeyGen for dramatically better voice than HeyGen's native cloning.
• Runway ML — unlimited background removal in the base plan; also inpainting from Topic 8.
• Veo 3 — for generating a walking/movement clip you can composite around your HeyGen clone when you only have a stationary sample.

Steps Discussed

1. Generate the script: open your ICP chat thread, paste the Rant Script Generator prompt from the X100 prompts page. Output: 10 rant scripts.
2. Pick one script. If it's too short, paste it back and ask ChatGPT to lengthen it.
3. AI Influencer video in Captions AI:
   1. Open Captions (iOS app preferred). Menu → All Tools → AI Creators → Prompt to Video.
   2. Choose an avatar (e.g., Bailey). Click "Custom" on the top-left to paste your own script verbatim.
   3. Click Generate Video and wait (~1–2 min).
   4. When done, choose AI Edit; pick a style (e.g., "Paper Two"). Captions adds captions, B-roll, animations, image overlays.
   5. Scroll the timeline; tap any image to Replace (upload or AI-generate a substitute).
   6. Export. Copy the auto-generated transcript on the right to feed back to ChatGPT for the post caption.
   7. Note: avatar selection is locked once you start generating — to change avatars, regenerate from scratch.
4. AI Clone in HeyGen — first-time setup:
   1. Open HeyGen → Create Avatar → "Start from video" (more realistic than "Start from photo").
   2. Shoot or repurpose a 2–3 minute video. Single subject. Look at camera the entire time. No hands across face. One language. Moderately lit, plain wall preferred. Lapel mic if outdoors.
   3. Speak with 40% more energy than feels natural. Match the emotional tone you want the clone to have.
   4. Upload the footage (preferred), record via webcam, or scan the QR for phone recording.
   5. Verify identity via webcam, reading the on-screen consent script + passcode.
   6. Name the avatar (e.g., "[Name] Studio v1" or "[Name] Outdoor").
   7. Wait for processing. You can create up to 100 avatars per person — repeat the process for different looks (studio, outdoors, casual, suited).
5. HeyGen video creation (ongoing):
   1. Open your avatar → Create with AI Studio.
   2. Paste your script. Preview voice; if HeyGen mispronounces a word, respell it phonetically.
   3. Insert ellipses (…) or m-dashes for pauses; periods between acronym letters; phonetic URLs.
   4. Connect ElevenLabs voice: My Voices → upload your sample to ElevenLabs → copy API key into HeyGen → choose your ElevenLabs voice.
   5. Generate, then export.
6. Edit the HeyGen export in Captions AI for B-roll, captions, and motion polish (same AI Edit flow as above).
7. If the background needs replacing: send the clip through Runway ML's background removal first, then add new background in Captions or Premier Pro.
8. For multi-person brands (e.g., Dave's Jimmy + Danny book launch): create one avatar per person from solo footage of each, write per-person scripts in HeyGen, and edit the outputs together as if it were a multi-camera shoot.

Resources

• X100 Prompts Page
• Captions AI (use iOS version for best results)
• HeyGen

Key Takeaways

• Two archetypes, two tools. Captions AI for AI-influencer (stock avatars). HeyGen for cloned-you.
• Use iOS Captions, not desktop. Features hit iOS first; desktop and Android lag months behind.
• Sample video quality = clone quality. Stiff sample = stiff clone. Bring 40% more energy than feels normal.
• One sample, one avatar. Different looks (studio, outdoor, suited) each need a separate sample/avatar. Don't mix into a single training set.
• ElevenLabs > HeyGen voice. Connect via API key for a step-change in realism.
• AI Edit in Captions does hours of editing in seconds. Add it as the polish step after HeyGen export.
• Existing professional footage can be the sample. Course shoots, podcast recordings, keynote footage all work — pick 2–3 uninterrupted minutes.
• Captions AI gives you the transcript. Don't rewrite the post caption — feed the transcript into ChatGPT to draft the social copy.
• Be early on new ad platform features. Meta typically rewards early adopters with better CPMs for ~12 months. (From the Advantage Plus stuck-point answer.)

Questions to Explore Further

• How do you handle natural body language across cuts when the HeyGen output is essentially a static-headshot mask over the original sample's gestures?
• What's the best way to A/B test AI-influencer vs AI-clone on the same script — does the same audience respond differently to a stock avatar than to a recognizable founder?
• For multilingual campaigns, does HeyGen + ElevenLabs preserve enough vocal identity across languages, or does the cloned voice sound generic outside the trained language?
• How long until Captions/HeyGen avatars are indistinguishable enough that disclosure becomes mandatory — and what's the ethical default now?
• What's the credit budget reality at scale — if a team ships 5 clone videos per founder per week, what's the realistic monthly HeyGen + ElevenLabs spend?

Steps I Need to Do on My End

1. Create 5 AI micro-influencer videos in Captions AI using the rant-script prompt and the AI Edit feature.
2. Create 5 AI clone videos in HeyGen (clone yourself or your founder first if you haven't yet).
3. Drop each finished video into the AI WhatsApp group for feedback — explicit assignment: "if you can do five of each, please do send them in the group."
4. If you don't have a clone yet, shoot a 2–3 minute sample video this week: one camera angle, look at camera the whole time, 40% extra energy.
5. Connect ElevenLabs to HeyGen via API key so every clone video uses the higher-quality voice.

Meeting Anatomy

Stuck-points round (Taylor on Vapi AI voice calling integrated with HubSpot / GoHighLevel, Jacob/Genis on realistic AI images / Shutterstock-trick / Midjourney reference image training) → framing of the four-pillar method with this session at the Distribution layer → walk-through of the Gary V content pyramid + the two-secret workflows (blog → social, podcast → social) → live demo of the BuzzFeed blog prompt → blog-to-script conversion → Quso long-to-shorts → editing handoff to Captions AI → conversation on podcasts as B2B lead-gen (Scott's Ford Motor anecdote, Zoom side-by-side recording setup, $250 podcast-studio-by-the-hour option) → assignment.

Summary/Recap

The thesis: content creation isn't the bottleneck anymore — distribution is. Marketing teams can already produce posts, images, and clips; what's missing is a system that takes one pillar piece and produces 90 days of distributable content. Two AI-amplified workflows handle that:

Secret #1 — Blog → Social. Run the BuzzFeed blog prompt inside your ICP/messaging thread → ask ChatGPT to "convert this blog into a video script" → film yourself reading the script with a teleprompter app (Prompt Smart paces scroll to your voice; Captions AI has one built in) → drop the long video into Quso → Quso outputs 5–12 short reels with virality scores → run each reel through Captions AI for AI-Edit finishing (captions, B-roll, music). One blog post becomes: long-form YouTube video, 7 short reels, 7 static-image posts, a 7-touch email sequence — and the original blog itself.

Secret #2 — Podcast → Social. Podcasts are free (Zoom side-by-side speaker gallery is plenty) and they double as B2B lead-gen — Scott landed Ford's North America team by inviting the agency president onto his "Game Changers" podcast (the prospect never asked how many listeners). Frame the podcast around the guest's needs, not your product. Send the guest the Quso-clipped content afterwards as a thank-you and co-marketing tool. Tool stack is ~$75/month total: ChatGPT + Quso + CapCut + Captions AI — replacing what used to be a dedicated editing team. For framing, use the rule-of-thirds grid in your iPhone camera: eye-line on the upper horizontal, shoulders on the vertical lines. Minimum iPhone 10 for usable footage. Anything bigger than 1 GB upload to YouTube as Unlisted and paste the link into Quso (Quso accepts up to 5 GB direct on the paid plan).

