MMarketing Against The Grain
← All frameworks
ProductivityRob Lennon

Explore, Synthesize, Then Compress

Discover deeply with separate prompts before encoding the workflow as a mega prompt.

Difficulty
Moderate
Time to result
~weeks to results
Steps
6
Confidence
94%

Explore, Synthesize, Then Compress separates AI work into discovery and production modes. First, the creator uses multiple prompts to investigate a topic, inspect intermediate answers and follow unexpected leads. Once the conversation contains useful material, the model is asked to summarize or outline the most important information, turning conversational excess into a manageable artifact. Only after the workflow has proved valuable are its repeated steps combined into a mega prompt that can execute several operations at once. The compressed prompt becomes a shortcut for routine use, while the original multi-step process remains available for high-value work. The central trade-off is explicit: mega prompts save time, but separate prompts expose more intermediate output and therefore more chances for inspiration.

Origin

Rob Lennon described moving from exploratory ChatGPT sessions to reusable mega prompts on Marketing Against The Grain.

Core principles

  • 01Explore before optimizing.
  • 02Separate prompts preserve more opportunities for inspiration.
  • 03Synthesis converts conversational sprawl into usable structure.
  • 04Mega prompts should encode a proven workflow, not an imagined one.
  • 05Speed and discovery require different prompting modes.

How to run it

  1. 1

    Explore in small moves

    Use separate prompts to investigate the audience, topic, assumptions and possible angles. Read each answer before deciding what to ask next.

    Pro tip Keep a scratch list of prompts that consistently produce useful transitions.

    Watch out Do not keep generating after you have stopped evaluating the answers.

  2. 2

    Follow productive branches

    Pursue surprising or important ideas with targeted follow-up questions. Abandon branches that add volume without improving understanding.

    Pro tip State what was useful in the prior answer before requesting the next refinement.

    Watch out Unbounded branching creates information overload.

  3. 3

    Synthesize the session

    Ask the AI to summarize the most important information or produce an outline from the conversation. Specify the intended final use of the synthesis.

    Pro tip Request decisions, evidence gaps and unresolved questions as separate sections.

    Watch out A summary can hide valuable nuance, so retain the original conversation.

  4. 4

    Identify the stable sequence

    Review which steps repeatedly created value and arrange them into a dependable order. Separate required operations from optional explorations.

    Pro tip Include evaluation criteria, not just generation instructions.

    Watch out Do not automate steps that still depend on unclear judgment.

  5. 5

    Build the mega prompt

    Combine the stable operations into one structured prompt that names the topic, transformations and desired output. Test it on a fresh subject.

    Pro tip Number the operations and define the output format explicitly.

    Watch out Compression can cause the model to skip reasoning or flatten distinctions.

  6. 6

    Choose depth or speed

    Use the mega prompt for repeatable, time-sensitive work and the separate workflow for high-stakes exploration. Periodically compare their results.

    Pro tip Maintain both versions rather than replacing the exploratory process.

    Watch out Do not optimize for speed when inspiration is the primary objective.

In the wild

Compressing a content research session

A creator separately asks for uncommon subtopics, audience fears and counterintuitive resolutions. After testing the sequence across several topics, the creator combines those operations into one structured mega prompt. Routine newsletter planning uses the shortcut, while flagship essays retain the longer interactive process so intermediate ideas can be examined.

Recurring ideation becomes faster without eliminating the deeper workflow used for original work.

Recovering an outline from a long chat

After an hour of research, a marketer asks the model to summarize the most important information and organize it as a content outline. The marketer checks the result against the original conversation, restores one omitted insight and uses the revised outline as the drafting basis.

A sprawling session becomes an actionable document while preserving human editorial control.

Common mistakes

Compressing before learning

A mega prompt built too early encodes assumptions about a workflow whose valuable steps have not yet been observed.

Ignoring skipped inspiration

Combining operations removes intermediate stopping points where unexpected ideas might otherwise be noticed and pursued.

Falling for the slot machine

Repeatedly generating without reading or deciding creates dopamine-driven activity rather than useful exploration.

Is it for you?

Best for

It is best for creators who repeat similar research or ideation tasks and need both depth and operational speed.

Not ideal for

It is not ideal for novel assignments where the useful steps and evaluation criteria have not yet been discovered.

From the transcript

Over time you can start to consolidate some of those things and build what I call mega prompts, where you have it skip steps or…

Rob Lennon · 10:30

by skipping steps, you're also skipping potential inspiration.

Rob Lennon · 11:30

Analysis paralysis and information overload are gonna be huge problems for early adopters of AI because it's just so good at spinning stuff out.

Rob Lennon · 05:30

From the episode

How To Delegate Your Work To ChatGPT (Use These Prompts) with Rob Lennon

Rob Lennon