# Design Deserves Better Questions.

We can now produce a thousand designs before we have understood the need for one. Speed has changed. Responsibility has not.

Style Science is an open, evidence-aware framework for design reasoning: helping AI agents express, test and revise design decisions in service of human needs. It complements design thinking. Style is our language; reasoning connects its choices to human needs. Our aim is to support human judgment, not replace it.

## 1. Begin With the Life It Enters.

A design arrives somewhere. On a small screen. In a crowded room. In the hands of someone who is tired, hurried or unfamiliar with it. Begin there. The prompt is a brief; the person is the reason.

## 2. Give Every Decision a Reason.

A type size establishes importance. A distance separates or connects. A color directs attention. These choices act on one another. Make their relationships explicit enough to question. A reason may be provisional. It should not be absent.

## 3. Make Room for Character.

The average of familiar designs is not a standard worth pursuing. Let the subject determine the voice. A quiet page can be exacting. A vivid page can be disciplined. Seek a distinction that belongs to the work, and carry it through the smallest details.

## 4. Measure What You Can Defend.

Measure contrast. Measure overflow. Check whether the promised words are present. Then be precise about what those checks establish. A correct ratio cannot tell us whether a page moves someone. A preference cannot excuse text that cannot be read.

## 5. Let the Rule Be Challenged.

A convention contains experience, but it can also preserve an old limitation. Record where a rule applies, where it fails and why an exception was made. Keep the evidence that changes your mind. A system that cannot be corrected will eventually become a style to escape.

## 6. Build Knowledge That Travels.

A useful idea should survive a change of tool. Share the relationship, the calculation and the test. Keep the particular needs of a page, a chart or an interface intact. We should be able to carry understanding from one system to another without carrying the same appearance everywhere.

## 7. Spend Attention Carefully.

Generation has a cost, even when another version feels free. It uses computation. It asks for review. It leaves things to maintain. Explore alternatives that answer different questions. Keep what earns its place. Stop when another variation teaches us nothing.

## 8. Judge the Thing That Reaches People.

The intention is not the result. Open the document. Use the interface. Read the chart at its real size. Invite people outside the project to judge the work. If our system does not help them, its elegance is no defence.

## Work That Informs Ours

Our research direction is informed by Penrose, Scout, C-K design theory, Design Model, AgentsORG / DESIGN, design-pact and the Design Tokens Community Group. We study their ideas, credit their contributions and test what we can build upon. This does not imply affiliation, endorsement or implemented compatibility.

[Sources and research questions](./related-work.md)

We do not need machines with impeccable taste. We need systems whose decisions we can understand, whose limits we can see, and whose work we can make better.

Our commitment is to study design with the care we expect from it.

---

An original Style Science manifesto. October 2026.