# Craig Perry Growth Thesis **Analyst:** Content Trainer Pro | **Date:** 2026-07-02 **Subject:** @profound_ideas on YouTube | 22.6K subs | 13 long-form videos **Purpose:** Explain why he is growing fast. Extract what transfers to Everyday AI Club. --- ## DAN KOE IDENTITY CONFIRMATION CONFIRMED. The interviewer in the Eden video "How To Write About Your Unique Interests And Still Go Viral" is Dan Koe. The Substack post of that same interview (Jul 1, 2026) is co-authored by "DAN KOE" at letters.thedankoe.com, listed as "future/proof." Craig's Substack recommendations page links to "future/proof by DAN KOE." His last name never appears in the interview transcript, but his identity is confirmed via cross-referencing the Substack publication byline and Craig's recommendation list. Anthony's assumption was correct. --- ## SECTION 1: FACTS (THE NUMBER PATTERN) ### Channel Snapshot (as of 2026-07-02) - Subscribers: 22.6K - Total videos: 13 - Format: 100% long-form, 20 to 50 minutes each. Zero shorts. - Substack: Profound Ideas, 44,000+ subscribers (scraped live, ideas.profoundideas.com) - Substack rank: #10 Rising in Philosophy - Publishing start: approximately May 2025 (confirmed in interview: "about a year now since May last year" [3:00]) - Total essays written: approximately 80 long-form posts [INTERVIEW-EXTRACT.md, 1:15:16] - YouTube growth rate: "a couple hundred subscribers a day" in the last 6 months at time of recording [INTERVIEW-EXTRACT.md, 4:02] ### View-to-Subscriber Ratios (the 18x anomaly explained) | Video | Views | V/S Ratio | Notes | |---|---|---|---| | How To Become Dangerously Self-Educated | 404K | 17.9x | 18x outlier. Search-captured via hook title. | | How To Remember Everything You Read | 149K | 6.6x | Driven by returning audience + search. | | More Educated Than 99% of People | 96K | 4.3x | TRANSCRIPT MISSING, pattern inferred from title alone. | | How to Think On Paper | 59K | 2.6x | PKM/Obsidian audience crossover. | | How to Become Dangerously Articulate | 50K | 2.2x | Strong power hook, philosophy mid loses some. | | If you are ambitious but stuck... | 24K | 1.1x | Generic title, least distinctive packaging. | | How To Increase your Intelligence | 19K | 0.8x | TRANSCRIPT MISSING. | | Manage Multiple Interests | 10K | 0.4x | Below sub count. Early video. | | 11 tools for reversing brain rot | 6.9K | 0.3x | Weekly cadence video. | | Feel Lost Do This For 90 Days | 6.2K | 0.3x | Strong content, new cadence, small window. | | Most Important High-Value Skills | 4K | 0.2x | Best structural content, weakest title match. | | How to achieve anything effortlessly | 3.5K | 0.2x | TRANSCRIPT MISSING. | | Getting Unstuck | 1.5K | 0.07x | Lowest. Over-competitive title territory. | **Key pattern:** The top 3 videos (404K, 149K, 96K) all use broad universally-felt desires as titles. The bottom 5 are either too niche in their title promise or too competitive. Content quality does NOT explain the gap. Packaging does. ### Publishing Pattern (batch to weekly shift) - Apr 2: 3 videos dropped same day - May 2: 2 videos dropped same day - Jun 2: 3 videos dropped same day - Jun 4, 11, 18, 24: weekly individual drops began This is not accidental. Craig confirms in the interview he writes one essay per week [3:00]. The early batch drops were pre-produced Substack essays converted to YouTube en masse. The shift to weekly individual drops in June coincides with his YouTube becoming a primary focus: "I'm kind of shifting towards YouTube" [4:16]. The batch-to-weekly shift is a platform maturation signal, not a pivot. ### The 18x Anomaly 404K views on a 22.6K-sub channel is not a fluke. It is the result of a video performing on YouTube search and browse for a non-subscriber audience. The self-education video has the channel's strongest hook mechanism (pattern-recognition game), the channel's most broadly felt desire ("self-educated"), and a single power adjective in the title ("Dangerously") that makes the concept feel transgressive. The hook captured cold traffic. The content retained them long enough for YouTube to keep amplifying it. Craig references "200,000 views on the self-education video" in the "ambitious but stuck" video (channel-mid.json, video boJpqVqEBcw) at [21:58], not in the Eden interview. In that video Craig says "my self-education video I put up a couple weeks ago. It's at 200,000 views, which is crazy." This suggests the 404K count