Tools

• ChatGPT — BuzzFeed blog prompt, blog-to-script conversion, multi-format repurposing prompts.
• Quso AI — long-to-short clipping; better sizing flexibility (1:1, 4:5, 9:16, 16:9) and built-in branding than Opus. Free plan watermark; paid removes it. Up to 1 GB upload free / 5 GB paid; or paste a YouTube link (max 5 GB).
• Opus Clips — alternative to Quso; sometimes produces more or fewer clips per source. Test both.
• Captions AI — AI Edit step after Quso for B-roll, captions, music, animations. Also has a teleprompter.
• HeyGen — if you don't want to film yourself, generate the long-form video using your clone (Topic 9) before clipping.
• Prompt Smart — teleprompter app that paces scroll to your voice.
• CapCut / Adobe Premier Pro — manual edits when AI tools aren't enough.
• Descript — edit video by editing the transcript.
• Zoom / Google Meet / Teams — free podcast recording (use Zoom's side-by-side speaker view).

Steps Discussed

1. Blog → social workflow:
   1. Open your ICP/messaging chat thread or custom GPT.
   2. Paste the BuzzFeed blog prompt from the X100 prompts page. Set the number to 7 and pick mistakes over secrets for your first attempt.
   3. Save the resulting blog to a Google Doc. (You also get SEO keyword, meta description, schema markup, image prompt.)
   4. In the same thread, prompt: "convert this blog into a video script."
   5. (Optional) Same thread: "convert this blog into 7 LinkedIn posts" and "convert this blog into a 7-touch email sequence using the soap-opera framework."
   6. Open Prompt Smart (or Captions AI's teleprompter), paste the video script.
   7. Set up your iPhone (10+) on a tripod. Turn on the camera grid: eye-line on top horizontal third, shoulders on the vertical lines.
   8. Hit record and read the script with energy. If you don't want to film, generate the long-form video in HeyGen using your clone.
   9. Upload the long video to Quso → Long-to-Shorts → Generate Clips. Wait 3–5 minutes.
   10. Review each clip with its virality score. Edit captions inside Quso OR turn captions OFF and download the raw clip.
   11. Pass each raw clip through Captions AI → AI Edit → pick a style for the finishing layer.
   12. Schedule/publish directly from Quso to TikTok, Facebook, Instagram, LinkedIn, X, YouTube, Pinterest.
2. Podcast → social workflow:
   1. Pick a guest — ideally an ICP-adjacent decision-maker you wouldn't otherwise reach. Frame the podcast around their wins, not your product.
   2. Record on Zoom. In View, switch to Side-by-Side Speaker so the recording captures both faces cleanly.
   3. (Or book a local podcast studio — many cities have ~$250/hour studios that hand you the file.)
   4. Upload the recording to Quso (or paste an Unlisted YouTube link if > 1 GB free).
   5. Quso returns 12–20+ short clips. Use the highest virality-score clips first.
   6. Send the guest the clipped content as a co-marketing thank-you — increases reach via their feed and often leads to introductions.
3. Vapi inbound-call routing (from the Taylor stuck-point):
   1. Set up Vapi as an inbound AI-voice that calls leads instantly when a Meta lead-form is submitted.
   2. If unanswered, retry; if answered, record and transcribe the call.
   3. On affirmative ("yes I want to speak to sales"), send the transcript + recording into the CRM (HubSpot / GoHighLevel) and book a sales-rep appointment.
   4. Use ChatGPT as a setup consultant — paste screenshots and ask it to walk through the Vapi config step-by-step.

Resources

• X100 Prompts Page
• Quso
• Captions AI
• HeyGen

Key Takeaways

• Content creation isn't the bottleneck anymore — distribution is. A pillar piece + an AI repurposing stack solves it.
• One blog can become a full month of content. Long video + 7 reels + 7 static posts + 7-touch email + the blog itself.
• Quso beats Opus on flexibility, Opus sometimes returns more clips. Run the same source through both and keep the winner.
• Captions AI is "finishing school." Always finish a Quso clip in Captions for B-roll, music, animations.
• Frame the iPhone with the rule of thirds. Eye-line top horizontal, shoulders on the verticals. Editor needs space for captions.
• Podcasts are the cheapest B2B lead-gen tool. Zoom side-by-side + an interesting guest = relationship, content, and an entry path to their org.
• $75/month tool stack replaces an editing team. ChatGPT + Quso + CapCut + Captions.
• For tricky shoots, use HeyGen for the pillar video. No camera, no studio, just a script and your clone.
• For realistic images, train Midjourney on reference photos. The "ball pit" training is the example — indistinguishable from real photography.

Questions to Explore Further

• What's the optimal cadence — one new pillar per week (52 a year), or one per month with deeper repurposing?
• How do you avoid Quso fatigue when the same 7 mistakes get clipped into 7 reels — should the reel hooks be rewritten in ChatGPT first?
• For podcasts used as B2B lead-gen, what's the right time window from recording → outreach → meeting before the relationship goes cold?
• When Vapi handles inbound and qualifies a lead, what disclosure language is required so the prospect knows they're talking to an AI?
• Is there a measurable lift from the Captions AI finishing layer vs. raw Quso output — does the extra editing time actually move engagement?

Steps I Need to Do on My End

1. If you have a podcast (or want one): commit to a date for your first or next episode before next week's session. Even if you don't have a guest yet, lock the date.
2. Create 7 pieces of short-form content from a long-form source by next module — could be a blog you record, a HeyGen long video, or a recorded Zoom podcast.
3. Use the full chain: ChatGPT (script) → film or HeyGen (long video) → Quso (clip) → Captions AI (finish) → schedule.
4. Get really creative with who you put on the podcast — Scott's example: Daniel's high-end audio business could host a real-estate agent or home builder in Del Mar.

Meeting Anatomy

Stuck-points round (Matt on Descript transcript-based editing, Bill on Gemini AI Pro / Nano Banana 3 / VEO 3 access, group thumbs-up vote on a 3–4 part video editing series) → housekeeping (Bootcamp 2.0 release, December holiday adjustments, EOS coach Adam Pontrelli in the AI WhatsApp group) → recap of the EOS scorecard structure → live walk-through of the X100 Scorecard Analyst prompt against the Foundrs client scorecard → assignment.

Summary/Recap

This session sits at the Optimization pillar. The thesis: gut-feel optimization fails — you can't improve what you can't measure. The EOS scorecard (which all members have inside their Level 10 doc) is the source of truth. Adam Pontrelli's "desert island" framing: if you could only see 5–15 numbers and run your business off them, which would they be? Lean toward 5; bias toward leading indicators (activity-based) over lagging ones (revenue).

The new tool taught: the X100 Scorecard Analyst prompt (at the bottom of the X100 prompts page). It instantiates a ChatGPT persona that is simultaneously a data analyst and an expert social media marketer trained on Traction, Rocket Fuel, and the rest of the EOS canon. Drop in your scorecard (CSV, Excel, or screenshot), and it returns trend analysis, leak diagnosis, and prioritized action items. Critically, it interprets why a number is moving — e.g., clicks dropping while spend and impressions hold steady = creative fatigue, not targeting; impressions strong but clicks weak = creative problem; clicks strong but leads weak = offer problem.

