Prompt Sets Are the New PRD
Communicate product ideas by building the prototype, not writing the doc — demos before memos
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 4
- Confidence
- 90%
In the AI era the fastest, highest-bandwidth way to communicate what's in your head is to build it. Chennapragada insists new projects and features arrive with prototypes and 'prompt sets' rather than PRDs. Prototyping is a loop accelerator; the inner loop of building and iterating collapses, but because the supply of prototypes explodes, the bar to break through rises — so don't chase every idea, and keep a small taste-making core.
Origin
Aparna Chennapragada's operating rule at Microsoft; she pairs it with Reed Hoffman's 'if you're not embarrassed by v1 you launched too slow' and the 'demos before memos' maxim.
Core principles
- 01Building to see what you want to build is the fastest path to communicate it
- 02Prompt sets and prototypes replace the PRD for new work
- 03Time-to-first-demo shrinks but time-to-full-deployment lengthens — expect an uneven cadence
- 04A flood of prototypes raises the floor and the ceiling: breaking out is harder
How to run it
- 1
Prototype instead of specifying
For any new project or feature, produce a prototype and a prompt set as the primary artifact of the idea — the high-bandwidth way to communicate a vision.
Pro tip When in doubt: demos before memos.
Watch out This isn't about turning everyone into a software engineer — it's about the fastest path to seeing and sharing what's in your mind.
- 2
Use the shortened inner loop
Exploit that idea-to-clickable-prototype now takes minutes, and even run user research through AI conversations, to accelerate the build-iterate loop.
Pro tip Treat the prototype as a loop accelerator, not a finished deliverable.
- 3
Raise the bar for what ships to scale
Because prototypes are now cheap and abundant, hold a higher bar for what earns full deployment; the time-to-scale is longer even though time-to-demo is shorter.
Watch out Don't chase after every idea — supply of ideas is exploding and most won't rise above the noise.
- 4
Keep a small taste-making core
Let many people experiment, but concentrate the editorial/taste-making function in one or a few people at the heart of the effort.
Watch out Without a taste-making core you just get a Frankenstein product.
In the wild
Chennapragada built a personal Chrome extension (new tabs prompt 'How can you use AI to do what you're going to do right now?') using GitHub Copilot in about ten minutes.
→ Demonstrates idea-to-working-artifact collapsing to minutes — the premise behind prompt-sets-as-PRD.
Common mistakes
Writing memos instead of building demos
Defaulting to spec documents when a prototype would communicate the idea faster and more faithfully.
Chasing every prototype to scale
Treating the exploded supply of cheap prototypes as all worth deploying, instead of raising the bar for breakout.
Is it for you?
Best for
Product managers and builders proposing new AI-era products and features
Not ideal for
Deep changes in the 'bowels' of a mature product where the team already knows exactly what to do and a prototype adds nothing
From the transcript
“in this day and age if you're not prototyping and building to see what you want to build I think you're doing it wrong”
“I call it the prompt sets or the new PRDS”
“demos before memos”
“the time to first demo the is much shorter, right? But the time to like a full deployment is uh is going to take longer”
“it becomes even more important to have the territorial and taste making kind of u uh one or few at the heart of it uh…”
From the episode
Microsoft CPO: If you aren’t prototyping with AI, you’re doing it wrong
Aparna Chennapragada