reflects continued algorithm distribution well after that recording. --- ## SECTION 2: MECHANISMS (EVIDENCE-BACKED) ### Mechanism 1: The Broad Door, Niche Interior Rule Craig names this directly: "get people through the door first and then you can explore the house" [INTERVIEW-EXTRACT.md, 13:54]. Evidence: Compare the top and bottom performers by title type. - Top performer title: "How To Become Dangerously Self-Educated" (universal desire: learn without formal school) - Second: "How To Remember Everything You Read" (universal pain: forgetting what you read) - Bottom performer: "Getting Unstuck Is Hard Until You Follow This Exact System" (self-help cliche, crowded search category) - Second lowest: "Most Important High-Value Skills to Learn" (generic SEO title, mismatched to actual content depth) The 404K video is NOT about self-education as Craig defines it. It is packaged as self-education but the interior is about the Archivist vs Architect learning taxonomy, a specific named framework Craig invented. The broad title gets cold viewers in the door. The specific original interior retains them and earns shares. The inverse failure is visible in "Most Important High-Value Skills" (4K views). Craig says in its transcript: "I've written about 60 essays in the past year" [channel-recent.json, 1:40]. That video contains the most technically sophisticated content he has produced (encoding, retrieval, chunking, chain-linked skill architecture). But the title "Most Important High-Value Skills to Learn" is a generic SEO title that attracts the wrong viewer (someone wanting a listicle who bounces when they get a philosophy lecture). ### Mechanism 2: Named Concepts That Stick Every high-performing video coins at least one named taxonomy the viewer can take away and use. Evidence: - Self-educated video: "Archivist vs Architect" [channel-top.json, spine 2:58] - Remember everything video: Archivist vs Architect again, plus "hypercorrection effect" [channel-top.json, devices] - Think on paper video: "Mode 1 vs Mode 2," "cognitive offloading," "relationship mapping" [channel-mid.json, devices] - Articulate video: The Sisyphus frame as a reusable life metaphor [channel-mid.json, spine 15:08] - Ambitious but stuck video: "Synthesist vs Generalist" reframe [channel-mid.json, devices] Named concepts do three things simultaneously. They give the viewer identity vocabulary ("I am an Archivist and I need to become an Architect"). They make the idea repeatable across videos (Archivist/Architect appears in at least three separate essays). And they are the actual viral unit: a named concept can be screenshotted, quoted, and debated in comments in a way that a general idea cannot. Craig explains the philosophy behind this: "your mind is your niche" [INTERVIEW-EXTRACT.md, 5:44]. The named concept is how the niche mind announces itself. ### Mechanism 3: Shame Confession as Opening Hook Craig's most reliable hook device is not a promise and not a question. It is a personal failure story told in a specific number. Evidence: - Self-educated video: "I set the grades record at my secondary school in 2022 then calls it worthless" [channel-top.json, spine 1:02]. Confession: I beat the system and it meant nothing. - Remember everything video: "This time last year, I wrote 70,000 words worth of linear notes in Obsidian from three books." [channel-top.json, hook_30s, 0:30]. Specific books named (Russell, Plato, Aristotle). Cost reframed: "That is 35 two-thousand-word newsletters I could have written." [0:44]. - Think on paper video: The same 70,000-word confession again [channel-mid.json, hook_30s, 00:16]. Used as a recurring device because it is autobiographically true and perpetually relatable. - Feel Lost video: Full emotional disclosure about his mother dying of cancer when he was 12, used structurally at the ark-building argument moment [channel-recent.json, spine 9:07]. Not an opener but a mid-essay gravity anchor. The mechanism: Craig does not confess weakness as vulnerability theater. He confesses a specific measurable mistake ("70,000 words") and then converts it into an opportunity cost the viewer can feel ("35 newsletters I could have written"). This is a direct-response copywriting move applied to YouTube hooks. It works because it pre-answers the objection "who is this person to tell me how to learn" before the objection forms. ### Mechanism 4: Dan Koe as Accelerator, Not