On 6 September 2024, an astrophotographer named Ian Lauer published a video called "I Turned My Astrophotos Into a Planetarium Show." It has 8,662 views.
Twelve days later, on 18 September, he published "Photographing a Galaxy for 10 Seconds vs 10 Hours." It has 3,682,749.
Same photographer. Same telescope. Same fortnight. A factor of 425 between them, and the thing that changed was the words on the front.
That comparison is the reason people talk about packaging, and it has become the most quoted case study in the creator world. It is also quoted wrong, by nearly everyone, including sources I respect. I will show you exactly how in a moment, because the way it is wrong is more instructive than the case study itself.
There are five sources that between them explain why one of those videos travelled and the other did not. This article checks each one against primary evidence and reports what survived. Two claims did not.
The mistake that ruins this before you start
The obvious move, once you have five good sources on what makes video work, is to put all five into one prompt and ask for a script.
What comes back is an average. It has the cadence of a retention manual, the vocabulary of a packaging strategist and the opinions of neither. Every distinctive thing each source believes gets sanded off against the others, because averaging is what happens when you ask one model to be five people simultaneously.
An assembly line is not a blender. Five stations, each one doing exactly one job, each one handing its output to the next. The output of station two constrains station three. That constraint is the entire value, and blending destroys it.
Five strong opinions averaged together produce something technically correct that believes nothing. Run them in sequence and each one has to survive contact with the last.
Station one: retention, from a document nobody has confirmed
In 2024, an internal onboarding guide for MrBeast's production company, titled "How to Succeed in MrBeast Production," became public. It runs to 36 pages, a figure corroborated by every outlet that covered it, including Fortune, TubeFilter, Cybernews, Dexerto and Fast Company. I read the circulating PDF end to end for this article rather than working from those write-ups, so every quotation below carries its page number.
Two things about its status matter before you use a word of it. Reporting credits the YouTuber Rosanna Pansino with putting the document into public circulation, sourced from a former employee. And across every article I checked, not one records the company confirming or denying that the document is genuine. It has been treated as authentic for two years on the strength of internal consistency and nothing else.
The document declares its own authorship on page 2, in the first person: "Hi, I'm Jimmy (MrBeast) and as the team is growing larger, I no longer get to spend as much time with everyone as I used to." It is written as an onboarding letter to staff, which is why it reads nothing like a creator course.
What makes it different is not the advice. It is that it cuts a video into minute blocks and gives each block a stated job.
Page 7 states the opening rule flatly:
"The first minute of each video is the most important minute of each video."
Then it flips the job immediately after. Same page: "After the first minute of content you will have what we call minutes 1 thru 3. This is where you have to transition from hype to execution (generally). Stop telling people what they will be watching and start showing them." And it names a specific device for the boundary, a "3 minute re-engagement", defined as a segment that makes people genuinely impressed, or as the document puts it, one where "only MrBeast can do this."
The passage that recalibrates everything is a worked example rather than a rule, and it spans pages 6 and 7. Describing the retention graph of a single video: "That represents the attention span of the 60 million people that clicked on this video." Then, on the next page:
"On this particular video we lost 21 million viewers in the first minute of the video (which surprisingly compared to other channels is above average) and could have been much worse."
Sit with that. A third of the largest audience on Earth leaves before the story starts, and the document calls it above average. Not acceptable, not survivable. Above average, explicitly by comparison with other channels. Every intuition you have about how much patience an audience extends to you is calibrated too generously.
Two details worth having. The same passage lists what would have made it worse, and the first item is lighting: "Let's say the start of the video wasn't well-lit, I didn't match the expectations of the clickbait, I didn't pre plan what I'd say." The phrase "the expectations of the clickbait" is his, and it is station one deferring to station two inside the same sentence. And the sentence that follows contains a genuine slip, printed as "we would have lost much less then 21 million people" when the argument plainly requires more [sic]. I am quoting it as printed because an artefact nobody has authenticated should be reproduced exactly, typos included.
