Make Something People Want
Validate what people want before using AI to build it faster
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 91%
AI lowers the cost of research, prototyping, artwork, and software creation, but speed does not rescue an unwanted product. The decision rule is simple: keep your attention on what people want, then use AI to build and test that outcome faster. Start with a specific person and problem, collect evidence of demand, define the smallest useful result, and prototype it with the tools now available. Put the result in front of intended users and watch behavior rather than relying on the founder's enthusiasm. The mechanism joins demand discovery to cheap execution: market evidence chooses the direction, while AI reduces the time and technical skill needed to test it. This preserves the entrepreneurial advantage of rapid building without confusing easier production with a reason to produce.
Origin
Peter Norvig connected AI entrepreneurship to the Y Combinator maxim printed on his shirt: make something people want.
Core principles
- 01Demand comes before technical execution
- 02Building faster does not create market need
- 03Keep watching what people actually want
- 04Use AI to accelerate a validated direction
How to run it
- 1
Choose a specific user
Identify who has the problem and the context in which it matters.
Pro tip Narrow the user enough that you can speak to them directly.
Watch out A product for everyone makes demand evidence difficult to interpret.
- 2
Collect demand evidence
Look for repeated requests, workarounds, costs, or behavior that demonstrates what people want.
Pro tip Prefer observed behavior over compliments about the idea.
Watch out Founder conviction is not market demand.
- 3
Define the smallest useful outcome
Describe the minimum result that would solve a meaningful part of the user's problem.
Pro tip Cut features that do not test the core demand.
Watch out A larger prototype creates more work without necessarily producing better evidence.
- 4
Build the test quickly
Use AI for research, prototyping, artwork, or code to produce a usable demand test.
Pro tip Use AI to fill capability gaps rather than waiting for every specialist.
Watch out Do not mistake polished AI output for product-market fit.
- 5
Let behavior decide
Show the prototype to intended users and continue, revise, or stop based on what they do.
Pro tip Predefine the behavior that would count as evidence.
Watch out Positive words without meaningful action can be false validation.
In the wild
A biologist had long wanted an interactive map for exploring bird migrations but lacked the programming ability to create it. After trying an AI coding copilot, he built the app himself. The tool filled a technical capability gap and let the domain expert turn a concrete need into working software without first finding a technical co-founder.
→ A previously inaccessible product became buildable by the person who understood the problem.
Common mistakes
Starting with the tool
Choosing an AI capability before identifying a wanted outcome encourages impressive but unnecessary products.
Confusing speed with demand
AI can shorten development, but it cannot make customers want the result.
Is it for you?
Best for
Founders choosing, testing, or narrowing an AI-enabled product idea.
Not ideal for
Mandatory internal or regulatory projects whose value does not depend on market demand.
From the transcript
“you can now start doing things uh much more quickly you can prototype something and and go to a release product uh much faster”
“keep your eye on what it is that people want”
“make something people want”
From the episode
Peter Norvig: Simple Ways to Grow Your Business with AI
Peter Norvig