Label the Process Stage, Not the Polish
A production-looking prototype no longer means it's production-ready — say the stage out loud
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 3
- Confidence
- 88%
Historically the fidelity of an artifact signaled where you were in the process: something that looked like production meant assumptions were derisked and it was near shipping. AI severed that link — a polished-looking prototype can be a raw early exploration. Ambrosino's fix is to explicitly declare what stage a piece of work is at, because the medium no longer carries that signal for you.
Origin
Andrew Ambrosino (Codex app lead, OpenAI); connects to the 'primal mark' idea a prior podcast guest shared, and to the classic divergence/convergence design process.
Core principles
- 01Fidelity used to imply process stage; that coupling is now broken
- 02Polished exploration will get mistaken for 'ready to ship' unless you say otherwise
- 03Over-anchoring on a production-ready-looking artifact kills better ideas
- 04The 'overlay' of the process — knowing where you are in it — matters more than ever, even as the specific tools die
How to run it
- 1
Attach an explicit stage label to every artifact
When you share work, state plainly whether it's early multiplayer exploration, a derisking prototype, or a ship candidate — regardless of how finished it looks.
Pro tip Say 'this looks production-ready but it's actually early exploration' before anyone else assumes otherwise.
Watch out Enough people seeing a polished prototype will start asking 'can we release this now?' when you're actually still in early design.
- 2
Separate visual readiness from model/research/business readiness
Check whether the underlying model of where research is heading, what users ask for, and what's right for the business is actually right — not just whether it looks ready.
Watch out 'Visually it's ready for prod, but it's not actually the right model' is the failure mode to guard against.
- 3
Keep the process overlay even as you drop the old rituals
Abandon the rigid day-to-day design-process specifics, but retain the shared sense of 'we are at this point in the process.'
Watch out Throwing out the process overlay entirely leaves teams unable to tell exploration from commitment.
In the wild
Ambrosino describes 90 people having the same idea, producing something that looks really polished, and colleagues asking to release it — when in reality that polished output IS the early design process now, not the end of it.
→ Teams must verbally re-establish the stage because the artifact's polish lies about it.
Common mistakes
Reading polish as readiness
Assuming a production-quality-looking prototype has been derisked and reviewed leads teams to ship explorations that were never validated against research, users, or business goals.
Being stubborn that a rough result means a bad feature
Concluding 'this isn't working so it's a bad feature' when it may simply be too early for the current model conflates stage with verdict.
Is it for you?
Best for
Cross-functional teams where AI-generated prototypes circulate widely and get mistaken for finished work
Not ideal for
Solo builders or tiny teams where everyone already shares full context on an artifact's stage
From the transcript
“the medium implied it had baked in a lot of a lot of signal around where in the process something was”
“you do not want to over anchor on this thing that was meant to be an exploration, but now it looks so production ready”
“this is actually that that's the design process now”
From the episode
OpenAI Codex lead on the new shape of product work
Andrew Ambrosino