docs(amicode): ADR 0002 — the Brain becomes the Chat's living background

packages/app/CONTEXT.md

@@ -22,6 +22,18 @@ other surface opens beside or in front of the Chat; the Chat is never replaced b
 another surface, only overlaid. There is no view in which the Chat is absent.
 _Avoid_: Main view, editor, session pane (as the whole hub)
 
+**Brain**:
+The data-true living background of the **Chat** — a map of the current session's thought that
+fills the Chat behind every component and is never interactive. One per window, present wherever
+the Chat is (Landing included) and nowhere else.
+_Avoid_: Background, wallpaper, atmosphere, brain strip, canvas
+
+**Glass**:
+The frosted, translucent surface every chat component floats on above the **Brain** — tinted
+just enough that text stays legible over the moving graph. (How the blur and tint are derived
+is an ADR-0002 concern, not a glossary one.)
+_Avoid_: Card background, frosted panel, blur layer, tier
+
 **Rail**:
 The one narrow vertical strip of icons along the window's leading edge. Collapsed to
 icons by default; each icon reveals its text label on hover. The Rail is the only
@@ -68,6 +80,9 @@ only surfaces its state), setup wall
 ## Relationships
 
 - The **Landing** is a **Chat**; the window never opens on anything else.
+- The **Brain** is the background of the **Chat** and only the Chat; every chat component
+ floats on **Glass** above it, while the **Rail**, titlebar, and **Panels** stay solid and
+ frame the living pane.
 - Every **Rail surface** opens as a **Panel** beside the **Chat**; the Chat is never
  replaced, only overlaid, and at most one Panel is open at a time.
 - The **Rail** carries navigation surfaces above and the **Account zone** below; both
@@ -100,3 +115,6 @@ reader meeting them in older code or notes should map them forward:
  a possible future, not part of this redesign. See ADR 0001.
 - **Home cards** "Meet Amico" and "Jump back in" → folded into the first-run Landing and
  the Chats surface respectively; they are no longer standalone surfaces.
+- **Inline brain strip** (the Brain as a single row inside the message timeline) → absorbed
+ into the **Brain** background. One render loop, promoted from a timeline row to the room;
+ the timeline keeps only its text shimmer. See ADR 0002.