Cause CONFIRMED via Substack. Dan Koe's "future/proof" newsletter (letters.thedankoe.com) cross-posted the Craig interview on July 1, 2026, the same day it appeared on Craig's Substack. Craig's Substack recommendations include Dan Koe's newsletter. Craig states in the interview that he watched a Dan Koe video "about a year and a half ago" about writing online that changed his direction [Dan Koe interview transcript, 2:45]. This is a network amplification mechanism, not a growth cause. Koe's newsletter has hundreds of thousands of subscribers. A single cross-post of a 1:20 interview is a meaningful distribution event. But Craig's Substack was already at 44K+ before this cross-post. The pre-existing subscriber base was built without Koe's direct promotion. Koe is an accelerant on top of organic compounding, not the engine of it. ### Mechanism 5: Essay-First, Video-Second (Zero Production Tax) Craig confirms in the interview: three phases, the last of which is "write it by hand" and then read the finished essay to camera [INTERVIEW-EXTRACT.md, 11:01]. The YouTube video IS the essay, performed live. There is no separate scripting step. There is no visual production beyond the talking-head camera shot. No b-roll. No motion graphics. No editor required beyond a basic cut. Evidence in the data: every video transcript is a flowing essay with a clear argumentative arc. The "live demonstrations" (mind maps on screen, printed articles annotated) are the only visual elements added. These cost zero extra time because Craig uses them in his own learning process anyway. The production tax is near zero, which means the compounding is nearly frictionless once the essay is written. ### Mechanism 6: The Substack-YouTube Flywheel Craig's channel cannot be understood as a standalone YouTube channel. It is the second-stage output of a Substack that already had 44,000 subscribers before most of these videos were published. The Substack essay validates and pre-populates demand for the YouTube video. The YouTube video brings cold traffic back to the Substack. Evidence from the "Remember Everything" video CTA: "subscribe to Substack free tier for printable bookmarks plus a secret discount link to the paid tier in the welcome email" [channel-top.json, cta]. The video free tier drives email capture. The email captures funnel into a paid product. The paid product funds more content. The Substack is the business. YouTube is the top of the funnel. Evidence from the Substack scrape: "How To Remember Everything You Read" has 26,440 likes and 4,455 reads on Substack (scraped live). "How To Understand More of What You Read" has 8,949 likes and 1,603 reads. These are massive engagement numbers for a Substack. The essay performance on Substack predicts the video performance on YouTube because they are the same content, same audience, different packaging. ### Mechanism 7: Validation Filter as Research Protocol ATTRIBUTION CORRECTION: The specific validation filter numbers (300-1,000 Substack likes, 100K YouTube views) were stated by the INTERVIEWER at [26:28-26:48], not Craig. The transcript shows a `>>` speaker change at [26:28] before these numbers are given. Craig uses the concept of "validated ideas" throughout the interview and endorses high-engagement filters implicitly, but these specific thresholds are the interviewer's stated practice. The mechanism is still real and transferable, but it should not be cited as Craig's explicit personal rule. The validation filter concept: consume only validated content (high engagement) to train your taste. Endorsed by both Craig and the interviewer in the interview. Specific thresholds per the interviewer: Substack posts must have 300 to 1,000 likes, YouTube videos must have 100,000-plus views [26:41, 26:48]. This trains taste to match what the market already rewards. --- ## SECTION 3: HYPOTHESES (LABELED GUESSES, RANKED BY CONFIDENCE) ### H1: The 404K video was a search event, not an audience event (confidence: high) The self-education video outperforms his subscriber count by 18x. His subscriber base at the time of publication was much smaller than 22.6K (he started in May 2025; the April 2 batch was his YouTube launch). A 404K view count on a near-zero subscriber base means YouTube's algorithm distributed it to non-subscribers via search and browse. The title "How To Become Dangerously Self-Educated" has high search intent (people search "how to self-educate," "self-education," "learn on your own") plus the "Dangerously" modifier makes it distinctive in the browse feed. UNVERIFIED: actual YouTube search impression data is not available. ### H2: The Substack subscriber base was the real channel subscriber list before YouTube (confidence: high) Craig's Substack launched approximately May 2025. His YouTube batch drop began April 2026, almost a year later. His Substack had a large pre-existing audience by then. When he uploaded videos, he likely mentioned them to his newsletter. A 44,000-person email list watching your first YouTube video batch would give the videos a massive engagement spike in the first 48 hours, which is the signal YouTube uses to decide how widely to distribute new content. The "launch day" views on the April 2 batch were probably seeded by the newsletter. UNVERIFIED: no Substack send data for that date available. ### H3: The shift to weekly drops in June signals YouTube is now the primary platform (confidence: medium) The transition from batch drops (Apr 2: 3 videos, May 2: 2, Jun 2: 3) to individual weekly drops (Jun 4, 11, 18, 24) matches Craig's interview statement: "I'm kind of shifting towards YouTube now" [4:16]. Weekly individual videos are better for YouTube's algorithm than batch drops because each video gets its own 48-hour discovery window. The recent videos (4K, 1.5K, 6.2K views) are underperforming relative to the earlier batch, likely because Craig is still building his YouTube subscriber base and the weekly cadence has not yet fully replaced the Substack email-seeding effect. UNVERIFIED: view trajectory over the next 30 days would confirm or deny. ### H4: "Dangerously" as a title adjective modifier is his highest-converting packaging word (confidence: medium-high) Two of his top three videos use the word "Dangerously" (Dangerously Self-Educated: 404K; Dangerously Articulate: 50K). The "99% of People" video (96K) uses a different but equally transgressive frame (elite percentile claim). His lowest performers use neutral title language (Getting Unstuck, Ambitious But Stuck). UNVERIFIED: A/B test data does not exist, but the pattern is consistent across all 13 videos. ### H5: The "writing about ideas, not a niche" positioning is itself a novelty signal (confidence: medium) On YouTube, the self-improvement space is crowded. "How to study," "how to learn," "productivity" are saturated categories. Craig's positioning ("your mind is your niche" [INTERVIEW-EXTRACT.md, 5:44], "if an idea is profound to me I will write about it, that's my only rule" [5:52]) is an unusual signal in a space full of topically siloed channels. His content spans learning science, philosophy (Camus, Epictetus, Aristotle), articulation, self-education, and writing. A channel that crosses these in a unified essay voice is genuinely distinctive in the browse feed. UNVERIFIED: no comparative data on similar channels' performance in the same period. --- ## SECTION 4: THE TRANSFER LIST **Format:** Each tactic states what it is, what Craig does that proves it, and which of Anthony's existing principles it extends (Ramonov 1-21) or if it is NEW. --- ### T1: Write the essay first. Read it on camera. Skip the script. **What Craig does:** He writes a long-form essay on Substack, then reads it to camera for the YouTube video. The essay is the script. No separate scripting step. No teleprompter. He reads from what he wrote, which is already in his voice because he wrote it before any video planning happened. **Why it matters:** Anthony's Sabrina engine (batch filming, Ramonov system) assumes a separate scripting phase. If Anthony's essays for the Everyday AI Club newsletter became the YouTube scripts word-for-word, the production overhead drops to near zero. He is already writing long-form content. The camera pickup is incremental, not additive. **Extends:** Ramonov 3 (batch filming). One batch session per week filming the essay that was already written for the newsletter reduces the time-per-video from hours to 30-60 minutes of camera time. **Executable next week:** Write one Everyday AI Club newsletter essay. Sit down with a camera. Read it. Upload. Do not write a separate script. --- ### T2: Open with a shame confession in a specific number. **What Craig does:** "This time last year, I wrote 70,000 words worth of linear notes in Obsidian from three books." [channel-top.json, hook_30s, 