The Seed — atom-first design
The pipeline has two lawful entries. process-intent-to-brief parses a
request into a brief — concept-first, the classic door. This module is the
other door: the project begins with an atom — a small authored thing,
iterated until it’s good in isolation — and the concept is grown out of it.
A protagonist silhouette becomes a platformer; a wistful motif becomes a
farewell game; a card that reads at a glance becomes a deckbuilder. Gamedev
does not always start with a design document. Sometimes it starts with
something you made on a Tuesday and couldn’t put down.
Both doors lead to the same gate: no loop assembly until a spine holds
(process-the-spine). What changes is the direction of travel — intent-first
derives elements from a tension; seed-first discovers the tension inside an
element that already exists.
Why an atom outranks a pitch
The spine’s old ordering law said never twist-first, or you get a pitch in search of a loop. The refined law is never pitch-first — and an atom is not a pitch. A pitch is an unfelt sentence; an atom is felt evidence. It has survived iterations. Someone kept touching it. It carries real information about what is interesting — information no brief can contain, because briefs describe intentions and atoms demonstrate results. That evidence is exactly what earns the atom its entry: you are not designing from a guess, you are designing from the one part of the game that already works.
The method
1 · Author the atom until it’s good alone. In the project, from day zero —
an atom is a module in atoms/, visible in the Workshop the moment it exists
(see the project anatomy in docs/CONVENTIONS.md). Iterate with the knobs, the
vision judge, the audio gates — whichever proof fits its kind. Do not start
concepting from a mediocre atom; the whole method rests on the atom being
genuinely good.
2 · Interrogate what it radiates. An atom that’s good in isolation is radiating something — name it. Ask, in order:
- Fantasy — who do I become when I look at / listen to / play this?
- Verb — what action does it imply? (A silhouette with a lantern implies carrying; a rising motif implies ascending; a card implies committing.)
- Tension — what could make this thing load-bearing? What would it cost, what would it renew, what would oppose it?
Write the answer as one line in the atom’s declaration (radiates:) — the
machine-readable hook the tooling surfaces. If, after honest iteration, the
atom radiates nothing — it’s pretty but implies no fantasy, no verb, no
tension — archive it with a note. A dead seed costs a day; a game grown from
a dead seed costs a month.
3 · Run the spine backwards. process-the-spine normally derives
elements from a tension. Here you invert it: audition tensions in which this
atom is structural, not decorative. For each candidate spine, ask the seed
gate:
If I removed this atom from this spine, would the spine survive unchanged?
If yes, the atom is decoration in that spine — the exact failure
antipattern-decoration names — and you keep auditioning. You want the
spine where the atom sits inside the tension: the lantern-carrier’s flame IS
the scarcity; the motif’s swell IS the renewal beat; the card’s cost IS the
spend. From there the spine runs forward as normal — objective · superpower ·
scarcity · obstacle · renewal, the gate, the resonance table.
4 · Write the concept into the Timeline. When a spine holds, open the
project’s TIMELINE.md and write the Original Concept as its first entry —
dated, honest, and allowed to stale. The Timeline is a log, not a contract:
later pivots are appended as new entries, never edited into the original. The
concept you write today is the record of what you meant, not a rail the
project must run on.
5 · Proceed as usual. Compose, shape, handoff — the rest of the pipeline
(process-refine-and-handoff) neither knows nor cares which door you came
through.
The pivot rule
Before loop assembly, atoms outrank concepts: when iterating an atom teaches you something the concept can’t absorb — the character you can’t nail gives way to a cooler one that implies a different game — the concept yields. Re-run this module from the new atom and append the pivot to the Timeline. This is why atom iteration happens before assembly: it is when pivots cost a paragraph instead of a rewrite.
After loop assembly, the ranking flips: the spine outranks atoms. A new atom that fights the shipped tension is scope creep wearing inspiration’s clothes — park it in the Timeline’s Future section for the next game.
Traps
- Pitch smuggling. “My atom is the idea of a game where…” — no. An atom is
authored and iterated, not described. If you can’t put it in the Workshop,
it’s a pitch, and pitches enter through
process-intent-to-brief. - Polish as procrastination. Six atoms, no spine, no game. The Workshop’s missing Play tab is the tell. The discipline: every atom names what it radiates or gets archived; the seed exists to grow, not to be tended.
- Decoration marriage. Forcing the atom into a spine it doesn’t bear. Run the seed gate honestly — a good atom in the wrong spine makes a worse game than no atom at all.
- Concept archaeology. Treating the Timeline’s first entry as binding. It’s a dated record of a past belief; the log grows, the project moves.