AI Creative Muscle Workout
Reverse-engineer excellent work, recreate it with AI, and diagnose every gap
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 6
- Confidence
- 96%
The AI Creative Muscle Workout uses deliberate reconstruction to develop creative skill alongside technical fluency. Begin with an excellent image, video, advertisement, or other artifact and ask an AI assistant to describe its composition, style, and construction. Attempt to reproduce it with a generative model, then compare the result with the reference. Each discrepancy becomes a diagnostic signal: perhaps the model lacks brand context, or the operator lacks vocabulary for typography, lighting, pacing, or composition. Study that missing component, enrich the instructions or context, and attempt the recreation again. Repeating this cycle across progressively stronger examples builds a working library of techniques, reveals what current models can do, and expands the marketer's sense of what is creatively possible.
Origin
Nicholas Holland described practicing this method while attending his children's sporting events, where he photographs interesting visuals, asks AI to reverse-engineer them, and then tries to recreate them with image models. Extracted from Marketing Against the Grain.
Core principles
- 01AI performance improves through deliberate practice
- 02Excellent outputs reveal techniques worth reverse-engineering
- 03Every failed recreation exposes a specific skill or context gap
- 04Creative range expands as tool fluency grows
How to run it
- 1
Choose an excellent reference
Select a strong image, video, advertisement, or other creative artifact that contains techniques you want to understand.
Pro tip Build a reference set from the best work in the world rather than limiting yourself to your own industry.
Watch out Do not select mediocre work merely because it is easy to imitate.
- 2
Reverse-engineer the construction
Give the reference to a multimodal assistant and ask it to describe the composition, style, elements, and likely production choices.
Pro tip Ask for language that could be reused as generation instructions.
Watch out A surface-level description may omit typography, lighting, hierarchy, or brand conventions.
- 3
Recreate it with AI
Use an appropriate image, video, or text model to produce your own version from the reverse-engineered description.
Pro tip Preserve the underlying technique while changing proprietary content or branding.
Watch out Do not present a close imitation as original commercial work.
- 4
Compare and diagnose
Compare the generated result with the reference and name the most important differences. Determine whether each gap comes from missing context, weak instructions, tool limitations, or your own knowledge.
Pro tip Diagnose one component at a time so each revision teaches a clear lesson.
Watch out Do not dismiss a weak result by simply declaring that AI cannot do it.
- 5
Learn the missing component
Study the relevant creative concept, acquire missing context, or test a more capable model before trying again.
Pro tip Use AI as a tutor for unfamiliar concepts such as typography, composition, or video pacing.
Watch out Repeatedly changing prompts without learning from the gap produces random iteration rather than practice.
- 6
Repeat at higher difficulty
Continue the cycle with more complex references and different media until the techniques become reusable instincts.
Pro tip Track recurring discoveries in a personal creative playbook.
Watch out Occasional experimentation will not build the same fluency as consistent practice.
In the wild
Holland described seeing a complex visual at a gymnasium, photographing it, asking AI how it was made, and using that analysis as the basis for reconstruction practice. The purpose was not merely to obtain one asset but to strengthen his ability to recognize and reproduce creative techniques.
→ The exercise expanded his understanding of both creative construction and the capabilities of current AI systems.
A marketer gathers 100 exceptional advertisements, asks AI to break down the creative thinking behind each, and recreates selected techniques in new campaign concepts. When a result falls short, the marketer studies the missing brand context, typography, or production technique before retrying.
→ Repeated analysis and reconstruction produce stronger creative judgment and a reusable vocabulary for directing AI.
Common mistakes
Treating generation as a one-shot task
Accepting the first output prevents comparison, diagnosis, and skill development. The value comes from cycling between reconstruction and analysis.
Blaming the model for missing context
A poor result may reflect absent brand guidance, source material, or domain language rather than an absolute model limitation.
Practicing only familiar techniques
Recreating work that already matches your abilities does little to expand your creative horizon. Strong references should expose skills you have not yet mastered.
Is it for you?
Best for
It is best for marketers who understand basic AI tools but struggle to produce distinctive, high-quality creative work.
Not ideal for
It is not ideal for teams seeking instant polished assets without investing time in experimentation and skill development.
From the transcript
“I will take a picture of something.”
“I will put it in chat GPT or Breeze to re to basically reverse engineer. I say, "How would you describe this image?"”
“you just have to work on these muscles and it helps you understand all sorts of things.”
From the episode
This is How the Top 1% Marketers are Using AI