0:30]. The number is specific, the failure is personal, the cost is calculated ("35 newsletters I could have written"). **Why it matters:** The confession hook is more reliable than a curiosity gap hook or a promise hook because it pre-answers the credibility objection before the viewer forms it. For Anthony's AI education lane, the analogous confession is: "I spent three months using ChatGPT like Google Search. I asked it questions and forgot the answers five minutes later. That was wrong and I can show you why." **Extends:** Ramonov 1 (copy proven hooks). The shame confession with a specific number is a proven hook pattern. This is not about copying someone else's hook. It is about applying the confession structure to Anthony's actual experience with AI tools. **Executable next week:** Write one hook where you confess a specific mistake you made with AI tools (wrong tool, wrong use, wasted time) and give it a number. "I spent 14 hours building an AI agent that did what one Claude prompt does in 30 seconds." Then pivot to the lesson. --- ### T3: Coin one named concept per video. Make it a self-diagnostic. **What Craig does:** Every high-performing video coins a named taxonomy the viewer can immediately apply to themselves. "Archivist vs Architect" works because the viewer can diagnose which one they are in 5 seconds. The named concept is the shareable unit of the video. **Why it matters:** Anthony's Everyday AI Club teaches AI tools to non-technical people. Most AI tool education is instruction-mode (here is how to use this feature). Named concepts are identity-mode (here is who you are in relation to this technology). "The AI Hoarder vs the AI Architect" (you collect tools vs you build systems) is directly analogous to Craig's taxonomy. Identity concepts are shared. Instructions are not. **Extends:** NEW. The Ramonov system describes hook and format selection but does not specify named-concept creation as a content building block. This is an addition. **Executable next week:** In your next long-form video, before you explain anything, ask: "What are the two types of people in my audience and what are they doing wrong?" Name both types. Give them a label the viewer can claim. Build the rest of the video around the gap between the two. --- ### T4: Apply the Broad Door rule to every title. **What Craig does:** Title = broadest possible framing of the desire. Body = specific angle Craig personally cares about. "Get people through the door first and then you can explore the house." [INTERVIEW-EXTRACT.md, 13:54]. **Why it matters:** Anthony's existing titles for the Everyday AI Club are likely tool-first: "How to Use Claude Projects," "Claude vs ChatGPT," "Build an AI Agent with These Tools." These are niche doors. The broad door equivalent: "How to Stop Forgetting What AI Tells You" (broad pain), "Why Smart People Get Worse Results From AI" (percentile frame), "How to Think 10x Faster" (broad desire, AI is the interior mechanism). The specific tool lives inside the essay. The broad desire gets cold traffic through the door. **Extends:** Ramonov 1 (copy proven hooks). The hook is the title for YouTube. Broad title = broad hook reach. **Executable next week:** Take your next planned video and ask: "What does a non-AI-person want that my AI solution gives them?" Write five broad-door titles aimed at that desire. Pick the one that would make someone who has never heard of Claude click. --- ### T5: Add one power adjective to your title. Make the skill feel dangerous or transgressive. **What Craig does:** "Dangerously" appears in his two top titles. "More Educated Than 99% of People" is the percentile-transgression version. Both frames make learning feel like gaining power that the establishment does not want you to have. **Why it matters:** Anthony's AI education lane has a natural power adjective available: "dangerously," "unfair," "illegal" (used carefully), "that your boss does not know," "that makes you harder to replace." The AI topic is already inherently about power dynamics (who has access, who does not). The adjective makes the implicit power dynamic explicit in the title. **Extends:** Ramonov 1 (copy proven hooks). This is a single-word implementation of the hook principle. **Executable next week:** In your next batch of five title options per video, write at least one version with a power adjective. Test against the