The three secrets: (1) Diagnose before you fix — symptoms are not causes; (2) Evolution — every 90 days, take the analyst's output and feed it back into your custom GPT's knowledge files so the system compounds learnings; (3) Scale the winners — re-run earlier content/ad prompts with the new learnings baked in. Owner rule, restated from EOS canon by Adam: a measurable can only have one owner — if two people "own" a number, no one does. Start small (3–5 metrics), be religious about weekly fills, expand only after consistency lands.

Tools

• ChatGPT + the X100 Scorecard Analyst prompt — at the bottom of the X100 prompts page. Run inside your ICP/messaging thread or custom GPT.
• EOS Scorecard (linked from inside your Level 10 doc → Rocks section) — owner + measurable + baseline + weekly goal across 5–15 metrics.
• Custom GPT — preferred over a chat thread; feed the Scorecard Analyst's quarterly output back into the GPT's knowledge files so learnings compound.
• EOS Coach (Adam Pontrelli) — in the AI WhatsApp group; ping with quick questions.
• Descript (mentioned during stuck-points) — transcript-based video editing.
• Gemini Pro / Nano Banana 3 / Kling AI / VEO 3 (stuck-point sidebar) — for product video generation; if you have a Google Workspace account, Gemini Pro is included.

Steps Discussed

1. Find your EOS scorecard — it lives in the Resources / Rocks section of your Level 10 document.
2. If you don't have one, build it:
   1. Pick 3–5 metrics to start (max 15). Lean toward leading indicators.
   2. For each metric: owner (one person, not two), measurable, baseline (your last 90-day average), weekly goal.
   3. Example marketing metrics: media spend, impressions, clicks, leads, sales-qualified leads / sales calls booked.
   4. Commit to filling it out on a fixed day each week. Consistency > comprehensiveness.
3. After ~90 days of weekly data, run the Scorecard Analyst:
   1. Open your ICP chat thread or custom GPT.
   2. Copy the X100 Scorecard Analyst prompt from the bottom of the X100 prompts page.
   3. Paste, hit enter. The prompt will ask for your scorecard.
   4. Upload as CSV, Excel, or a clear screenshot.
4. Review the analyst's output:
   1. Trend analysis — which metrics are up, flat, down vs. baseline.
   2. Diagnostic pass — why a metric is moving (creative fatigue vs. offer mismatch vs. targeting drift).
   3. Prioritized action items — concrete next-step changes ranked by impact.
5. Take action on the top 1–3 recommendations. Don't try to fix everything.
6. Every 90 days, save the analyst's output as a Word doc and upload it into your custom GPT's knowledge files — the system gets sharper each quarter.
7. Re-run earlier prompts (Gary V social, Joanna Wiebe email, Don Draper messaging) with the new learnings already baked into the GPT — outputs improve without manual prompt-engineering.

Resources

• X100 Prompts Page

Key Takeaways

• Diagnose before you fix. The dashboard shows symptoms; treating symptoms wastes the next 90 days.
• 5 numbers measured weekly beat 15 measured monthly. Adam's desert-island framing — pick the minimum viable set and hold it.
• One owner per measurable. Two owners means zero accountability.
• Leading indicators over lagging ones. Activity-based metrics (DMs sent, posts published, impressions) tell you about next quarter; revenue tells you about last quarter.
• The Scorecard Analyst is a data analyst + social-media expert in one prompt. It knows the EOS canon AND the digital-marketing playbook, which is rare.
• Quarterly knowledge-file refresh = compounding GPTs. Every 90 days, drop the analyst's report into your GPT's knowledge so future outputs get smarter.
• Common diagnostic patterns: clicks down + spend flat = creative fatigue; impressions up + clicks down = creative problem; clicks up + leads down = offer problem.
• Strategy ↔ execution lockstep is non-negotiable. Offshore/VA teams that diverge from the founder on ICP/messaging produce content that no analyst can save.

Questions to Explore Further

• When the Scorecard Analyst flags creative fatigue, what's the right re-prompting protocol — generate 4 new Gary V variations and A/B test, or wait for the analyst to suggest a creative pivot?
• How do you keep the scorecard alive when the marketing function is a one-person VA with no founder involvement — does the weekly fill survive without an enforcement loop?
• For multi-product or multi-brand businesses, do you run one scorecard with 15 metrics or N scorecards with 5 metrics each — and how do you weight cross-brand insights?
• What's the right cadence to feed the analyst's output into the custom GPT — 90 days as taught, or sooner if a clear win lands at day 30?
• How does the Scorecard Analyst handle conflicting signals (e.g., paid ads dropping while organic spikes) — does it propose attribution rebalancing, or just flag both?

Steps I Need to Do on My End

1. If you've been maintaining your EOS scorecard: run the X100 Scorecard Analyst prompt against it and get a list of prioritized action items. Explicit assignment.
2. If you haven't been maintaining it: design your scorecard this week — pick 5 (up to 15) metrics with owner + measurable + baseline + weekly goal.
3. Lean toward leading indicators (activity-based) and one owner per metric.
4. Commit to filling out the scorecard on a fixed day each week — "promise to yourself you will fill it out on time every single week."
5. Plan to feed the analyst's quarterly output back into your custom GPT's knowledge files so learnings compound.

Meeting Anatomy

Stuck-points round (Robert's account health on Instagram — drop from 30K followers to 2K views, advice on stitches with relevant creators + alias-account split test; Muzammil on the Prompt Engineer prompt as the meta-prompt for custom use cases) → handoff to the Module 11 (Bonus 1) landing-page training → walk-through of the AI Website / Landing Page Builder prompt's ~36 design questions → masterplan.md generation in ChatGPT → live build in Lovable.dev → iteration via chat (color change demo) → deployment + CRM connection conversation (Armando on Go High Level) → assignment.

Summary/Recap

The bonus module: replace the $5K agency landing-page build (developer + UI/UX + copywriter + PM, ~$5K+ minimum) with ChatGPT + Lovable.dev in ~30 minutes. The X100 prompt that powers it is the AI Website / Landing Page Builder prompt on the prompts page — long, with Russell Brunson's frameworks baked in. Run it in the same ChatGPT thread as your ICP, core messaging, and lead magnets so all the answers inherit that context.

The prompt asks ~36 sequential questions about the page: type (landing / home / sales / funnel step), primary purpose, CTA, target audience, offer name, tone, emotional angle, big promise, story elements (Scott's "billion-dollar man" angle is in his thread), media (photos / video / testimonials / logos), competitor pages, model-of-success references, color palette (screenshot a brand image and ask it to extract), typography, animation style, hero layout, look/feel, AI brain section, headline, testimonial-section behavior, thank-you page, upsell CTA, etc. Answer A/B/C/D for the multiple-choice ones, or pick "other" and describe. Most questions take ~30 seconds.

When the questionnaire ends, ChatGPT generates a giant masterplan.md — copy everything from "masterplan.md" downward (strip the preamble) and paste it into a new Lovable.dev project. Lovable returns a first-pass page in 2–5 minutes. Iterate via chat only — describe each change as if briefing a developer. Don't touch the visual editor: it poisons the underlying code and forces a restart. For exact brand colors, screenshot a brand asset, run it through a hex picker, paste the hex code into Lovable chat. Free plan = 5 credits (5 iterations); paid = more. Publish via Publish → Custom Domain → connect to Lovable's domain flow (or buy through Lovable). For CRM integration, ask Lovable to wire to Make.com, Supabase, or a webhook → then route into Go High Level / HubSpot / Salesforce. Lovable outputs React/JS; cannot be converted to WordPress. For Go High Level users: export Lovable's code via the Code button (or GitHub), and either paste the HTML/CSS into GHL's custom code block, or feed the full code export to ChatGPT and ask it to adapt for GHL.

