Create Once, Transform by Medium
Create the original insight yourself, then use AI to adapt it for each channel.
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 96%
Create Once, Transform by Medium separates original thinking from content adaptation. The creator first develops a high-quality canonical artifact—such as a Twitter thread—using current knowledge, firsthand experience, and personal judgment. Once satisfied with the substance, the creator gives the complete artifact to an AI and requests versions designed for specific channels, such as a blog post or LinkedIn post. The larger context window helps the model retain the source's argument while changing structure, length, and presentation. Human review then protects the original point of view and removes distortions. This workflow increases publishing volume without replacing the source insight with generic model-generated material, and it avoids the weak practice of pasting identical content into channels with different audience expectations.
Origin
Nathan Labenz described this division of labor for repurposing his AI analysis threads during Marketing Against The Grain.
Core principles
- 01Original insight and point of view come before automation.
- 02A finished source artifact gives the model reliable context.
- 03Each medium requires a distinct format rather than a direct paste.
- 04AI is strongest at transformation when it does not need to invent the thesis.
How to run it
- 1
Create the Canonical Piece
Write and refine the original artifact using your own expertise, experience, and point of view.
Pro tip Choose the medium in which you think most clearly.
Watch out Do not ask the model to manufacture personal experience.
- 2
Lock the Core Meaning
Identify the thesis, essential evidence, tone, and claims that every adaptation must preserve.
Pro tip Include these requirements alongside the source artifact.
- 3
Choose One Destination
Define the next channel and its native conventions, audience, and length constraints.
Pro tip Adapt one medium at a time for greater control.
Watch out A generic request to repurpose everywhere produces generic output.
- 4
Generate the Transformation
Ask the AI to recast the approved source for the destination without adding unsupported claims.
Pro tip Supply the entire source when the context window permits.
- 5
Review and Publish
Verify fidelity, restore personal phrasing, and approve the adaptation before publication.
Pro tip Compare every major claim against the canonical artifact.
Watch out Fluent output can still subtly change the creator's position.
In the wild
An AI analyst writes a current, opinionated Twitter thread without model assistance. After approving it, he supplies the thread to GPT-4 and requests a structured blog post and a concise LinkedIn version, preserving the analysis while changing the presentation for each audience.
→ One piece of original thinking supports several native channel assets with much less reformatting work.
Common mistakes
Outsourcing the Thesis
When the model lacks current information or firsthand experience, asking it to create the core argument produces generic or inaccurate content.
Copying Without Adapting
Pasting the same artifact into every channel ignores differences in audience, structure, and reading behavior.
Skipping Fidelity Review
The transformed version may introduce claims or emphases the creator never intended.
Is it for you?
Best for
Experts and creators with strong original material but limited time to reformat it for multiple platforms.
Not ideal for
Creators who lack a substantive source artifact and expect AI to invent a credible personal perspective.
From the transcript
“So oddly enough, I don't really use AI at all in writing my threads.”
“So it should allow me to do more, you know, it should be helpful in my content creation, but I'm not expecting anything this year…”
“It's going to be more about transforming it after it's created.”
From the episode
GPT-4 Beta User Reveals What Jobs It Will Destroy In 2023 with Nathan Labenz (#103)
Nathan Labenz