Hormozi's $4M Content System: What an AI Retention Tool Can Show
By Matt Cullerton, founder of Mavric and Apso
A TikTok from @angusthenontechnical argues that Alex Hormozi's content machine is a testing loop you can rebuild with AI: post, scrape your retention data, have a model name the phrases that lose viewers, then run trial reels. The $4 million figure is real. The claim that most of it went to testing has no source, and the report shown on screen says it cannot identify a weak phrase.
Did Alex Hormozi really spend $4 million on content?
Yes, over a shorter window than the video says. The video's own card cites "The Game, Ep. 750 · September 2024". Apple Podcasts lists that episode as "Part 1 - How I Gained 7.8 Million Followers In 40 Months | Ep 750," released September 30, 2024.
A transcript of the episode records Hormozi describing "$4 million in team, equipment, vendors, studios, and software, 1,000 hours of me on-camera recording, and 35,000 pieces of content" over "40-ish months." That transcript is a third-party one from podscripts.co, so treat the wording as close rather than exact.
The voiceover says "Four milli in the past four years." Forty months is three years and four months. The video's own graphic reads "Across roughly 40 months," so the correct number is on screen while the audio says something else. Stretching the window to four years lowers the implied yearly spend by about 17 percent.
Was most of the $4 million spent on testing?
No source says so. The video claims the money was "mostly spent on the testing side of things," and uses that to argue "you can vibe code his whole system."
Hormozi named five cost categories: team, equipment, vendors, studios and software. Testing is not one of them. The video's own card prints a footnote that concedes the point: "Conceptual emphasis, not a budget split. Testing is the largest visual area; no category dollar amounts or shares were disclosed."
Can AI map which phrases make viewers leave?
The tool in the video does not claim to. The voiceover says the AI can "literally map out which phrases cause the most people to leave." The report it cuts to prints "No clear underperforming passage" and "Early data cannot identify a reliable worst passage," with a badge reading "Early data · descriptive comparison."
The per-passage numbers explain why. Each passage is compared against 20 prior videos:
- 4 to 5 seconds: raw loss 3.0 points against a usual loss of 6.5.
- 5 to 7 seconds: raw loss 9.0 points against a usual loss of 7.4.
- 7 to 9 seconds: raw loss 4.0 points against a usual loss of 4.9.
Those gaps are too small to name a culprit. The largest single drop, 22.0 points, sits in the first two seconds, where short videos lose people who swipe away before anything starts.
The Claude Code thread shown later is more careful still. It calls a proposed cut "a hypothesis" and says "The archive tests also didn't find a cause." A closing card reads "Measure the revised cut before promoting it. The observed loss is rounding-sensitive." The tool reports uncertainty well. The narration sells it as a phrase-level drop-off detector.
Where does the retention data come from?
The workflow scrapes TikTok Studio through a logged-in browser. Its code launches Chromium with a saved login and the flag --disable-blink-features=AutomationControlled, opens https://www.tiktok.com/tiktokstudio/analytics/${itemId}, and keeps any response containing video_retention_rate_realtime. It also runs a detectBlockSignal check.
Each of those choices points to an undocumented surface. The flag hides the automation fingerprint, the saved profile supplies a session, and the block check exists because the operator expects to be blocked. TikTok supplies the retention curve. Matching that curve to spoken phrases is the creator's own layer, and that layer is the one reporting no clear weak passage.
Are trial reels a TikTok feature?
No, trial reels are an Instagram feature. The video tells viewers to post on TikTok, then says to "post two versions as trial reels" and that "whichever version wins goes to the main page." The screen at that point shows Instagram's "Trial reels" page with a "Create trial reel" button, followed by the Instagram profile @angus.sewell.
Instagram's creator blog describes trial reels as a way to "share reels with people who don't follow you." It says a trial reel "will not appear in your profile's reels tab or your main grid unless you later decide to Share to everyone," and that Instagram can "automatically share your trial reel with followers if we determine it's performing well based on the views it receives within the first 72 hours."
TikTok's documented two-version testing lives in Ads Manager. Its help article says to "Use the split test toggle in the TikTok Ads Manager when creating a campaign" and to set the test for "at least 7 days."
Can NotebookLM coach you from a lessons table?
Yes. The video's last step feeds a findings table into NotebookLM and asks for an audio overview. Google introduced NotebookLM on July 12, 2023 as "an AI-first notebook, grounded in your own documents," and it accepted only Google Docs at launch.
On November 13, 2025, Google added Sheets so users could "Add structured data and ask information on key statistics or generate summaries from your spreadsheets." The current help page lists CSV among supported sources, with limits of "up to 500,000 words or up to 200MB for uploaded files" per source and "up to 50 sources (for Free users)." A five-row lessons table fits easily.
Bottom line
Hormozi's $4 million over roughly 40 months is documented, but no source says most of it went to testing. The retention tool in the video is more honest than its pitch and declines to name a losing phrase on thin data. The trial-reel step runs on Instagram, and TikTok's split testing is an Ads Manager feature.
Sources
- Acquisition.com: names Alex and Leila Hormozi as co-founders.
- The Game, Ep. 750 on Apple Podcasts: episode title and September 30, 2024 release date.
- Ep. 750 transcript, podscripts.co: the $4 million quote and the 40-month window (third-party transcription).
- Instagram Creators: Trial reels: how trial reels work and the 72-hour auto-share.
- TikTok Ads Manager: How to create a split test: TikTok's paid split-test feature.
- Introducing NotebookLM, Google blog: the July 12, 2023 launch.
- NotebookLM adds more source types, Google blog: Sheets support added November 13, 2025.
- NotebookLM Help: Add sources: supported file types and size limits.
Published October 22, 2026. Based on a TikTok by @angusthenontechnical. Full video notes (markdown)