Meta-method: the AI Website Builder prompt itself was built by interviewing a website developer / CTO on Zoom about every question they'd ask a client, dumping the transcript into the X100 Prompt Engineer prompt, and letting ChatGPT reverse-engineer the expert prompt. The same method built the "Steve Jobs prompt" for app-building from a YouTube video of someone speed-building an app in 8 hours.

Tools

• ChatGPT + AI Website / Landing Page Builder prompt — run inside your existing ICP/messaging chat thread.
• Lovable.dev — AI website builder. Free plan: 5 chat-iteration credits. Paid required for custom domain.
• Prompt Engineer prompt — the meta-prompt for building new domain-specific prompts from expert interviews or transcripts.
• Cloudflare / GoDaddy / NameCheap — for connecting an existing domain to Lovable.
• Make.com / Supabase — connect Lovable forms to your CRM via webhook or database.
• Go High Level / HubSpot / Salesforce — destination CRMs.
• GitHub — export Lovable code for full control or migration.
• Hex picker (any web hex-from-screenshot tool) — brand color matching.

Steps Discussed

1. Open your existing ICP/messaging chat thread in ChatGPT (so the page inherits persona, story, lead magnets).
2. Copy the AI Website / Landing Page Builder prompt from the X100 prompts page → paste → hit enter.
3. Answer ~36 questions sequentially:
   1. Page type → primary purpose → desired user action (CTA) → target audience (reference your ICP).
   2. Offer name / title → tone & personality → emotional angle → big promise.
   3. Story elements (which angles of your backstory to feature).
   4. Media elements (photos, video clips, testimonials, logos). Answer with comma-separated letters (e.g. A, C, D) for multi-select.
   5. Competitor pages and "model of success" references (e.g., apple.com).
   6. Color palette — upload a brand screenshot and ask it to extract; choose "keep core but slightly enhance."
   7. Typography, animation style, hero section layout, look & feel (dark mode / futuristic), AI brain section, headline approach, testimonial section behavior, thank-you page, upsell CTA.
4. When ChatGPT confirms "I have everything I need, generating masterplan.md" → say yes if needed → wait for the long block.
5. Copy the masterplan.md block. Strip the preamble; the prompt should start with "masterplan.md".
6. Open lovable.dev → New Project → paste the masterplan.md → wait 2–5 min for the first-draft build.
7. Iterate via chat only:
   1. "Change the headline to X."
   2. "Replace the gradient with solid #0F8A4D."
   3. "Add a testimonial section with 3 quotes."
   4. Never click into the visual editor — describe everything in chat.
8. For brand colors: screenshot a brand asset → hex picker → paste the hex string into Lovable chat.
9. Connect a CRM via chat: "Wire the lead form to Make.com via webhook" (or Supabase, or directly to a known integration).
10. Publish: Publish button → Custom Domain → connect via Cloudflare/GoDaddy/NameCheap, or buy a domain inside Lovable. Requires paid plan.
11. For Go High Level: click the Code button in Lovable → export the HTML/CSS → paste into GHL's custom-code block. If languages don't match, paste the export into ChatGPT and ask it to adapt for GHL.
12. (Meta-method) To build a new domain prompt: interview an expert on Zoom (or download a relevant YouTube video) → transcribe → paste into the Prompt Engineer prompt → ChatGPT produces the expert-grade prompt.

Resources

• Lovable
• X100 Prompts Page

Key Takeaways

• Run the AI Website Builder prompt in your ICP thread. The 36 questions inherit your persona, story, lead magnets — that's where the conversion comes from.
• Never use Lovable's visual editor. It poisons the underlying code; iterate via chat only.
• Screenshot + hex picker = brand-perfect colors. Paste hex codes into Lovable chat for exact matches.
• Russell Brunson frameworks are baked into the prompt. You don't need to know them; the prompt enforces them.
• 5 free credits = 5 chat iterations. Plan your refinement messages carefully on the free plan.
• Cannot convert to WordPress. Lovable outputs React/JS — pick Lovable hosting or export to GHL/HubSpot/static.
• CRM integration is a chat instruction. "Wire to Make.com via webhook" usually works.
• The Prompt Engineer prompt is the unlock for everything. Want a tax-prep / podcast-launch / app-build prompt? Transcribe an expert, pipe through Prompt Engineer.
• Coded landing pages load faster than WordPress. Lovable's React output beats most CMS pages on mobile/desktop speed.

Questions to Explore Further

• What's the right cutoff between "iterate in Lovable" and "export and finish in code" — at what point does iterating via chat get more expensive than handing off to a developer?
• For multi-page websites (vs. single landing pages), does the masterplan.md flow scale, or do you need to chunk it into per-page generations?
• How well do Lovable-built pages perform in Google Search vs. WordPress — does the React rendering hurt SEO, and how do you mitigate?
• For Go High Level users, is the round-trip (Lovable → ChatGPT adaptation → GHL custom code) actually faster than building natively in GHL?
• What's the right way to A/B test variants — fork the masterplan.md and build two Lovable projects, or do everything in one Lovable page with conditional rendering?

Steps I Need to Do on My End

1. Build at least one landing page in Lovable using the AI Website / Landing Page Builder prompt before next week. Explicit assignment: "Try to build a landing page before next week if you can."
2. Run the 36-question prompt inside your existing ICP/messaging thread so the page inherits your persona and story.
3. Iterate via chat only — do not touch Lovable's visual editor.
4. Drop the finished page link into the AI WhatsApp group for feedback.
5. (Optional) Use the Prompt Engineer prompt to build one new domain-specific prompt of your own this week.

Meeting Anatomy

Guest-led session by Dave (SEO/GEO specialist, based in Toronto) — five-fundamentals SEO foundation → three SEO pillars (on-page, off-page, monitoring) → live audits of attendee websites (Diego/Blue Valley Tractor for PageSpeed Insights; Ann/Kline Products for SEMrush keyword research + LLMTEL.com app-of-the-day; Lauren/Christina for schema.org validator) → AI core-knowledge vs AI-search distinction → Wikipedia / Wiki data / Common Crawl strategy → PR and backlink profile shape (natural vs engineered) → directories and entity consistency → Q&A.

Summary/Recap

Dave reframed SEO around GEO — Generative Engine Optimization: getting found inside AI models and AI-augmented search, not just blue links. Five foundational truths: (1) entity authority beats page authority — AI cares who you are, not which URL ranks; (2) Google's EEAT (Experience, Expertise, Authoritativeness, Trustworthiness) is the rubric for surfacing in AI results; (3) SEO + PR = SEO for AI — independent third-party coverage drives entity authority; (4) AI has two layers — core knowledge (set in stone after the training cutoff; GPT-5 released Aug 2025 was trained to Oct 2024) and AI search (variable RAG that fetches live information per query); (5) different team members are at different points on the AI adoption curve — fear → skepticism → "oh my goodness" → investigation → adoption → mastery.

Three SEO pillars: on-page (technical + content + style), off-page (backlinks, PR, Wikipedia, knowledge graphs, directories), and monitoring (Search Console, LLMTEL, etc.). On-page essentials: PageSpeed Insights target < 2s mobile load (ideally ~1s); proper H1/H2/H3 hierarchy with keywords early; alt text on every image; schema markup for entities/logo/breadcrumbs; clear short paragraphs at a middle-school reading level (matches how LLMs surface text). Screaming Frog audits site-wide technical issues.

