Design your wings.
Feathers & Wax is a design research and product company. We’re building tools that help people find original ideas they actually want, and turn them into real designs.
- 19
- research papers
- 2.4M
- contest entries studied
- 1,759
- people’s ratings analysed
Original ideas are scarce
AI has made ideas cheap. It has also made them alike: ask a model for a design idea many times and most answers converge on the same one. Revising human work with AI pulls it toward that same answer.
We’re building Feathers & Wax to make ideas that are new to the person they are for, chosen by that person, and developed until they hold up. We start with architecture, and plan to extend to objects, brands, spaces and any creative brief.
26 of 30
answers from a frontier model to one skyscraper brief were the same idea.
51%
of the distance between human competition entries and the model’s usual answer disappeared after one AI revision.
How we work
Three steps, each based on something our research found. Read about the process
01
Generate widely.
Every idea starts from a different seed, so the set covers far more ground than a model’s usual answer.
02
Let the person choose.
People pick blind. Their picks teach the system which way they lean, and how far.
03
Develop with care.
The chosen idea becomes a model and drawings. A critic reviews every round, and new outside material keeps it from going generic.
Studies from our first brief
The three ideas behind The Unfinished Assembly, each pushed to an extreme on one question: territory, gravity and time. The final design combined them. These are generated concept studies, not renderings of a finished building.
Taste is personal
Unusual is not enough. In a museum competition with 615 entries, the most unusual-looking boards were nearly three times as likely to land in the jury’s bottom tiers as the most typical ones.
Most of taste belongs to the person. Across 1,759 people, the stable personal part of a rating is about twice the shared part. So we design for one person at a time.
Entries the jury placed in its bottom two tiers
What decides one person’s rating
- 44% stable and their own
- 23% shared with everyone
- 33% noise, the person disagreeing with themselves
Framework
Our first product. Framework writes down how a person thinks, so a model can reason the way they do. The same format will hold yours.
Research is better when shared
We publish our experiments, including the ones that did not work.
- What breakthroughs have in commonWe coded the documented processes of 74 creators in 11 fields. No single move is universal, but four jobs recur in every field: feed, reframe, import, and vary and select.Study
- What art survives automationIf making becomes nearly free, what is left for art? History suggests five traits that machines do not erase, and one risk that they create: everything starting to look the same.Essay
- The open loopIdeas revised in a closed loop drift back to the typical. People escape that by taking in real outside material. Living abroad is the clearest natural case, and it suggests a residency for our workflow.Essay
- Method, not motifFrameworks of invented people carry as well as famous ones, so the framework text works without the model’s memory. Told to carry how someone thinks and nothing from their field, frameworks made the designs a critic preferred.Study
Join us
Feathers & Wax was founded by Sean Choi, an architect, and works from San Francisco and Seoul. We’re looking for a cofounder who has trained preference models, and for people who want ideas made for them.
Sign up to hear when we publish research or release a product.