Somebody asked me a while back how long it takes me to write a blog post, and I gave a bad answer, because the true answer is that I don't write them, and the question has an assumption buried in it that took me a while to notice. I record them, on a walk, and then I edit. The writing is maybe a fifth of the job and it's not the part that decides whether anything gets published.
Most people treat publishing as one skill, the way you'd treat playing an instrument. It's closer to running a kitchen, where the sourcing and the prep and the cooking and the plating and the service are five different jobs, and a restaurant that fails is almost never failing at all five. It's failing at one, and usually at prep.
So here are the pieces, what each one does, and which one is actually stopping you.
The five layers
The thought capture. Getting it out of your head and into a recording. Ten minutes talking through questions you wrote down beforehand. This is the raw material and it's the only step nobody else can do for you.
The transcript. The recording becomes text, because text is the format every other layer is built out of. This used to be expensive and slow, and now it runs on a laptop for nothing, which is a bigger deal than it sounds like and I'll come back to it.
The long piece. One substantial thing per cycle, built from the transcript. An essay, a video, a proper guide. This is the source of truth, and it's the only thing in the system you compose rather than extract.
The cuts. Everything else, pulled out of the long piece. The social post, the short, the quote card, the newsletter section. These get extracted, never written fresh, and that distinction is the difference between a system and a treadmill.
The distribution. Where it goes, in what order, and whether you can tell afterward what worked. Most people do the first part of that and skip the last part entirely.
That's it. Five layers, and every content operation you've ever admired is some arrangement of those five, whether or not the people running it would describe it that way.
Which one is your bottleneck matters more than any of the advice you've been given, because the advice is almost always aimed at a different layer than the one that's stuck. If you have no recordings, better writing tools are irrelevant. If you have forty hours of recorded calls and nothing published, you don't have a capture problem, you have everything else.
The bonus asset
There's a sixth thing that isn't a layer exactly, and it changed how I work more than the other five combined.
Keep every transcript. Dated, filed, never edited, never overwritten. Not the polished draft, the raw thing you actually said.
The reason is that a transcript library is a history of your own thinking, and it accumulates. Six months in, you stop starting from nothing on any topic you've ever talked about, because you've already talked about it and the words are sitting there in your own phrasing. A year in, it's the most valuable document in the business, and it's the one nobody thinks to build because audio isn't searchable and text is.
This is the same idea as version control for code, which is the thing that makes a website ownable rather than rentable. Nothing gets overwritten, everything has a date, and you can go back and look at exactly what you thought in March. That comparison isn't decoration. It's the same mechanism producing the same benefit in a different medium, and I wrote about the code version of it in the website primer.
Almost every business I talk to is already sitting on a version of this and doesn't know it. Recorded sales calls, webinars, workshops, internal decks. That is usually the largest content asset in the building and it's invisible, because nobody has had a reason to turn it into text.
What I use, and why
| Layer | What I use | Why |
|---|---|---|
| Thought capture | Phone voice recorder, on a walk | No setup, no excuse. Walking makes me explain things instead of writing them. |
| Transcript | An open-source speech model on my own laptop | Free, unlimited, and nothing gets uploaded anywhere, which matters when the recording mentions a client. |
| Long piece | Claude, drafting from the transcript, then a real editing pass by me | It's editing, not authorship. The arguments are already in the transcript. |
| Cuts | Descript for video, and the same transcript for the written ones | Text-based video editing, so cutting a sentence out of the video is cutting it out of the transcript. |
| Distribution | My own site first, then the platforms pointing back at it | The site is the thing I own. Everything else is an on-ramp. |
| Library | Dated raw transcripts, filed and never edited | The compounding part. |
About $20 to $55 a month for all of it, and the floor is real. The full breakdown, plus every price, is in the step-by-step version of this.
What this looks like on one real piece
To make it concrete, one set of mine, by the numbers.