Off-page essentials: build a knowledge graph by getting Wikipedia (or Wikidata if you can't get the page) coverage — Wikipedia is a heavily curated source LLMs rely on. Check Common Crawl to see if your site is in the canonical web crawl LLMs use. Get covered in Google News (PR drives notability, which drives Wikipedia eligibility). Build a natural backlink profile (Majestic.com visual should look organically clustered, NOT a perfect geometric pattern — that's engineered link-building and gets penalized). Keep NAP (Name/Address/Phone) identical across directories. Podcasts are indexed in ~4 days — fastest authority signal you can buy. Monitor with Google Search Console, Google Analytics (sources now include ChatGPT, Perplexity), and LLMTEL.com (10 free reports testing 17 LLMs).

Critical when-not-to-use-AI guidance: use AI for base-level evergreen content; do NOT use AI for news, thought leadership, or industry-specific innovations. AI output is the statistical average of the internet — and "average" can't differentiate you in search. The average is rising fast, so you have to be above it.

Tools

• PageSpeed Insights (pagespeed.web.dev) — Core Web Vitals + performance diagnosis. Target < 2s mobile load.
• Screaming Frog — site-wide technical SEO crawl (page titles, blocked URLs, header structure).
• Schema.org Validator — confirms structured data is parseable.
• Google Rich Snippets / Rich Results Test — checks Google-side schema visibility.
• SEMrush — paid; keyword research, competitor analysis, domain authority.
• Majestic — backlink profile visualization (natural vs engineered).
• LLMTEL.com — 10 free reports testing how you show across 17 LLMs.
• Google Search Console + Google Analytics — track performance, including AI-source referrals.
• Google Trends + Google Keyword Planner + Google autofill — free keyword research.
• GPT for Sheets — ChatGPT formulas inside Google Sheets via API.
• news.google.com — check if you have PR coverage (notability test for Wikipedia).
• Common Crawl — confirm your site is in the LLM-canonical web crawl.
• Wikipedia / Wikidata — entity-authority anchor.

Steps Discussed

1. Audit on-page:
   1. Run your URL through PageSpeed Insights. Aim for < 2s mobile (ideal 1s); pass Core Web Vitals; hand the diagnostic list to your developer.
   2. Run Screaming Frog. Fix: page titles outside <head>, missing H1s, internal pages blocked by robots.txt, missing meta descriptions.
   3. Put keywords early in title tags and meta descriptions.
   4. Verify proper H1/H2/H3 hierarchy on every page (like a Word doc outline).
   5. Add alt text to every image with a keyword-rich description.
2. Add schema markup:
   1. Visit schema.org validator; paste your URL.
   2. Confirm webpage, organization, logo, breadcrumbs are present.
   3. Add sameAs entries pointing to your LinkedIn, Facebook, X, etc., so LLMs know they're the same entity.
   4. Aim for zero errors and zero warnings AND maximum detail.
3. Keyword research:
   1. Drop the URL into SEMrush; pull competitors and their keyword overlap.
   2. For each candidate keyword, check Difficulty %, search volume, and competitor backlink count — low difficulty + reasonable volume + low competitor backlinks = winnable.
   3. Cross-check with Google Trends, Keyword Planner, and Google autofill.
   4. Survey your customers and check Reddit / Quora for question patterns.
4. Build entity authority:
   1. Search yourself on news.google.com. If you have PR coverage, you're notable enough for Wikipedia.
   2. If you have a Wikipedia-worthy story, get a page (Wikidata as fallback if the page can't be created).
   3. Check Common Crawl for inclusion.
   4. Get on a podcast (or host one) — indexes in ~4 days.
   5. Audit Majestic backlink profile — natural cluster pattern, NOT geometric/engineered.
5. Directory hygiene: NAP (Name/Address/Phone) must be byte-identical across every directory. Inconsistency dilutes entity authority.
6. Use AI for base content, NOT differentiation: evergreen pages on water trucks / mortgages / hardwood = AI is fine. News, thought leadership, proprietary innovations = write yourself.
7. Monitor:
   1. Set up Google Search Console and put the snippet on the site.
   2. Set up Google Analytics; watch the Sources panel for ChatGPT, Perplexity referrals.
   3. Run free LLMTEL.com reports periodically — shows how you appear (or don't) across 17 LLMs and what corrections you need.

Resources

No external resources for this module.

Key Takeaways

• Entity authority beats page authority. AI ranks who you are, not which URL is "best."
• SEO + PR = SEO for AI. Independent third-party coverage is now load-bearing.
• Core knowledge ≠ AI search. Core is fixed at training cutoff; AI search is RAG over the live web. If LLMs have you wrong in their core knowledge, you can't directly edit — you have to flood the web with correct signals.
• Use AI for base content only. AI output is the average of the internet — average doesn't differentiate.
• Wikipedia is the biggest single LLM signal. If you can't get a Wikipedia page, get Wikidata.
• Podcasts index in ~4 days. Fastest authority signal available.
• Natural backlink profile beats engineered. Majestic visualization should be clustered, not geometric.
• NAP consistency across directories is non-negotiable. Same name, same address, same phone, everywhere.
• Target < 2s mobile load. Ideally 1s. Big background images are the #1 killer.
• Write at middle-school reading level. LLMs surface content at that level — match it.
• Add sameAs to schema. Links your website entity to your social profiles in the knowledge graph.

Questions to Explore Further

• For non-notable B2B businesses (small water-truck manufacturer, niche industrial service), what's a realistic 12-month path to Wikipedia or Wikidata eligibility — is it press releases, podcast appearances, or industry-association coverage?
• When LLMs have an entity wrong in their core knowledge, what's the most direct correction path — Wikipedia edits, schema updates, or new PR coverage to outweigh the old signal?
• Does Dave's "podcasts index in 4 days" claim hold across all niches, or is it specific to topics LLMs already weight heavily?
• For Lovable-built React landing pages (Topic 12), do crawlers reliably parse schema, or does the JS rendering create blind spots that need pre-rendered HTML?
• What's the right Search Console + Analytics setup to attribute traffic correctly when a user discovers you on ChatGPT/Perplexity and then converts via direct URL paste?

Steps I Need to Do on My End

No homework assigned in this session.

Meeting Anatomy

Framing of agents in the broader "post the pillars" automation arc → three-secret structure (Tools, Framework, Build) → live build of a Rant Script Generator AI agent in Make.com from scratch (OpenAI connection with API key + Org ID → GPT-5.1 module → Rant Automation Prompt with ICP/messaging pasted in → Repeater set to 5 → Google Sheets output) → live test (5 scripts populate the sheet) → extension demo (add LinkedIn / Gmail output) → preview of a more advanced 3-stage ABC pipeline → Q&A (Matt on Pabbly, Robert on multi-account posting, Kevin on scheduling) → assignment.

Summary/Recap

An AI agent is anything that does things for you without you being involved. Three secrets:

Secret #1 — Tools (Brain + Body). The Brain is the LLM (ChatGPT / Gemini); the Body is the automation layer (Make.com, Zapier, Pabbly Connect, n8n). Choose the body that has connectors to every tool you need — if Pabbly lacks a LinkedIn connector and your agent posts to LinkedIn, you've picked the wrong body. Zapier has the broadest connector library but is the most expensive. Make.com free tier = 1,000 ops/month (often enough to run permanently).

Secret #2 — Framework. Write down every manual step you'd take to do the task by hand. Each step becomes an automation module. For the rant-script use case, manual steps were: open ChatGPT → load ICP/core messaging → research the trending topic → generate five rant scripts → drop into a Google Sheet → send to Scott. Those four-to-six steps become four-to-six Make.com modules.