neutral title to see which feels more clickable to Anthony's honest judgment as a viewer. --- ### T6: Build one named concept that recurs across 3+ videos. Make it your signature. **What Craig does:** "Archivist vs Architect" appears in the self-education video, the remember everything video, and the think on paper video. It is not repeated as lazy filler. It is a foundational taxonomy that his whole worldview builds on, and each video deepens it from a different angle. **Why it matters:** The signature concept is the thing your audience uses to explain you to other people. "Craig Perry is the Archivist/Architect guy." If Anthony's signature concept becomes "The AI Thinker vs the AI Tool Collector" or "The AI System vs the AI Trick," it gives returning viewers a framework to organize everything else they learn from the channel. It also makes new viewers who encounter video 7 want to go back and watch video 1 where the concept was introduced. **Extends:** Ramonov 3 (batch filming). Shooting multiple videos that build on one concept is a natural batch. Film the concept introduction first, then the two or three applications. **Executable next week:** Identify the one idea about AI that Anthony genuinely believes that most of his audience does not. Name both sides of it. Test whether it is self-diagnostic (can a viewer tell which side they are on in 5 seconds?). If yes, that is the signature concept. Introduce it in the next long-form video. --- ### T7: Use the validation filter before building your swipe file. **What Craig does (attribution note):** The specific threshold numbers below were stated by the INTERVIEWER, not Craig, at [26:41-26:48] in the Eden interview. Craig uses the concept of validated ideas throughout and implicitly endorses high thresholds, but did not personally state these exact numbers. The tactic is still real and endorsed in the conversation. Substack: 300 to 1,000 likes minimum. YouTube: 100,000-plus views minimum. Below that, it does not enter the swipe file regardless of how interesting it seems personally. **Why it matters:** Most creators build swipe files from content they find interesting, which trains their taste to match their current level, not the level they are trying to reach. Anthony's Ramonov system emphasizes hook copying from proven content (Ramonov 1). The validation filter is the upstream discipline that ensures the swipe file contains only proven hooks, not just interesting ones. **Extends:** Ramonov 1 (copy proven hooks). The filter is how you ensure you are copying from the right source tier. **Executable next week:** Pull up your current swipe file. Remove any video or post that does not clear 100K YouTube views or 300+ Substack likes. What is left is your actual research pool. Add a filter rule to your content system: nothing enters the swipe file below those thresholds. --- ### T8: Read one essay sentence by sentence once per week. Extract one craft principle. Apply it that week. **What Craig does:** "I like to read a paragraph or a section from a post and go, why did that sound so persuasive? Really just trying to understand the craft." [INTERVIEW-EXTRACT.md, 1:13:42]. One post per week, printed out, analyzed at the sentence and mechanism level. Extract one principle. Apply it in that week's essay. Repeat. He has done this across approximately 80 essays. "You only need one permanent 1% improvement every week." [1:15:03]. **Why it matters:** Anthony's Ramonov system describes production cadence and batch filming. The craft improvement loop is missing from the production system. Craig's 1% compound principle is what separates his writing quality from channels that film more but do not improve. For Everyday AI Club, one session per week reading a top-performing AI education script sentence by sentence and extracting one mechanism is the equivalent practice. **Extends:** NEW. The Ramonov principles cover velocity, batch production, and hook selection but do not include a weekly craft improvement loop. This is a genuine addition to the system. **Executable next week:** Print one top-performing transcript from Anthony's target (it could be Craig Perry's own 404K video). Read it paragraph by paragraph. Ask "why did that land?" after each section. Write one sentence at the end describing what you learned. Apply it in that week's video. --- ### T9: Use Problem, Insight, Solution as your default essay