packages/app/docs/adr/0002-brain-as-chat-background.md

@@ -0,0 +1,125 @@
+# 0002 — The Brain becomes the Chat's living background; components float on legibility-floored glass
+
+- Status: Proposed
+- Date: 2026-07-24
+- Deciders: Kate Bonner (Head of Product)
+- Tracking: harmoniqs/opencode#56
+
+## Context
+
+The amico **Brain** — the data-true, live map of a session's thought — currently renders
+as an inline strip *inside* the message timeline (a native canvas engine; the `/brain.html`
+iframe was already retired). It is one row among many, easy to scroll past, and it competes
+with the messages rather than framing them.
+
+An earlier redesign (`kate/chat-brain-atmosphere`, six commits, never pushed or merged)
+explored promoting the Brain to a chat-wide background behind *tiered per-bubble glass*
+("atmosphere"), reaching a resolved design and a Phase-5 build. Kate reopened the design
+fresh rather than continuing that branch. The open question: does the Brain stay an inline
+timeline element, or become the **permanent background of the Chat** with every component
+floating on frosted glass — and if so, how is legibility *guaranteed* when text sits on a
+translucent surface over a moving graph? "Movement comes through" (translucency) and
+"highly legible" (contrast) are in direct tension, and the earlier build hit exactly one
+legibility failure: muted grey text not clearing WCAG-AA on its most translucent tier.
+
+This decision governs the **Amicode app UI** (the opencode fork). It sits under ADR 0001
+(chat-first shell): the Chat is the one persistent hub, and this ADR settles what the Chat
+*looks like*.
+
+## Decision
+
+**The Brain becomes the permanent, full-bleed background of the Chat** — Landing,
+conversation, and composer all sit on it. There is **one Brain**: the inline timeline strip
+is absorbed into the background (one render loop, promoted from a timeline row to the room).
+The **Rail, titlebar, and Panels stay solid**, framing the living pane. The Brain is a fixed,
+viewport-anchored backdrop; cards scroll over a stationary Brain.
+
+**Every chat component floats as a frosted-glass card, and legibility is guaranteed by
+construction — not by eye:**
+
+- **Blur** — a *high, constant* gaussian blur on every card. It provides perceptual calm and
+ kills the high-frequency motion detail that hurts reading. It is cheap (no per-frame logic)
+ and it is *not* what makes text legible.
+- **Tint** — contrast comes entirely from a tint whose **opacity is pinned to clear the WCAG-AA
+ floor (with a small safety margin, not exactly at the floor) over a fixed reference frame** —
+ a synthetic Brain at peak bloom in which the sampled backdrop is the brightest palette value
+ **the Brain actually paints in that theme** (peak `#fff676` on dark; peak `#8f8000` on light —
+ the light-mode Brain never paints `#fff676`, so validating against it there would derive a
+ near-transparent tint that silently fails over the real light frame). Blur smooths the backdrop but does not darken it; a card over a bright-yellow
+ cluster is still on bright yellow until the tint pulls the composite toward the ink. The tint
+ set is **keyed to the chat's own theme** — the chat runs a theme independent of the app shell
+ (default oc-2) — so a chat-theme change re-keys the Glass and re-validates contrast. **Two
+ tints:** *standard* (prose, bubbles, composer) holds body text ≥4.5:1; *dense* (code, diffs,
+ run-plots) is more opaque and holds code/diff text ≥4.5:1 and graphical marks — syntax colors,
+ diff fills, plot lines — ≥3:1 (WCAG 1.4.11), so colored content stays crisp.
+
+**The Brain is data-true with an adaptive heartbeat.** Nodes light up as amico reads files,
+runs tools, and touches the vault; the graph grows with the conversation, **densifying in place
+within the fixed, viewport-anchored frame** — bounded by a node cap / recency window so a long
+session neither scrolls the Brain nor grows its per-frame cost without limit. Motion follows
+activity: a lively tempo when amico is working, ~8fps breathing at rest, and a **hard pause**
+when the window is hidden or `prefers-reduced-motion` is set. At rest it shows a **sparse
+seed** — a near-empty Landing whose graph grows *only* from the current session's real
+thought, with **no cross-session persistence** (a fresh session opens nearly empty). One Brain
+per window shows the **active** session; switching the titlebar session tab swaps the Brain to
+that session's graph.
+
+**Pre-agreed perf fallback (non-negotiable mid-build):** the frosted look is protected. At
+rest, full blur and full motion always apply. A frame-time monitor trips the ease when p95
+frame time stays over budget for a sustained window (with hysteresis before restoring, to
+avoid oscillation): the Brain's *motion* eases (calm/slow) and returns to full tempo at rest.
+If easing motion to a full stop still misses budget, the Brain **hard-pauses to a static