Secret #3 — Build. Wire up the modules. To connect OpenAI to Make: API Key + Organization ID (both retrieved from OpenAI's developer page; organization must be verified). Pick the model — GPT-5.1 for quality, GPT-4o for speed/cost. Paste the Rant Automation Prompt with your ICP, core messaging, and website URL filled in. Output to a Google Sheet (or Airtable) with a "Repeat 5 times" iterator to get five scripts per run. Schedule the scenario to run every morning at 8am and you've replaced your social media VA's research-and-scripting role.

To extend: add a LinkedIn "Create User Text Post" module wired to the same ChatGPT output for auto-posting; add a Gmail module to email the scripts; add per-account LinkedIn connections to fan out across multiple founders. A more sophisticated 3-stage ABC pipeline appears in the X100 vault: A = research agent (ChatGPT → Airtable), B = content production (caption + image → Google Drive → Airtable), C = distribution (Airtable → LinkedIn after human approval). The human approval row in Airtable is a guardrail — without it, LinkedIn flags you for posting every ~90 seconds.

Tools

• ChatGPT / OpenAI — the brain. Used via API from Make.com.
• Make.com — the body. Free tier = 1,000 credits/month. Each module action = 1 credit.
• Zapier / Pabbly Connect / n8n — alternatives to Make. Choose by connector coverage. Zapier = widest but priciest.
• Google Sheets / Airtable / Microsoft Excel — destination for generated content.
• LinkedIn / Gmail / Facebook Pages / Instagram — output destinations available as Make modules.
• X100 Rant Automation Prompt — on the prompts page; the seed prompt for the live demo.
• OpenAI API Key + Organization ID — retrieved from OpenAI's developer dashboard; required once to connect.

Steps Discussed

1. Pick the body: confirm Make.com has connectors for every system your agent needs (LinkedIn, Sheets, Gmail, etc.). If a connector is missing, switch bodies.
2. Create a Make.com account → New Scenario → name it (e.g. "AI Agent for Creating Rants"). Blank canvas appears.
3. Connect the brain:
   1. Add module → search OpenAI → choose "Generate a Response."
   2. Click "Add Connection." Use the two hyperlinks Make provides — one opens the OpenAI API Keys page, the other opens the Org ID page.
   3. Create an API key (name it anything; copy immediately — only visible once). Paste into Make.
   4. Verify your organization on OpenAI if needed; copy the Org ID; paste into Make. Save.
4. Configure the OpenAI module:
   1. Model: GPT-5.1 for quality, GPT-4o for speed and lower credit cost.
   2. Prompt Type: Text Prompt.
   3. Paste the X100 Rant Automation Prompt; fill in placeholders for ICP, core messaging, and website CTA URL.
5. Add the output destination:
   1. Pre-create a Google Sheet "Rant Automation AI Agent" with a "script" header.
   2. In Make: add Google Sheets → Add Row module. Connect your Google account. Pick the drive + sheet + Sheet1.
   3. Map the script column to the OpenAI module's Output → Content → Text field.
6. Repeat to generate 5 scripts per run: Insert a Tools → Repeater module at the front; set Repeats to 5.
7. Click Run Once. Confirm five rows appear in the sheet, one rant script per row.
8. Schedule: at the bottom-left of the scenario, set the cadence — e.g., every 24 hours at 7:30am. The agent now runs unattended.
9. Extend:
   1. Add a LinkedIn "Create User Text Post" module wired to the OpenAI output → auto-posts each script.
   2. For multi-account posting, add multiple LinkedIn connections (one per founder/account).
   3. Add a Gmail module to email the scripts to the founder instead.
   4. For an approval gate, route through Airtable with an "approved Y/N" column; only post when the column flips to Y.
10. For image-input prompts (custom GPTs that accept images): in Make's OpenAI module, switch from "Generate a Response" to "Message an Assistant" — it behaves like a Custom GPT callable from Make.

Resources

• Prompts Page
• Presentation

Key Takeaways

• An agent = Brain + Body + Framework. Pick the brain first (LLM), pick the body by connector coverage, then write the manual workflow and replace each step with a module.
• Pick the body by what it can connect to. Don't pick Make if your destination isn't in Make's connector list.
• Verify your OpenAI organization. Without verification, no Org ID, no API connection.
• Use GPT-5.1 for quality, GPT-4o for cost. Free OpenAI credits run out fast on the flagship model.
• Repeater modules turn "1 script" into "N scripts" cheaply. Wrap the OpenAI module in a Repeater to multi-output.
• Always include a human approval layer for posting. LinkedIn / Meta will throttle or ban accounts that post every 90 seconds.
• "Generate a Response" is fine for text. For image inputs, use "Message an Assistant." It behaves like a Custom GPT callable from Make.
• Same agent, different outputs. The same chain can post to LinkedIn, email a founder, or drop into a sheet — just swap the destination module.
• Make.com free tier is genuinely free. 1,000 ops/month covers many real agents.

Questions to Explore Further

• When does an agent become worth the build vs. just running the prompt manually — what's the threshold of runs/week where Make.com pays for itself in time saved?
• For the ABC pipeline, what's the right human-approval cadence — does a daily batch-approval flow work, or does the agent need per-post approval to keep quality up?
• If your social platform isn't a native Make connector (e.g., Threads, BlueSky), is it faster to switch bodies or to wire via a generic webhook?
• How do you keep the ICP + core messaging fresh inside an agent — refresh the prompt template quarterly, or pull from a versioned doc that Make reads on each run?
• For multi-founder agencies, what's the right account architecture — one Make.com workspace with multiple scenarios per client, or one scenario per client with shared modules?

Steps I Need to Do on My End

1. Build your first AI agent this week. Could be a rant-script generator, a daily research email, a content drip, or anything else. Explicit assignment: "create your first AI agent."
2. Start by writing down the manual workflow you'd take to do the task by hand — that becomes your module list.
3. Verify your OpenAI organization and grab the API key + Org ID so Make can connect.
4. Pick GPT-4o initially to conserve credits; switch to GPT-5.1 once the agent works and you want quality.
5. If the agent posts to social, add a human-approval gate (Airtable column) to avoid platform bans.

Meeting Anatomy

Guest-led session by Nick Lopez (Meta ads expert) — five-step framework (objective, budget, audience, creative, measurement) → prerequisites checklist (Business Manager, Pixel, CAPI, payment) → meta-ads taxonomy (account → campaign → ad set → ad) → objective-picker chart → budget math with the 50-conversions-per-week rule → live diagnostic-style questions from Matt (local service B2B), Kevin (fencing), Muzammil (Google → Meta retargeting), Sanchez (A/B test setup) → close with creative/measurement guidelines.

Summary/Recap

Nick laid out a five-step process for getting started on Meta. Step 1: pick the objective — sales, leads, awareness, traffic, engagement, or app. Choose based on what you actually want; awareness compliments paid search rather than replacing it.

Step 2: budget. Daily for evergreen, lifetime for date-bounded promos. Minimum ~$20/day so the algorithm has room to learn. The load-bearing rule: Meta needs 50 conversions/week to exit the learning phase, so weekly budget = CPA × 50. If your target CPA × 50 exceeds what you can afford (Nick's strawberry example: $50 CPA × 50 = $2,500/week, vs $3,000/month budget), don't lower the target CPA — switch to an easier upstream conversion event (add-to-cart, key page view, sign-up) until the algorithm is trained, then promote it back to the harder conversion.

