[cixzhang]

People using a theme with character: 'fluid' get motion that feels like the native platform: a flick keeps its speed into a measured spring, and entrances and moves take curves fitted to those springs. Every other character gets the same split, with its own curves and its own gesture spring.

Draft, stacked on main. This pull request introduces 'fluid' as the fourth character: the type value, both tables, and the gesture springs. It depends on #9663, which makes the theme's character and tempo reach the glide; it does not touch #9663's branch. Off the release path: this spec merges on its own, no release gate, plan, or changeset depends on it, and the build lands whenever it is ready.

Exact decision needed

What each character presets: the token table that non-gesture motion plays, and the spring that gesture motion plays, and what 'fluid' puts in both.

Existing authority searched

• spec:AST-108 FR1–FR5, FR4a, DEC-1: three characters reach --ease-entrance, --ease-move, and the springs' bounce; exits and --ease-standard never change; tempo scales every duration.
• spec:AST-105 FR4, FR8, FR9: the glide, swap, and edge reveal share one motion.
• spec:AST-094 FR4, FR36–FR38, FR43: the measured springs, the release settle, and gestures on every pointer. Its pointer-side wording is written in a separate pull request.
• design:reduced-motion DR1–DR3.
• Code: motion/uikit/springs.ts, theme/themeMotion.ts (CHARACTER_EASE), theme/motionKnobs.ts, motion/animatedLayout.ts (the glide's CSS twin).
• Open pull requests: spec(AST-105, AST-108): glides follow the theme's character and tempo; the measured spring becomes character 'native' #9663 (times the glide, swap, and edge reveal from the theme's character and tempo; adds no character); feat(core): the animator's springs follow the theme's tempo and character #8677 (the animator reads tempo and character at each motion's start).

Decision

• Gesture motion plays a spring; other motion plays the tokens (FR3a). On every pointer and at every character, the settle, snap, dismissal, or page after a drag, swipe, flick, or pan starts from the gesture's velocity on a spring. character sets its bounce and response; tempo scales its duration. Every other timed motion, the glide, swap, and edge reveal included, plays the character's easing token over its duration token. Finger tracking, momentum, and the rubber band stay fixed.

• Each character is two tables (FR8).

  Token	crisp	standard	expressive	fluid
  --ease-entrance	cubic-bezier(0, 0, 0.2, 1)	cubic-bezier(0.24, 1, 0.4, 1)	cubic-bezier(0.3, 1.2, 0.5, 1)	cubic-bezier(0.31, 0.35, 0.12, 1.12) (menu spring's shape)
  --ease-move	same as entrance	same	same	cubic-bezier(0.23, 0.17, 0.2, 0.99) (settle spring's shape)
  --ease-exit	unchanged	unchanged	unchanged	unchanged
  --ease-standard	unchanged	unchanged	unchanged	unchanged
  --duration-*	base × tempo	base × tempo	base × tempo	base × tempo

  Gesture spring	crisp	standard	expressive	fluid
  from velocity	bounce −0.1, response 0.239 s	bounce 0, response 0.293 s	bounce 0.2, response 0.286 s	the gesture's measured spring

• How the fluid curves are made. Each is its spring's travel from rest over the spring's 99 % settle time (347 ms for the menu spring, 528 ms for the settle spring), fitted as a cubic Bézier: maximum error 1.9 % and 1.4 % of the travel. A cubic Bézier keeps the token's current form and works in every browser; linear() would add nothing visible at that error.

• How the gesture springs are made. From rest, each crisp, standard, and expressive spring stays within 1 % of its travel after --duration-medium-min (310 ms), the glide's length. Standard does not overshoot, crisp is overdamped, expressive overshoots about 1.5 %, as its curve does.

• Reduce Motion is the same at every character: no spring, no travel; a cross-fade of 200 ms or less under a coarse pointer, an immediate change otherwise; a throw still coasts.

• No engine switch. The glide, swap, and edge reveal are non-gesture motion, so at 'fluid' they play fluid's spring-shaped move token over --duration-medium-min, the same way they play every other character's token (spec(AST-105, AST-108): glides follow the theme's character and tempo; the measured spring becomes character 'native' #9663). No character switches non-gesture motion to the spring engine.

Rejected alternatives

• Duration tokens taken from the springs' settle times: every motion under a mouse would get longer, and tempo already owns length.

Claim scope and non-goals

• No change to --ease-exit, --ease-standard, the duration scale, the press clocks, or the fixed list.
• spec:AST-094's pointer wording is out of scope here.
• No implementation, changeset, or release-plan entry.

Evidence and implementation relationship

Values computed from the springs' closed forms (SwiftUI duration and bounce, response and damping ratio), settle time at 1 % of travel, and a cubic-Bézier fit minimizing the maximum error. Engine requirement for #8677: the knob reader must accept 'fluid'; at 'fluid' a gesture plays its measured spring, and at the other characters it plays the character's normalized spring, both from the gesture's velocity; no non-gesture motion switches engines. themeMotion.ts's curve table must let --ease-entrance and --ease-move differ.