Twenty minutes writing questions produced 2,678 characters of prompts. Walking and talking through them produced 10,983 characters of transcript. That transcript became a published essay of about 8,200 characters and a spine for a video, and the essay barely shrank on the way, which is the part worth noticing. It was a trim, not a rewrite. I wasn't short of content, I was short of a recording.
Then the cheap thing went out first. A forty-second teaser on July 30, the essay on August 5, and the video wasn't shot until August 13, by which point I already knew the hook worked on strangers. That ordering is most of the value of thinking in layers.
The piece was AI Doesn't Have Taste, and the whole trail, including the part where the video came in at nearly twice its target length, is in the deep dive.
Two people worth listening to, who don't agree
I'd rather hand you a real disagreement than a consensus that isn't one, and on this the disagreement is about machinery.
Alex Hormozi runs Acquisition.com and wrote $100M Offers and $100M Leads, and his position is that free content is the most reliable lead source available and you should industrialize it. His stated rule is to give away the secrets and sell the implementation, on the grounds that showing people what you know is what earns their attention and eventually their money. He puts numbers on the volume, most famously a rule of a hundred, meaning something like a hundred minutes a day making content or a hundred outreach attempts, every day, measured. He does say to pick one platform first rather than spraying all of them, which people misremember. But the thesis is that output volume is the constraint and the answer is a machine.
Seth Godin has published a post on his own blog every single day since January 2002, which is over twenty years and well past seven thousand posts, and he has done it with no atomization apparatus at all. No clip machine, no cross-platform pyramid. A blog he owns and an email list of people who asked to hear from him, which is the asset he has argued for since Permission Marketing in 1999. His position is that the daily practice is the thing, and that the surface you publish on should be one nobody can take away from you.
Where they differ is where the leverage sits. Hormozi says in the volume and the measurement. Godin says in the consistency and in owning the ground you stand on.
I've landed in a specific place after running both instincts badly. Hormozi is right about the bottleneck for almost everybody reading this, which is that you don't have enough raw material and the fix is to record more, more often, with less ceremony. Godin is right about the destination. Every piece I make points at a page on a domain I control, because the platforms are rented distribution and the archive is the thing that compounds. What I don't do is Hormozi's volume, because it assumes a team, and one person hitting that number produces thin work that teaches people to skim you.
Record like Hormozi. Publish like Godin.
What I'd do first
If you're publishing nothing, this is not an emergency and I'm not going to pretend it is. But it does compound, so earlier is better than bigger.
- Find out which layer is actually stuck. Count what you have. Recordings with nothing published is a downstream problem. Published pieces nobody saw is a distribution problem. Nothing anywhere is a capture problem. Fix the one you have.
- Write five questions before you record anything. Twenty minutes, on paper, on a topic you'd argue about. This is the step everybody skips and it's the one that makes the recording usable.
- Record ten minutes on the first question and don't publish it. Just prove to yourself the raw material exists, because most people are convinced it doesn't and they're wrong.
- Start the library today, even if it's a folder. Dated files, raw text, never edited. It's worth almost nothing this month and quite a lot next year.
- Find the archive you already have. Old recorded calls, webinars, workshops. Run one of them to text and read it. That's usually the moment the whole thing clicks.
How this connects to everything else
This piece is about the machine. It deliberately doesn't answer what to point the machine at, which is the more important question and the one people skip, and I wrote that up separately in Content Scaffolding: A Five-Minute Primer. Read that one first if you haven't decided what you're the authority on, because a working production system aimed at nothing in particular just produces more of nothing in particular.
For the reason any of this ends up findable, there's SEO: A Five-Minute Primer on how search works and AEO: A Five-Minute Primer on the version where an AI answers instead of listing links.
And if you want the actual order of operations, with the tools, the prices, and the four places it breaks, that's How I Actually Make Content.
If you'd rather hand the whole thing to someone, that's what I do.
Positions attributed to Alex Hormozi (Acquisition.com, "$100M Leads") and Seth Godin (Seth's Blog, "Permission Marketing") are my summary of their published work, not quotes.
