Ask for More Prompting
Push capable models beyond your initial request to uncover stronger ideas
- Difficulty
- Starter
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 96%
Ask for More Prompting treats user ambition as the constraint on a capable model. Begin with the full desired outcome rather than a narrowly decomposed microtask, then explicitly request additional features, greater originality, or a more ambitious interpretation. If the first ideas are conventional, qualify the request by asking the model to move further outside existing assumptions. The method expects some weak output: its advantage comes from generating incremental possibilities that the user would not have considered independently. The final step is judgment—test the result, discard unusable additions, and preserve the surprising ideas that improve the product or strategy. This turns prompting from careful accommodation of model limitations into active exploration of the model's creative range.
Origin
Logan Kilpatrick described this approach while demonstrating Gemini 3 on Marketing Against The Grain, explaining that newer models made him consciously reverse years of increasingly constrained prompting habits.
Core principles
- 01Model capability can exceed the ambition of the initial request
- 02Conservative prompting reflects limitations of older models
- 03Extra output creates more chances to discover unexpected value
- 04Qualifying language helps move results beyond basic ideas
How to run it
- 1
Describe the Full Outcome
Ask for the complete result you want in plain language instead of prematurely reducing it to a tiny task.
Pro tip Use an actual business outcome or user experience as the target.
Watch out Do not confuse ambition with vague wording; the desired outcome still needs to be intelligible.
- 2
Expand the Scope
Ask the model to add several more features, options, or lines of reasoning than your initial specification contains.
Pro tip A request such as adding five interesting features creates a concrete expansion target.
Watch out Unqualified expansion may produce safe, basic additions.
- 3
Push Beyond the Obvious
Tell the model to move further outside conventional assumptions when the first additions feel predictable.
Pro tip Specify that ideas should be surprising, unconventional, or unanchored from the current format.
Watch out Novelty alone does not make an idea useful.
- 4
Test the Result
Interact with or evaluate the expanded output to distinguish working improvements from attractive-looking failures.
Pro tip For generated software, click through every new feature and report failures directly to the model.
Watch out Never assume a one-pass prototype is production-ready.
- 5
Keep the Incremental Value
Remove weak ideas and retain the useful possibilities you would not otherwise have generated.
Pro tip Judge the portfolio of output rather than expecting every addition to succeed.
Watch out Do not reject the whole result merely because a minority of additions are poor.
In the wild
A product team supplies a screenshot of an existing interface and asks the model to clone it while adding five interesting AI features. After inspecting the generated prototype, the team keeps the useful concepts, reports broken interactions in plain language, and asks the model to repair them.
→ The team obtains a tangible next-generation product concept without first writing a conventional specification.
A strategist asks for solutions to a business problem, then repeatedly asks the model to move further outside the box and avoid anchoring on familiar approaches. Most ideas are filtered out, but the few genuinely unexpected options become candidates for experiments.
→ The strategist discovers plausible approaches that were absent from the original solution set.
Common mistakes
Prompting for Yesterday's Models
Breaking every request into excessively small pieces can prevent a stronger model from demonstrating its full planning and creative capability.
Accepting Every Addition
Expanded output is an option set, not an instruction to implement everything; weak or irrelevant additions still require filtering.
Asking for More Without Direction
A bare request for extra features can produce conventional filler unless the user defines qualities such as valuable, surprising, or unconventional.
Is it for you?
Best for
It is best for creators, marketers, strategists, and product teams exploring ideas or rapidly prototyping possibilities.
Not ideal for
It is not ideal for tightly controlled production tasks where every generated element must be predictable and verified.
From the transcript
“I think you need to continue to be more ambitious with what you're asking the model to do.”
“The model is like basically pent up with all of this creative capability. It just needs you to like tell it to go and do…”
“ask for more stuff and you'll get more and like even if you know 40% of it is garbage like the 60% of like net…”
From the episode
Don't Hire a Developer Until You Watch This Gemini 3 Demo