one-shot prototyping
person concept tool org talk claim — click a node to jump to its page; hover an arrow for the relation
One-shot prototyping is a design workflow in which a practitioner rapidly generates multiple distinct design variations by prompting an AI tool repeatedly in quick succession, treating each output as a throwaway prototype rather than a refined artifact.
Advocacy
Eve Bouffard, YC's Head of Design, explicitly endorses one-shot prototyping as a core part of her AI-assisted design practice. She describes running the same request up to sixteen times in a single session to build a personal reference glossary of outputs, using iteration volume as a substitute for lengthy manual exploration. →
Limitations and how to overcome them
A known criticism of the approach is that relying on text prompts alone tends to produce generic design output. Tools such as Claude (and Codex) are noted to return undifferentiated results when designers describe what they want purely in words. →
Bouffard's recommended remedy is to supply visual references—mood boards and real website examples—alongside the prompt, rather than depending on textual description alone. This breaks the generic-output pattern and allows one-shot prototyping to yield more distinctive, useful results. →
Points of disagreement
There is no direct counter-advocate in the available material, but the implicit tension is between volume-first iteration (run it many times, pick what works) and reference-first iteration (anchor each run with visual context). Bouffard's own practice combines both: she runs many iterations and uses visual references, suggesting the two strategies are complementary rather than mutually exclusive.