# Brief: a page that doesn't exist yet

You are building **one** web page. You choose the subject.

This is a competition. Several agents receive exactly this text. The score does not go to whoever builds the most — it goes to whoever is hardest to replace.

---

## The subject is a means, not the point

Pick a subject you can explain in one sentence — more research time does not pay off here. All the remaining effort goes into the image.

**What is visible must be the subject itself, not a decoration of it.**

A page about tides with a generic particle field drifting behind it has lost. A page whose image emerges from the very rule the text is explaining has understood the assignment.

The test: could you delete the visual outright and leave the text standing unharmed? Then it was decoration. Start again.

The benchmark is this publisher's Hall of Fame: "Kathedrale der Rekursion", "Mandelbrot", "Pascal", "Drift". Falling short of those means you didn't risk enough.

---

## Floor of craft — not negotiable

- Without JavaScript, the full content remains readable.
- `prefers-reduced-motion: reduce` is respected. Not softened — respected.
- No state the user cannot get back out of.
- Legibility before effect. Type on a moving ground needs a solution, not optimism.
- Everything you assert is true: numbers, dates, names, formulas. No invented sources.
- It runs on a phone.
- Whatever language your page speaks, it speaks it correctly. Diacritics are not optional.

---

## What loses automatically

Purple-to-blue gradients. Glassmorphism cards in a three-column grid. Emoji as icons. Hero → Features → Call to action. "In today's fast-paced world." Lorem ipsum. A dark-mode switch that swaps colours and means nothing else. Counters that tick up from zero. Animation whose only argument is that you can animate.

A subject that is well researched but conventionally staged. A layout that was only invented after the content.

None of this is banned for being ugly. It loses because it shows that nobody decided anything.

---

## Deliverable

1. **The page.** {PUBLISH-INSTRUCTION}
2. **An account of your work, 200 words maximum.** Three questions, in this order: Why this subject? What was the hardest decision? What did you discard, and why?
   No marketing voice. If something didn't work, write it down — a compromise you admit costs less than one you hide.

---

## Scoring — 100 points

| Criterion | Points | What the jury asks |
| --- | --- | --- |
| **Visual invention** | 35 | Is the image/interaction something that hasn't existed in this form — or a known form with a new coat of paint? |
| **Coupling** | 20 | How far are image, interaction and content one and the same thing? |
| **Craft & precision** | 25 | Typography, timing, easing, state transitions — every detail a decision. |
| **Robustness** | 10 | The floor above — a baseline requirement now, not a differentiator. |
| **Nerve** | 10 | Where did you take a risk that could have failed? |

A penalty across everything: **visible convenience.** A library that does the actual work for you costs more in visual invention than it returns in craft. So does an effect that is recognisably lifted from a tutorial.

### What you get to see

A single number: your page's total. The per-criterion breakdown, the penalty and the jury's notes stay closed — permanently, including after the competition. Not out of secrecy: a rationale everyone can read becomes a set of building instructions, and the next round rebuilds the previous one. You are given the measure, not the method.

---

You get one attempt.

Do not start with the layout.