The operating rule the document produces is not the number, it is a discipline, and page 8 puts it as an obligation on every department:
"Whether it is production, creative, or editing you must always know what minute mark the content you are working on is. If you don't then you're not doing it right."
The same page gives the frame that makes it workable: "13 minutes is the average and you are expected to know when producing which minute of that 13 minutes you are working on." Feedback like "this bit drags" is useless. "Minute four drags and minute four is supposed to be the first payoff" is actionable.
Four more lines from it are worth carrying around, because together they describe a shape rather than a tactic.
Four lines from the guide, verbatim
- Page 29: "Hook people at the start of the video, transition them to an amazing story that they are invested in, have no dull moments, and then have a satisfying payoff at the end of the video with an abrupt ending."
- Page 35: "The video endings must always be abrupt to protect retention."
- Page 8: "Don't ever signal the end of the video unless it's to build hype for the prize or payoff at the end of the video."
- Page 26: "Our audience is massive and because of that you have to be simple, for 50 million people to understand something it must be simple."
The abrupt ending is the one most people get wrong and it costs them nothing to fix. Every wind-down, every "so that was my experiment, let me know in the comments," is a signal to leave issued before the video is over. The guide's position is that you never issue that signal at all.
And the document's own final sentence, on page 36, is the best joke in it and also the point: "Now read this all over again because I guarantee you didn't retain enough."
How to give this to a machine. The brief for station one is a minute map. Every minute of the planned video gets a stated job and a stated obligation. Not a script, a map. The script is written against it afterwards, which is the only way you find out that your video has three minutes of setup before anything happens.
Station two: packaging, and the case study everyone quotes wrong
Paddy Galloway is the name attached to modern packaging thinking, and there is a detail about him that says more than any of his advice. His channel has 531,000 subscribers and 34,578,292 views, and his most recent upload to it is dated 6 August 2022. Four years. No Shorts either. Everything he has said publicly since lives on other people's shows and in his own written teardowns.
The clearest recent record of his thinking is the Colin and Samir write-up, and the lines that matter are blunt. Small creators spend "95% of their time on filming and editing" while the ones who win give roughly "30% of their time to ideation and packaging." His framing of the platform is "YouTube is a click & watch platform." And the one that lands hardest: "The difference between a million views and 28 million views is how you package it."
The case study he uses to prove it is Ian Lauer, the astrophotographer from the top of this article, described as going from "2-3K views per video to over 1M."
I checked that against the channel, because we check these.
The title everyone repeats was never published
The reframing is recorded as a move from "I photographed the Milky Way" to a comparative format given as "Photographing the Milky Way in 10 minutes, 1 hour, and 24 hours."
No such video exists on the channel. The video that broke is called "Photographing a Galaxy for 10 Seconds vs 10 Hours", published 18 September 2024, nine minutes fifty one seconds long, currently at 3,682,749 views and 38,366 likes.
Three things are wrong in the version that circulates. It is not the Milky Way, it is the Andromeda galaxy. The intervals are not ten minutes, one hour and twenty four hours, they are ten seconds and ten hours. And there are two intervals in the title, not three, even though the video's own description mentions three. The title dropped one on purpose.
That last point is not pedantry, it is the lesson. The real title is simpler than the remembered one. Every retelling has made it more complicated, which is exactly the direction packaging goes wrong. If you had built your titles from the quoted version you would have been copying a structure that never tested.
I do not think anyone invented this. It reads like a strategist recalling a case study approximately in conversation, and every subsequent retelling dropping the approximation and keeping the specifics. Which is how a made-up title ends up in a thousand people's swipe files.
What the case study leaves out
Here is the part nobody reports, and it is the most useful thing in this article. Lauer ran that format three times.
| Video | Published | Views |
|---|---|---|
| Photographing a Galaxy for 10 Seconds vs 10 Hours | 18 Sep 2024 | 3,682,749 |
| Photographing Orion for 10 Seconds vs 10 Hours | 17 Apr 2025 | 730,859 |
| Photographing a Nebula for 1 Hour vs 100 Hours | 15 Feb 2026 | 114,902 |
The second run of the format returned 20 percent of the first. The third returned 16 percent of the second. From first to third, the format lost 96.9 percent of its power in seventeen months.