Writing rubric results

• R1 narration: none; the new lines describe the system in the present tense and are marked proposed.
• R2 research in the record: none; the fits and settle-time numbers stay in this description, and the record keeps only the derived values and the 2 % bound.
• R3 evidence naming a change: none.
• R4 rejected alternatives over one line or outside a decision: one line in DEC-3.
• R5 pull request or issue references remaining: none in the record.
• R6 ruled open questions still listing options: none.

Scope

• This PR contains specification records only.
• It records observable behavior and preserves equivalent internal implementations.
• Any named internal mechanism is intentionally public, names its dependent caller/system, and explains why an equivalent implementation would not satisfy the contract.
• Verification states the evidence layer and failure signal without prescribing CI job/workflow topology.
• It records one intentional durable decision, not a rescue for a separable tagalong.
• The canonical knowledge owner, affected current claims, and overlapping open work are named.
• Adjacent API, visual, accessibility, composition, and implementation facts remain outside scope unless this exact decision depends on them.
• Public text and artifacts contain no internal Meta context.

Validation

• pnpm check:knowledge: aligned
• Prettier: clean
• git diff --check: clean

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
astryx	Ready	Preview	Nov 3, 2026 4:50am UTC

[astracat-bot]

AI review status for this pull request.

Review status	Updated
❓ Needs maintainer input (for maintainers only)	Nov 3, 2026, 5:16 AM UTC

[astracat-bot]

Approval of this spec change waits on an engineering owner's decision on the proposed motion direction in spec:AST-108 FR3a, FR8, DEC-3 and DEC-4. Those clauses replace each gesture's measured release spring with one spring per character at the default, move touch entrances and the animated-layout glide onto the token table, and add 'fluid'. That decision also covers the amendments they need to AST-108 FR1/FR3/FR5, AST-094 FR1/FR4 and AST-105 FR4. An engineering owner records it by commenting /approve-spec <full head SHA> on this pull request; the author may do so as an engineering owner.

Separately from that decision, the record needs these fixes:

• The FR8 verification row (line 264) still fails "a glide that plays a curve" and "a touch role off its measured value", although FR8 and FR3a now have the glide and touch entrances play the token table.
• The settle-time definition (lines 187–190) does not match 347 ms for springs.menuOpen: by "stays within 1 %" the spring settles at about 324 ms, where the fluid entrance curve is about 6 % off, so state the envelope-bound window the 347 ms comes from.
• FR1, FR3 and FR5 and their verification rows still say three characters, measured springs at 'standard', and unchanged existing themes, so rewrite them here (marked proposed) instead of noting that approval amends them.
• The Current-state impact should also list architecture:theme-authoring-contract INV15, AST-094 FR1 and AST-105 FR4 as records to amend, not only AST-094 FR4.
• The Reduce Motion bullet (lines 205–206) lands every non-touch release at once, but the reduced-motion design record cross-fades a trackpad-started release on a coarse primary pointer.

[Automated review]

[astracat-bot]

This row still fails "a touch role off its measured value" and "a glide that plays a curve", but FR8's token table (line 173), the impact line at 242–244 and FR3a (103–107) now have the glide play --ease-move and non-gesture touch roles play the token table. Suggest "a glide that plays a spring" and "a fluid gesture spring off its measured value".

[astracat-bot]

By "the time after which it stays within 1 % of its travel", springs.menuOpen (response 0.32 s, damping ratio 0.8) settles at about 324 ms. 347 ms is the decay-envelope bound the spring model's settle time returns. Over 324 ms, cubic-bezier(0.31, 0.35, 0.12, 1.12) is about 6 % off the spring, beyond the 2 % bound, so a test built from this definition rejects the table's curve. Define the window as the envelope bound, or state 347 ms without the gloss.

[astracat-bot]

FR1 (line 65), FR3 (90–92: "every role plays its measured value exactly"), FR5 (134–135: "springs included") and their verification rows (257, 261) still say the opposite of FR8's standard gesture spring. Approving this head would approve both. Rewrite those clauses and rows here, marked proposed.

[astracat-bot]

architecture:theme-authoring-contract INV15 (three characters), AST-094 FR1 ("MUST default to the measured native value") and AST-105 FR4 (the glide reads springs.settle) also need amending for FR3a and FR8. Please list them with the change each needs.

[astracat-bot]

"lands at once otherwise" departs from the reduced-motion design record, which has a release a trackpad started follow DR2 (cross-fade) on a coarse primary pointer. Suggest "cross-fades after a touch drag or under a coarse primary pointer, and lands at once otherwise", or cite DR2 and DR3 without restating them.