Step 3: audience. Go broad and let Advantage Plus do the work — modern Meta targets via creative far better than via interest lists. Step 4: creative. Launch 5–10 ad variants, run 7 days untouched (don't disturb the learning phase), then keep the top 2–3. Always produce both 1:1 (feed) and 9:16 (Reels/Stories) versions. Step 5: measurement. Track CPC, CPA, ROAS, CTR, conversion rate; tag every URL with UTM parameters so Google Analytics knows the source.

Key sub-lessons from Q&A: Meta is passive consumption, Google is active intent — they complement, don't replace. For local services (Kevin's fencing example), awareness on Meta lifts branded search later. For retargeting, the Meta Pixel fires whenever someone lands on your site — regardless of whether they came from Google ads, email, or direct, so a Google → Meta retargeting cycle works (Muzammil's question). Conversions API (CAPI) is strongly recommended for tracking accuracy under privacy restrictions; Shopify/WordPress/Squarespace now have native integrations. From Topic 9's stuck-points, Nick added: new Meta features get better CPMs for ~12 months — be early. Lead-form quality dies without automated follow-up — Matt's "12:19 AM half-drunk filling the form" leads were low-intent; the fix is automated sequences, not better targeting.

Tools

• Meta Business Manager — required. Check at business.facebook.com.
• Meta Pixel — site-side tracking; native integrations on Shopify, WordPress, Squarespace.
• Conversions API (CAPI) — server-side complement to the Pixel; strongly recommended for accuracy post-iOS-14.
• Advantage Plus — Meta's AI-driven audience and placement layer.
• Facebook Ads Library — competitor research; look for old launch dates (sustained = winning).
• Google Analytics + UTM parameters — required for attribution.
• Vapi / automated follow-up (cross-reference Topic 10) — for converting Meta leads who arrived low-intent.

Steps Discussed

1. Prerequisites: Meta Business Manager + Facebook Page + Instagram (connected via Page Settings) + Pixel installed + Conversions API connected + payment method on file.
2. Understand the taxonomy: Ad Account → Campaign (objective + total budget) → Ad Set (audience, schedule, placements) → Ad (creative + landing page). Same shape applies to TikTok, Google.
3. Step 1 — Objective:
   1. Pick from Sales / Leads / Awareness / Traffic / Engagement / App.
   2. Map your business goal to the objective: market penetration → Awareness; video views → Engagement; website traffic → Traffic; messages → Engagement; calls → Traffic; e-comm → Sales; B2B forms → Leads.
4. Step 2 — Budget:
   1. Choose Daily (evergreen, ad-set rotates) or Lifetime (date-bounded promos like Black Friday).
   2. Apply the 50-conversions-per-week rule: weekly budget = target CPA × 50.
   3. If that exceeds reality, drop to an upstream event (add-to-cart, signup, key page view) so cost-per-conversion fits the math.
   4. Minimum ~$20/day per ad set.
5. Step 3 — Audience:
   1. Go broad. Use Advantage Plus audience by default.
   2. Don't dilute spend by splitting one $80 ad set into two $50 + $30 ad sets (from Topic 2's live diagnostic).
   3. Carve out ~15–20% of budget as a "testing" ad set off to the side for experimental targeting.
6. Step 4 — Creative:
   1. Launch 5–10 creative variants at once.
   2. Run for 7 days untouched — do not pause, do not edit (kills the learning phase).
   3. Produce both 1:1 (feed) and 9:16 (Reels/Stories) versions for every creative.
   4. After 7 days, kill the bottom; keep the top 2–3.
   5. Set up a testing budget alongside the core campaign to trial new creative without disturbing winners.
   6. Add a retargeting layer once the prospecting funnel is dialed (Pixel-based, retargets anyone who hit the site regardless of source).
7. Step 5 — Measurement:
   1. Track at the ad-set level: impressions, reach, frequency, CTR, CPC, CPA, ROAS, conversion rate.
   2. Tag every destination URL with UTM parameters for Google Analytics attribution.
   3. Watch for cross-channel signals — ChatGPT and Perplexity now appear in Google Analytics sources.
8. Competitor research: Facebook Ads Library → competitor page → About → Page Transparency → See All Ads. Look at launch dates — nobody keeps a losing ad live for 6 months, so the oldest active ads are the winners.
9. Lead-flow fix (Matt's pattern): if leads arrive low-intent, the targeting isn't the problem — the follow-up is. Add an automated sequence (Vapi voice call + SMS + email) so 200 leads → automatic 50 qualified → manual sales touch on the survivors.

Resources

• Prompts Page

Key Takeaways

• 50 conversions/week is the algorithm's diet. Without that throughput, Meta never exits the learning phase.
• If you can't afford the target conversion, drop one level up. Train on add-to-carts or sign-ups first; promote to purchases later.
• Don't split the dollar across ad sets. One $80 ad set beats two $50 + $30 ad sets every time.
• Target via creative, not interests. Advantage Plus does the targeting; your job is making the creative do the filtering.
• 5–10 creative variants → 7 days untouched → keep top 2–3. That's the iteration loop.
• 1:1 AND 9:16, always. Don't skip Reels-aspect — that's where the cheap impressions live.
• Meta = passive, Google = intent. Awareness on Meta lifts later branded Google search; they're additive.
• Older ad launch dates = winners. Competitor research via Ads Library is about longevity, not creative copying.
• Pixel fires from any traffic source. Meta retargeting captures site visitors regardless of whether they came from Google ads, email, or direct.
• Automate follow-up or Meta leads die. A weak offer + low-intent leads + manual call list = 0% conversion. Automation flips it.
• New Meta features get cheaper CPMs for ~12 months. Be early.

Questions to Explore Further

• For service businesses with high CPAs (e.g. $500+ legal/B2B leads), is the upstream-event trick (train on add-to-cart) still effective, or does the gap between proxy and real-conversion erode quality past a certain threshold?
• How long should you let an Advantage Plus campaign run before deciding it's not working — Meta's official guidance is 7 days untouched, but is that enough at lower budgets?
• When the Pixel and CAPI report different numbers, which one should you trust as the source of truth for ROAS — and how do you reconcile?
• For local-service businesses (Kevin's fencing case), what's the right cadence between Meta awareness, paid search, and the automated follow-up loop — is there a measurable lift from running all three vs. just paid search?
• How early is "early" for new Meta features? Should you allocate 20% of budget to whatever Meta released last quarter just to harvest cheaper CPMs while they last?

Steps I Need to Do on My End

No homework assigned in this session.

Meeting Anatomy

Stuck-points round (Q1 rocks confirmation across attendees, Matthew on Nick Lopez Pixel follow-ups, Dave on GoHighLevel agency migration, group thumbs-up vote on a future GoHighLevel 101 training, LinkedIn-ads creative-review request) → handoff to Module 17 → Rishab teaches the LOCK framework with Scott adding the "K" mid-session → live walk-through of Seamless AI for list scraping → live demo of the X100 Hormozi Grand Slam Offer Builder → cold email copy generation in ChatGPT → deliverability stack (warmup, validation, sender setup) → assignment + Q&A.

Summary/Recap

The diagnostic framework when email isn't working: L.O.C.K. — List, Offer, Creative, Knowledge/tech. Diagnosed in that order — if the list is wrong, no offer/creative can save the campaign; if the offer is weak, great creative on a great list still flops. Scott added the K mid-session: deliverability tech (warmup, validation, ESP choice) is often the silent killer.