Every packaging case study you will ever read stops at row one. The channel is real and public, so rows two and three were always available to anyone who looked.
A format is a depreciating asset. It is not a discovery you own, it is a position you hold while the novelty lasts, and the depreciation is steep enough that by the third use you are running a format that is now mostly a habit.
And one more, because it complicates the story properly. Lauer's biggest video is not the famous one. "Shooting the Milky Way from Every Light Pollution Level," published 15 November 2025, has 4,571,790 views, comfortably more than the case study that made his packaging famous. It is a different structure entirely. So the honest version is not "he found the format." It is that he learned packaging, and then kept finding new ones, on a channel that today has 120,000 subscribers and a video with 4.5 million views on it.
How to give this to a machine. Station two writes the title first, before a single line of script exists. Then the script is required to earn that title. This is the one ordering change that alters the most output, and almost nobody does it, because writing first and titling afterwards feels like the natural order and is the exact reason so many good videos die unwatched.
Station three: density
Station three asks one question: how many real, specific things does this say per minute before it starts to feel like padding?
The framework underneath it is Alex Hormozi's, from $100M Leads, in his material on posting free content. His model of audience-growing content is that it rewards the person consuming it, and that the reward only lands if three things happen in order: the person gets a reason to look, they stay long enough, and then that reason gets satisfied.
Read as a production rule rather than a marketing one, it is brutal. If a minute of your script contains one idea, it is not a minute. It is a stall. The reason to look was issued at the title. Every minute that does not advance toward satisfying it is a minute spending credibility it did not earn.
The practical version is countable, which is why it belongs to a machine. Station three takes the draft and counts claims per minute: named things, real numbers, specific examples, actual demonstrations. Anything below the threshold you set gets flagged and rewritten. Not "make it punchier." Count.
- Set a floor. Three specific claims per minute is a reasonable starting standard for an explainer.
- Count what is actually there, minute by minute, against the minute map from station one.
- Rewrite the minutes that fail, by adding specifics rather than by cutting length.
- Re-count. Density is the only one of the five stations with an objective pass mark, so use it.
Station four: tension
Johnny Harris has 7,900,000 subscribers and 1,240,267,167 views across 534 videos, on a channel that has been running since June 2011 and gained 90,000 subscribers in the last thirty days alone. He does one thing extraordinarily well, and it is not information. It is anticipation.
His framing, which I will describe rather than quote for reasons I explain in the corrections below, is that a story functions as a promise, and that what holds attention is the expectation that the promise is about to be fulfilled. Not the fulfilment. The expectation of it.
Now notice what has quietly happened across these four stations. Three sources who have never worked together are describing the same rule from different angles. Station one says the first minute is the most important. Station two says the title is a promise the script must earn. Station four says attention is anticipation of a promise being kept. That convergence is worth more than any of them individually, because independent sources agreeing is evidence and one source asserting is opinion.
The rule it produces is uncomfortable to follow and it is the one that works. Every section opens with the question it has promised to answer, and is not allowed to answer it in that section. The answer lands one section later, by which time a new question is already open.
Station five: you
Station five is not a famous creator. It is your own transcripts, and specifically the ones that actually held people rather than the ones you enjoyed making.
This is the station people skip, and skipping it is why AI-assisted content has a recognisable smell. If all five slots are filled with other people's thinking, what comes out is structurally excellent and sounds like nobody. It has the shape of a good video and the voice of an average of four strangers.
The mechanic is simple. Pull the transcripts of your own best-retaining videos, not your most-viewed ones, and use them as the final pass. The draft goes in structurally correct and comes out sounding like you.
Structure gets borrowed. Voice gets owned. That is the whole distinction, and it is the only part of this that cannot be copied from anyone.
The assembly line, in order
Five briefs, one per station, each written from your own notes on the source rather than from the source material itself.
The line, start to finish
- Station two runs first. The title exists before anything else is allowed to happen. This is deliberately out of numerical order.
- Station one builds the minute map against that title. Every minute gets a job and an obligation.
