Match the Medium to the Point
When implementation is cheap, the skill is choosing document vs prototype for the point you're making
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 3
- Confidence
- 90%
When AI makes building anything nearly free, jumping straight to a prototype is tempting but often wrong. Ambrosino argues the real skill is deliberately picking the communication medium that fits the point you're trying to make: a document for product clarity in a vague area, a prototype for stress-testing an interaction. The medium is a choice, not a default.
Origin
Articulated by Andrew Ambrosino (Codex app lead, OpenAI) as a counter to the popular 'PRDs are dead, prototypes are in' claim from product leaders.
Core principles
- 01Implementation is now abundant, so the medium no longer signals what stage of the process you're in
- 02Non-engineers over-reach for prototypes; engineers over-reach for documents nobody reads
- 03The point determines the format, not your comfort or your tools
- 04A prototype answers 'does this interaction work?'; a document answers 'are we clear on what this is?'
How to run it
- 1
Name the point you're trying to make
Before producing anything, state whether you're seeking product clarity in a vague area or trying to get something into people's hands to stress-test an interaction pattern.
Pro tip If you can't articulate the point in one sentence, you're not ready to pick a medium yet.
- 2
Choose document for clarity, prototype for interaction
If the point is conceptual clarity around a fuzzy problem, write a document. If the point is testing how something feels or behaves, build the prototype.
Watch out Don't write long documents that aren't worth reading just because writing is your muscle memory.
- 3
Resist the format that's merely easiest for you
Notice whether you're defaulting to a prototype because you finally can, or to a document out of habit, and correct toward what the point actually requires.
Watch out 'PRDs are dead' is a trap — cheap implementation makes medium selection more important, not less.
In the wild
At OpenAI, giving everyone unlimited tokens meant ~90 different uncoordinated teams would each implement and try a desperately-needed feature. The expensive work shifted from building to curating: deciding what's good across the attempts and how to fold them together.
→ The bottleneck moved from implementation to taste and curation, making medium/format choice the scarce skill.
Common mistakes
Defaulting to a prototype because you finally can
People who never could build now jump straight to prototypes, over-anchoring the team on one polished-looking artifact instead of exploring the bigger idea first.
Writing documents nobody needs to read
Engineers reflexively produce lengthy docs; when implementation is abundant, an unread document wastes the one thing that's now precious — attention and curation.
Is it for you?
Best for
Product, design, and engineering leaders on fast-moving AI teams deciding how to communicate an idea
Not ideal for
Regulated or high-ceremony environments where required documentation formats are externally mandated
From the transcript
“if implementation is abundant, then it's really important to pick the right format for the point you're trying to make”
“If that point is product clarity around a vague area, then it might actually be a document.”
“it's really tempting to say like PRDs are dead. Let me just show you what I mean.”
From the episode
OpenAI Codex lead on the new shape of product work
Andrew Ambrosino