List — three acquisition paths: earned (lead magnets, opt-ins; best quality), scraped (Seamless AI, Apollo, ZoomInfo; great for cold B2B, niche-able), or bought (gray area, beware). Earned beats everything; scraped is the realistic starting point for someone without a list. For Dream 100 outreach, scrape by ideal-customer firmographics, role, industry, and revenue band.

Offer — use the X100 Hormozi Grand Slam Offer Builder prompt. Apply the Hormozi value equation: maximize Dream Outcome × Perceived Likelihood; minimize Time Delay × Effort/Sacrifice. Name the offer specifically — not "free strategy call" but something like "Master Your Next Difficult Conversation Strategy Call." Naming gives the offer teeth and converts dramatically better.

Creative — cold email is plain text. No links, no images, no banners, no HTML signatures, no logos. ESPs flag stranger-emails with links as spam instantly. The first email's only CTA is "reply and I'll send you X" — a free guide, the case study, the spreadsheet. The actual link lands in the reply. Always include a P.S. (the second-most-read part of an email after the subject line).

Knowledge/Tech — warm up the sending domain/inbox for 2–4 weeks with a service like Warmy or Warm before sending volume. Validate every scraped list through ZeroBounce or Hunter.io to strip dead addresses (bounces destroy sender reputation). Use a cold-email-specific tool like Lemlist or Instantly for outbound (DON'T send cold from Mailchimp / ActiveCampaign — they ban accounts). Use GoHighLevel / ActiveCampaign / Mailchimp for warm/earned-list nurture sequences. SPF + DKIM + DMARC records configured on your domain. Sending volume ramp: start < 50/day per inbox, scale slowly to ~200/day max per inbox.

Compliance warning: email cold-outreach is legal in most US/EU jurisdictions with unsubscribe + identification; unsolicited SMS and AI voice calls are lawsuit risk. Get explicit opt-in before phone/SMS outreach. Vapi auto-calls are fine for inbound (you call them after they fill a form); they are NOT fine for outbound cold.

Tools

• Seamless AI — primary list scraper demoed in the session. Niche down by ICP firmographics + role + industry.
• Apollo / ZoomInfo — alternatives to Seamless. Different price points and database coverage.
• X100 Hormozi Grand Slam Offer Builder prompt — on the prompts page; generates a named, value-equation-aligned offer.
• ChatGPT (in your ICP/messaging thread) — write the actual cold email copy and follow-up sequence.
• Warmy / Warm — inbox warmup tools (2–4 weeks before scale).
• ZeroBounce / Hunter.io — list validation; strip dead addresses before sending.
• Lemlist / Instantly — cold email sending tools (purpose-built for outbound).
• GoHighLevel / ActiveCampaign / Mailchimp — for warm/earned-list nurture (NOT cold).
• SPF / DKIM / DMARC — domain auth records; required for deliverability.

Steps Discussed

1. Diagnose with LOCK — when an existing email program is failing, check List first, then Offer, then Creative, then Tech. Order matters.
2. Build/acquire the list:
   1. If you have a lead magnet, route it to a landing page + opt-in form (earned). Best quality.
   2. Otherwise, open Seamless AI. Filter by industry, role, company size, revenue band, geography.
   3. Export the scraped list as CSV.
   4. Run the CSV through ZeroBounce or Hunter.io to remove invalid emails.
3. Craft the offer using the Hormozi Grand Slam prompt:
   1. Open your ICP/messaging chat thread.
   2. Paste the X100 Hormozi Grand Slam Offer Builder prompt.
   3. It outputs an offer using the value equation; name it specifically (not "free call" — name what the call delivers).
4. Write the email creative:
   1. Plain text only. No links, no images, no HTML signature.
   2. Personalize line 1 (name, company, or specific recent event).
   3. State the problem you solve in one sentence (pulled from Top 10 objections / pain points).
   4. Introduce the named offer in one sentence.
   5. CTA = "Reply and I'll send you [the asset]." The link lives in your reply.
   6. Always add a P.S. — second most-read line in any email.
5. Build the deliverability stack:
   1. Buy a separate sending domain (don't burn your main one).
   2. Configure SPF, DKIM, DMARC.
   3. Connect inbox to Warmy/Warm and warm up for 2–4 weeks before sending real volume.
   4. Send cold campaigns from Lemlist or Instantly only. Don't use Mailchimp / ActiveCampaign for cold.
   5. Ramp volume slowly: < 50/day per inbox to start, scale to ~200/day max per inbox.
6. Sequence: send a 5–7 touch cadence (initial + 4–6 follow-ups spaced 2–4 days apart). Keep each follow-up plain text, no links, varied subject line. Track replies, not opens (opens are inflated by ESPs prefetching).
7. For warm lists: switch to a soap-opera Smart 90 sequence (from Topic 6) in GoHighLevel / ActiveCampaign — those tools handle drip + automation cleanly.
8. Don't do SMS or AI voice cold outreach without opt-in. Vapi for inbound only.

Resources

• Seamless AI

Key Takeaways

• List beats Offer beats Creative. The LOCK diagnostic order is non-negotiable.
• K = Knowledge/Tech. The silent killer is deliverability — warmup, validation, sending tool choice, domain auth.
• No links in the first cold email. ESPs flag stranger-emails with links as spam immediately.
• Name your offer specifically. "Master Your Next Difficult Conversation" converts; "Free Strategy Call" doesn't.
• Apply the Hormozi value equation. Maximize Dream Outcome × Perceived Likelihood / minimize Time Delay × Effort.
• P.S. is the second-most-read line. Always include one; restate the offer or add urgency.
• Don't send cold from Mailchimp / ActiveCampaign. They'll ban your account. Use Lemlist/Instantly for cold; warm tools for nurture.
• Validate every list before sending. Bouncing addresses destroys sender reputation faster than anything else.
• Warmup is 2–4 weeks of patience. Skipping warmup = straight to spam.
• Email = legal with care. SMS / voice cold = lawsuit risk. Get opt-in first for phone-based outreach.
• GoHighLevel consolidates ESP + CRM + funnels + courses. Worth a dedicated training (group voted in favor).

Questions to Explore Further

• For agencies running cold email at scale across 10+ clients, what's the right per-client domain/inbox architecture — one inbox per client, or one pooled sending infrastructure with per-client domains?
• How do reply rates change when you A/B test specifically-named offers ("Master Your Next Difficult Conversation") vs. generic ones ("Free Strategy Call") — is the lift consistent across industries?
• If a client wants to migrate from Kajabi to GoHighLevel, what specifically fails to transfer (members, drip sequences) and what's the right manual-rebuild workflow?
• For warm-list re-engagement campaigns (subscribers who haven't opened in 6+ months), does the LOCK order still apply, or should you lead with Creative (subject line) since List is already "yours"?
• What's the realistic reply-rate floor for a well-warmed cold sequence with a named Hormozi offer — 1%, 3%, 5%? — and how does that vary by industry vertical?

Steps I Need to Do on My End

1. Create an email campaign end-to-end this week. Explicit assignment: "scrape a list, create an offer, generate an email campaign, and deploy that email campaign."
2. Build your irresistible offer using the X100 Hormozi Grand Slam Offer Builder prompt — name it specifically, not generically.
3. Scrape your list with Seamless AI (or run a lead magnet on a landing page to start an earned list).
4. Validate the list (ZeroBounce / Hunter.io) before sending.
5. Warm your sending inbox for 2–4 weeks before any volume; send cold only from Lemlist/Instantly, never from Mailchimp/ActiveCampaign.
6. Write plain-text emails with no links, a named offer, a reply-driven CTA, and a P.S.