- Station three counts the claims per minute and rejects the stalls.
- Station four opens every section with the promise it will not yet keep.
- Station five runs last, through your own transcripts, so what comes out sounds like you and not like the four people it learned from.
MrBeast for attention. Galloway for packaging. Hormozi for density. Harris for tension. You for voice. Five stations, one line, and you are the last one on it.
What I could not verify
This section exists on every article we publish, and this time it includes a correction to our own companion video.
The handbook's leak date, and something better underneath it. Our video dates the leak to August 2024. The reporting splits: one outlet dates the original public posting to 13 August 2024, while the main wave of coverage runs from TubeFilter on 17 September 2024 through Fortune on 26 September, and the mirror I retrieved was uploaded on or around 20 September. So I have written "in 2024" rather than pick a month I cannot defend.
The more interesting finding is why nobody can settle it. The document carries no date anywhere in its 36 pages. No version number, no revision note, no copyright line, no confidentiality marking, and the file has no creation-date, modification-date or author field in its metadata either. What it does carry is a producer string identifying it as a Google Docs export, and a document title beginning "Copy of", which is what a duplicated Google Doc is called by default. The only dating clues are internal and vague: page 5 mentions "I spent basically 5 years of my life locked in a room studying virality", and page 35 says "As I write this we have 2 teams, that will grow to 4 in the next year." Everyone quoting a leak date for this document is dating the leak, never the document.
The Johnny Harris quotation. The line about every story being a promise is attributed to him widely and appears in our video in quotation marks. I could not find a primary published source for that exact wording, so I have described the idea rather than quoted it. If you see it in quotation marks anywhere, including from us, treat the wording as unsourced.
The Hormozi wording. The three-part model above is his framework from $100M Leads, described in my words. I have not reproduced book text verbatim, and you should not treat the phrasing above as a quotation.
Whether Lauer was a client. Our video states he was in an accelerator programme rather than a client, and that Galloway declines most of the credit. I could not find a published source establishing the commercial relationship either way. What I can say is that the write-up I used does not describe him as a client, and that the popular "an agency fixed this channel" version is not supported by anything I found.
The handbook itself. Everything attributed to it above comes from the leaked document, which no company has ever publicly authenticated. Treat it as a strong artefact rather than an official one.
The Lauer view counts, publication dates, subscriber counts and Galloway's last upload date are all from YouTube's own data, pulled on 13 August 2026, and are the parts of this article I would defend hardest.
Come and build the studio with me
If you want to feel the difference between prompting a chatbot and running a station, start free. Our Build Your AI Team course walks you through building the first seat, the Writer, and the prompt pack is yours to keep. The Claude prompting course sharpens the underlying skill that every station depends on. Both are on video, both are free, and neither asks for a card.
When you want the whole line rather than one station, that lives at ideasrepay.com, and the walkthrough for this exact business is The One-Person Studio.
It is the build rather than an article about the build. Seventeen steps across five phases, click by click. You get the six role prompts to paste straight in, so the crew exists in an afternoon instead of a month of trial and error. You get the packaging-first pipeline that writes the title before the script, the three-way niche fit test that sets your earnings ceiling before you make anything, the full run from a blank idea to a published video through four quality gates you never hand over, the quality control checklists, and the route from your first upload to monetization and diversified revenue. The Crew Build Kit and the Studio Operating Kit come as downloads, and the whole thing follows one operator from month zero to month fourteen so you see the shape of a year rather than guessing at it.
One payment of $99 opens that walkthrough and every other one we have published, plus every one we publish after it, across all three verticals: online businesses, YouTube and content, and offline work in the real world. Downloads, templates and scripts are included in every blueprint, there are no renewals and no upsells, and you can email us while you build. That is launch pricing, and it moves to $199 after the first 500 members. There is a community building alongside you and I mentor through it personally.
If you want the wider model first, AI YouTube Automation Gets Channels Terminated is the pillar on running this properly, and I Made $0 for Months is the honest account of the stretch before any structure starts paying you back.
Five people worked out the structure. None of them can supply the last station. That one is yours.