超知能 · CHŌCHINŌ · A MIND YOU CAN OPERATE
SUPER INTELLIGENCE超知能
Don't watch it. Operate it.
// 00 — MANIFESTO
A superintelligence is not something you watch. It is something you operate. Every answer a machine gives you is shaped by a handful of levers: what you ask, who you tell it to be, how much chaos you allow, and what you chain it to. Below, every lever is live. Real models. Real tokens. Real images. Pull them.
// 01 — SPEAK · HANASU
Bend the mind.
Same question, different machine. A system prompt tells the model who to be. Temperature decides how predictable it is. Change them and ask again.
SYSTEM PROMPT · 人格
TEMPERATURE · 温度 0.70
WHAT IS ACTUALLY SENT · 送信
> channel open. model loaded on remote GPUs.
> ask anything. then change the persona or temperature and ask the same thing again.
// 02 — IMAGINE · SŌZŌ
Words into pixels.
Write an idea. Let a language model engineer the prompt for you — that's chaining two AIs. Then an image model paints it, and 95,000 particles rebuild it in three dimensions.
YOUR IDEA · 発想
STYLE · 様式 (adds words to the prompt)
FORMAT · 判型
FINAL PROMPT · 最終
—
Your image will appear here
— and in the particles behind you.
// 03 — LEARN · MANABU
Look inside.
This is a real neural network, training right now in your browser. No server. Click to drop cyan points, shift-click (or switch below) for coral. Watch it find the border. Each small square is one neuron, showing what it has learned to see.
// 04 — CREATE · TSUKURU
Give it a name.
One name, three models in a chain: a language model invents its kanji, its nature and its first words; then writes the prompt for its sigil; then an image model draws it. You keep the birth certificate.
// 05 — 段位 · DAN