You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Two Characters in genuinely different styles validate the Character Package abstraction
against real variance before the format is published. One style would only prove the
format fits itself.
Scope
Faithful Win95 — 16-color, hard pixels, dithering.
Modern pixel art — larger palette, smoother animation.
Each ships the eight required animations, a Personality Prompt, and declared Behaviors.
The two must move and react differently enough that switching feels like a different
companion rather than a reskin.
Acceptance criteria
Both packages load through the same loader with no special-casing.
Both are playable with the static Director alone.
Switching between them visibly changes idle life, not just artwork.
Neither package requires a Primitive that the other cannot use.
Tests
Both packages load and validate in Engine tests. Visual quality is judged by hand.
Spec: docs/SPEC.md · migrated from docs/issues/0009-shipped-characters.md
Verified on main (b0a6e67) by running each Character for 110 seconds with AI_BUDDY_TRACE_FRAMES=1 and reading what the Engine actually played.
The third criterion is the one that needed the Static Director. When #81 shipped the two Characters, their Behaviors differed on paper but none of them ever played: a Behavior reaches the Engine only as WorldSnapshot::proposal and nothing produced one. #82 landed the Director and gave BMO and Nim their weights, so they are selectable — StaticDirector::propose treats a Behavior with no trigger as always eligible (director.rs, None => true).
Both played every Behavior they declare.
BMO: fidget, greet x2, patrol, report x2, walk
Nim: amble, doze x2, greet, mumble, stretch x2
And the lives are visibly different, not just the drawing. Counting only animations a Behavior drove — the sprite Grounded, playing something other than idle, so the Engine's own State-driven animation is excluded:
driven by a Behavior
frames spent settling
BMO
land, react, talk, walk
0
Nim
land, sit, sleep, talk, walk, react
403
BMO never sits or sleeps of its own accord; Nim does both. The shape of the two runs differs as well: BMO spent 3189 frames Perched and 1341 walking, Nim 5078 Grounded and 4474 idling.
Worth separating from that, because it is easy to miscount: animation_for(State::Perched) is sit, so a perched sprite draws sit whoever it is. BMO logged 2842 such frames and none of them are the Character settling — they are the Engine's animation for standing on a window edge. The table above counts only what a Behavior put on screen.
Switching visibly changes idle life, not just artwork — the table above.
Neither needs a Primitive the other cannot use — structurally true: every Primitive either declares is in the Required Animation Set both supply.
One thing deliberately left open and tracked on #10: neither BMO nor Nim declares when conditions, so they have balance without arc — Nim can propose doze at someone actively typing. #9 does not ask for triggers, so it does not hold this open.
Activity
Replace Chip with BMO
Ship two Characters, not one twice
Claim only what the eye can check today
Replace Chip with BMO
omesser commentedon Aug 30, 2026
Verified on
main(b0a6e67) by running each Character for 110 seconds withAI_BUDDY_TRACE_FRAMES=1and reading what the Engine actually played.The third criterion is the one that needed the Static Director. When #81 shipped the two Characters, their Behaviors differed on paper but none of them ever played: a Behavior reaches the Engine only as
WorldSnapshot::proposaland nothing produced one. #82 landed the Director and gave BMO and Nim their weights, so they are selectable —StaticDirector::proposetreats a Behavior with no trigger as always eligible (director.rs,None => true).Both played every Behavior they declare.
fidget,greetx2,patrol,reportx2,walkamble,dozex2,greet,mumble,stretchx2And the lives are visibly different, not just the drawing. Counting only animations a Behavior drove — the sprite Grounded, playing something other than
idle, so the Engine's own State-driven animation is excluded:land,react,talk,walkland,sit,sleep,talk,walk,reactBMO never sits or sleeps of its own accord; Nim does both. The shape of the two runs differs as well: BMO spent 3189 frames Perched and 1341 walking, Nim 5078 Grounded and 4474 idling.
Worth separating from that, because it is easy to miscount:
animation_for(State::Perched)issit, so a perched sprite drawssitwhoever it is. BMO logged 2842 such frames and none of them are the Character settling — they are the Engine's animation for standing on a window edge. The table above counts only what a Behavior put on screen.All four criteria met:
One thing deliberately left open and tracked on #10: neither BMO nor Nim declares
whenconditions, so they have balance without arc — Nim can proposedozeat someone actively typing. #9 does not ask for triggers, so it does not hold this open.omesser commentedon Aug 30, 2026
All four acceptance criteria met and verified on main. See the run above.
fix(website): Swap Trump's Theatre critic line per copy review