architecture. **What Craig does:** "I break it down into the problem, the insight, and then the solution." [INTERVIEW-EXTRACT.md, 42:04]. This comes after the five-element hook (pattern interrupt, curiosity loop, visual cue, promise, plan) [37:30]. The essay body is always problem first (what is misunderstood), insight second (why it actually works the way it does), solution third (what to do with that insight). **Why it matters:** Anthony's Everyday AI Club content likely follows a tutorial format: here is the tool, here is how to use it, here are the steps. That is solution-only content. It answers "what" and "how" but skips "why people get this wrong" (problem) and "the actual mechanism" (insight). Craig's format includes two steps that tutorial content typically omits. Those two steps are what make the audience feel understood and stay for the solution. **Extends:** Ramonov 1 (copy proven hooks, proven formats). Problem-Insight-Solution is a format frame, not just a hook pattern. **Executable next week:** Take one upcoming video and rewrite the outline as: (1) What most people get wrong about this AI topic (2 minutes). (2) The actual mechanism that explains why (5 minutes). (3) What to do with that insight, step by step (10 minutes). Compare it to your current outline and see which feels more like something you would watch. --- ### T10: Build the essay around one concrete worked example, not a list of principles. **What Craig does:** The 404K video uses the guitar analogy throughout: "The archivist buys an 8-hour theory course and still cannot play a C chord; the architect Googles the C chord and plays it in a song." [channel-top.json, devices]. The 149K video shows a live mind map of The Myth of Sisyphus on screen. The 59K video draws an actual mind map of Nicomachean Ethics in real time. The concrete example is not an illustration of the principle. It IS the lesson. The principle is extracted from the example, not explained before it. **Why it matters:** AI education is dense and abstract (tokens, context windows, prompting techniques, agent workflows). A worked example makes the abstract tangible. "Here is how I used one Claude prompt to do what a 14-step Zapier workflow did before" is more retainable than "AI can automate workflows." For Anthony's Everyday AI Club, the worked example from Anthony's own work is more credible than any hypothetical because Anthony is running Orion AI Co, building KAIRO, and managing real clients. The real example is sitting on his desk every day. **Extends:** Ramonov 21 (build first, create second). The worked example is the thing Anthony built. The essay is the report on what building it taught him. **Extends also:** Ramonov 6 (document the journey). The worked example is the documentation. **Executable next week:** For the next long-form video, pick one real thing Anthony did with AI this week. A client workflow. A tool he built for KAIRO. A prompt that surprised him. Write the video as: here is the problem I had, here is what I tried, here is the specific thing that worked, here is what it tells you about how AI actually works. The worked example from your real life is the video. --- ### T11: CTA architecture: soft early, hard late, always with specific social proof. **What Craig does:** Every video has a soft CTA in the first 5 minutes (mentions the Substack or a guide without pressure), then a hard CTA at the end with specific numbers. The specific social proof in the hard CTA is not a general claim. "25,000 Substack subscribers in nine months via 60 to 90 minutes per day before work" [channel-top.json, spine 14:37]. He says exactly what he did, how long it took, and how many people it produced. **Why it matters:** Anthony's Skool principles already address this (the raw About link, the Skool funnel). The specific social proof format is the upgrade. Not "join thousands of people learning AI" but "I used this exact system to do X in Y days with Z people from zero." The specific proof is more persuasive than the generic proof. **Extends:** Skool principle covering the cover image as the ad and daily personal presence. The specific social proof in the CTA is the daily personal presence made concrete. **Executable next week:** Write one CTA for the Everyday AI Club that includes a specific number from Anthony's actual experience. Not the goal. The result already achieved. "I used Claude to cut my client