The 5-Prompt Sequence That Tells You If Your Ideas Hold Up
How to Stress-Test Any Idea With AI
Most people use AI to confirm what they already think. You have an idea, you ask Claude or ChatGPT to “explain” it, and the model dutifully hands back a polished version of your own opinion.
There’s a better way to use these tools. A single five-prompt sequence, build, challenge, destroy, rebuild, decide, will put any idea through more scrutiny than a week of back-and-forth with a colleague. You end up either genuinely confident in your position or convinced you need to change it. Either outcome is useful.
What You’ll Get
By the end of this sequence you’ll have:
A steel-manned version of your idea (stronger than you started with)
The three most credible objections, fully developed
A genuine attempt to demolish the whole thing
A rebuilt, tighter version that accounts for the weaknesses
A clear-eyed recommendation you can act on
The whole process takes about 20 minutes. You can run it on a business decision, a strategic opinion, a contrarian take you’re thinking of publishing, or an argument you keep having with someone.
The Real-World Scenario
Say you’re an entrepreneur and you’ve landed on a belief: “Hiring generalists is better than specialists for early-stage companies.”
You’ve argued this to your co-founder twice. You’ve half-written a LinkedIn post about it. But you’re not fully sure it holds up.
Run it through this sequence and find out.
How the Sequence Works
Each prompt does one job. Don’t skip ahead, and don’t combine them. The value comes from doing the stages in order. The output from each step becomes part of the input for the next.
Prompt 1: Build the Best Version of the Idea
What it does: Forces the model to construct the strongest possible case for your position before anyone attacks it.
When to use it: At the start, with whatever rough version of the idea you have. Don’t polish it first.
The Prompt:
I have a position I want to test rigorously. Here it is in rough form:
[PASTE YOUR IDEA IN 1-3 SENTENCES]
Your job right now is to build the best possible version of this argument.
Assume it is correct. Find the strongest evidence, the most compelling logic,
and the most persuasive framing. Give me the steel-manned case — the version
I'd be proud to defend publicly.
Write it as a clear, structured argument of around 300 words.How to use it:
Paste your idea in raw, unpolished form. Don’t pre-justify it.
Read the output and note anything that surprises you, ideas you hadn’t articulated yourself.
Save this output. You’ll reference it in Prompt 3.
Example input:
“Hiring generalists is better than specialists for early-stage companies.”
What you’ll get: A 300-word argument covering adaptability, resource constraints, the cost of over-specialisation, and why broad skills compound faster in a scrappy environment. Probably better-argued than your own version.
Pro tip: If the model’s steel-man surprises you with a point you’d never considered, that’s gold. It means there’s a stronger version of your idea than you’ve been defending. Use it.
Prompt 2: Find the Strongest Objections
What it does: Generates the three most credible counterarguments, written by someone who genuinely disagrees with you.
When to use it: Immediately after Prompt 1, before you’ve had time to feel good about the steel-man.
The Prompt:
Now switch sides completely.
Here is the argument you just helped me build:
[PASTE PROMPT 1 OUTPUT]
Your job is to be a sharp, well-informed critic who finds this argument
genuinely flawed. Identify the three strongest objections. For each one:
- State the objection clearly in one sentence
- Develop it in 2-3 sentences with specific evidence or logic
- Explain why it isn't easily dismissed
Don't steelman my position. Find the real cracks.How to use it:
Paste the Prompt 1 output directly.
Read each objection and honestly note: “Did I know this? Is it actually a problem?”
Mark any objection that you can’t immediately answer.
Example input: The steel-manned generalist argument from Prompt 1.
What you’ll get: Three developed objections, probably something like: specialists can do critical technical work generalists can’t, generalists plateau once the company scales, and in competitive hiring markets you can’t attract strong specialists later if you diluted your culture early.
Pro tip: The objection that stings most is usually the most important one. Notice your emotional reaction, that’s diagnostic.
You now have the best possible version of your idea and a cold, hard look at the three biggest reasons it might fail. Most people stop here; they either let their ego ignore the risks or they lose confidence and abandon the idea entirely.
Paid subscribers get the ‘Stress Test’ and the ‘Phoenix Rebuild’ to ensure their ideas actually survive the real world:
Prompt 3: The Demolition. A full adversarial attack to see if your logic holds up under professional fire. You need to know if it breaks before you ship it.
Prompt 4: The Phoenix Rebuild. How to synthesize the criticism into a tighter, more defensible position that accounts for every weakness we just found.
Prompt 5: The Final Call. A ‘No-Hedging’ prompt that forces Claude to stop saying ‘it depends’ and give you a concrete 48-hour action plan.
Bonus: The 1-Page Controversy Audit. My personal pre-flight checklist for any contrarian take or strategic shift.
Don’t just confirm what you already think. Upgrade to build ideas that actually stand up.