The machine that predicted 200 million protein structures

Work requirements

Details

ASK: one plate that makes the scale of a machine trying everything at once physically felt. FACT (locked): one AI system predicted and released more than 200 million protein structures - close to every catalogued protein - a jump of about 200 times from the roughly 1 million known before. Source: DeepMind and EMBL-EBI, 2022; the database is now about 214 million (Nucleic Acids Research, 2024). FACE: many attempts at once - the scale face, not the pricing shift. BEAT: where a person tries a handful, the machine tries them all. BAR: rewards close looking. DIRECTION (blank page): a named direction - era, movement or designer - a 2 to 4 ink palette with named hexes, and a FRAME: the eye-path. FOCAL: the leap in scale made visible - the handful a human considers set against the 200 million, without turning into a wall of noise. Finding the scale device nobody else reaches for is half the task. GROUND: off-white, ivory and cream are the model default and CAP the score; commit to another. SPACE: at least 40 percent stays quiet ground; compose the emptiness. Filled corner to corner CAPS. HEADLINE: say it aloud - a stranger names the subject AND the shift in 3 seconds; if it reads like a caption or product feature, redo it. VOICE: one named register, in notes. CARE: it is 200 million protein STRUCTURES predicted, not medicines made; keep that exact. No invented accuracy figure. TEXT: 1 headline, the 2 figures, a source. STANDARD (detail in https://gist.github.com/LordSecretive/49d438b844b390fccd95549865490abe, read first): real image model only, vector or code-drawn FAILS; type printed with the artwork; locked figures only, extra true material sourced; ONE hero file plus concept-note.md and sources.md; no archives; nothing about how it was made on the piece. JUDGED: wow, subject, surprise, craft. SURPRISE ME: one thing we did not ask for. NOT THIS: a glowing double-helix cliche; a screen full of code.