reporting time from 3 hours to 20 minutes. Here is how I taught 300 non-technical people to do the same thing." If the number does not exist yet, build toward the milestone that will produce it, then use that number the week it is real. --- ### T12: Two platforms only. Do not expand until one is fully compounding. **What Craig does:** Confirmed in the interview: "Only on Substack and YouTube" [3:48]. He had not considered X at the time. He deliberately stayed on two platforms for approximately one year until both were growing before contemplating expansion. **Why it matters:** Anthony is managing Everyday AI Club (Skool), TikTok (@ant the AI guy), YouTube (in development), and the newsletter simultaneously. Craig's signal is that compounding on two platforms to a meaningful level (44K newsletter, 22K YouTube) took a full year of weekly output with no distribution spreading. The implication for Anthony is not to abandon TikTok but to ensure the Ramonov batch system is feeding the right two platforms in order of compounding priority. The newsletter and YouTube are the flywheel. TikTok is awareness. Craig's system is clear on hierarchy. Anthony's system should be too. **Extends:** Ramonov 3 (batch filming, cadence discipline). Hierarchy of platform investment is implied by the batch system but not stated. **Executable next week:** Write down the priority order of Anthony's platforms. Which one is the subscription engine? Which one is the search engine? Which one is the awareness engine? Ensure the batch production system allocates filming time in that order. --- ## APPENDIX: WHAT IS NOT TRANSFERABLE (FAILURE MODES ADDRESSED) ### What Craig is doing that Anthony should not copy wholesale 1. Zero shorts. Craig's channel has zero shorts and is growing. This is a valid strategy for a Substack-first creator with an existing email list to seed YouTube. For Anthony without a 44K newsletter backing his YouTube, shorts are a legitimate cold audience acquisition tool. Do not remove shorts from the Ramonov system because Craig does not use them. His email list is doing what his shorts would otherwise do. 2. Pure philosophy positioning. Craig can write about Camus and Aristotle because "ideas" is his stated niche and it is credible coming from a 22-year-old who genuinely reads philosophy. Anthony's audience is non-technical people learning AI tools. The philosophical depth can be in the reasoning structure (Problem-Insight-Solution) but the surface content stays AI-practical. The Sisyphus metaphor can inform the essay arc but should not become the topic itself. 3. Reading essays to camera. Craig's audience has accepted this format because his Substack subscribers already read his essays and came to YouTube to see him perform them. Anthony's YouTube audience is likely arriving cold. Reading a written essay to camera may work once the audience has grown to the point where they follow Anthony across platforms. Before that threshold, the format works better as a transitional tool than as the primary presentation mode. --- ## SOURCES - channel-top.json (Craig Perry videos: 404K, 149K, 96K), collected 2026-07-02 - channel-mid.json (Craig Perry videos: 59K, 50K, 24K), collected 2026-07-02 - channel-recent.json (Craig Perry videos: 6.2K, 4K, 1.5K), collected 2026-07-02 - INTERVIEW-EXTRACT.md (Eden x Craig Perry interview, YouTube video Lht818MdTyQ), collected 2026-07-02 - Substack profile scrape: substack.com/@profoundideas, scraped 2026-07-02, result: 44K+ subscribers, #10 Rising in Philosophy - Profound Ideas publication scrape: ideas.profoundideas.com, scraped 2026-07-02, confirmed essay titles, dates, engagement numbers, and DAN KOE byline on co-authored July 1 post - Dan Koe identity: CONFIRMED via Substack co-authorship on letters.thedankoe.com and Craig's Substack recommendations listing "future/proof by DAN KOE" - Dan Koe interview transcript partial read: confirmed interviewer voice pattern and Eden product context - PINS-INVENTORY.md, reviewed 2026-07-02 **Data gaps:** - channel-top.json video 3 (96K, "More Educated Than 99%"): transcript MISSING due to NexLev rate limit - channel-recent.json video 3 (1.5K, "Getting Unstuck"): transcript MISSING due to NexLev rate limit - YouTube search impression data: UNVERIFIED (not available without YouTube Studio access) - Exact Substack subscriber count at the time of first YouTube batch drop (April 2026): UNVERIFIED