+blurred field** — the terminal valve — but **blur fidelity never degrades.**
+
+**Build by reuse.** A new, well-named branch off `local/amicode` reuses the de-risked native
+plumbing — the native Brain engine, the glass-token generator, and the mount wiring from the
+earlier branch (design-neutral, already hardened by an adversarial port-fidelity review) — and
+builds the *design* layer fresh. No iframe. Both light and dark themes are first-class from day
+one.
+
+See `packages/app/CONTEXT.md` for the new vocabulary (**Brain**, **Glass**).
+
+## Consequences
+
+- The **inline brain strip** is retired as a timeline row; the timeline keeps only its text
+ shimmer (the existing working indicator — the cycling gerund + live elapsed shown while a turn
+ works). One Brain, one render loop.
+- **No node-level interactivity is lost.** The strip's only affordance was a click-to-expand/
+ collapse toggle (it was a collapsible row); the Brain engine has no node hit-testing. As an
+ always-present background the expand toggle is obviated, not removed from a capable surface.
+- When a solid **Panel** covers the Chat, the Brain need not animate behind the occluded area —
+ a cheap perf saving, since Panels are opaque.
+- The *standard* tint's opacity is **bounded below** by the contrast floor — it cannot be made
+ ultra-transparent, or muted grey text fails AA over the reference frame (the exact
+ failure the earlier build hit). Secondary/muted text therefore rides the *dense* tint or a
+ locally-dimmed zone.
+- The Chat's cost profile changes: high-radius backdrop-blur recomputed per frame over an
+ animating canvas, across many cards, while scrolling, is the expensive combination. It must
+ pass a **laptop-webview 60fps gate** (the real target — headless Chromium is only a proxy);
+ the motion-ease fallback is the release valve.
+- The **Rail, titlebar, and Panels are deliberately excluded.** Nav chrome stays rock-solid,
+ and dense Panel lists (Chats history, Run gallery) don't fight a moving background.
+- The Landing (a fresh, empty Chat per ADR 0001) now opens on the sparse-seed Brain rather
+ than a blank surface — the window's first impression is the living background, calm and
+ nearly empty, not a widget wall.
+
+## Alternatives considered
+
+- **Keep the inline brain strip** (a timeline row). Rejected: it is one element among many;
+ the redesign's intent is the Brain as the room, not a row.
+- **One calm reading lane / unified glass sheet** (Brain only in the margins). Rejected in
+ favor of per-component floating cards — motion lives in the gaps between cards, which is
+ where "movement comes through" *without* the cards themselves having to be see-through to it.
+- **Dynamic contrast glass** (each card samples its backdrop per frame and adapts its tint).
+ Rejected: per-frame readback is the biggest perf risk and AA is hard to prove against every
+ frame; the static AA-floor tint is deterministic and cheap.
+- **Quiet zone behind cards** (dim/slow the Brain under each card's footprint via per-card
+ masking). Rejected: real engine complexity for translucency the gaps already provide.
+- **Breathing skeleton atlas at rest** (the full latent graph, dim, breathing). Rejected in
+ favor of the sparse seed — calmer, more literal, and it lets the floating cards dominate.
+- **Protect motion and ease blur under load**, or **degrade only while scrolling**. Rejected:
+ the frosted look is the aesthetic being protected; motion is the acceptable give under load.
+- **Rebuild the engine clean**, or **evolve the old `kate/chat-brain-atmosphere` branch**.
+ Rejected: reuse the de-risked plumbing (including its already-fixed port defects), but build
+ the design fresh on a new branch rather than inherit the old design structure or re-pay for
+ the engine port.

packages/app/CONTEXT.md

@@ -22,6 +22,18 @@ other surface opens beside or in front of the Chat; the Chat is never replaced b
 another surface, only overlaid. There is no view in which the Chat is absent.
 _Avoid_: Main view, editor, session pane (as the whole hub)
 
+**Brain**:
+The data-true living background of the **Chat** — a map of the current session's thought that
+fills the Chat behind every component and is never interactive. One per window, present wherever
+the Chat is (Landing included) and nowhere else.
+_Avoid_: Background, wallpaper, atmosphere, brain strip, canvas
+
+**Glass**:
+The frosted, translucent surface every chat component floats on above the **Brain** — tinted
+just enough that text stays legible over the moving graph. (How the blur and tint are derived
+is an ADR-0002 concern, not a glossary one.)
+_Avoid_: Card background, frosted panel, blur layer, tier
+
 **Rail**:
 The one narrow vertical strip of icons along the window's leading edge. Collapsed to
 icons by default; each icon reveals its text label on hover. The Rail is the only
@@ -68,6 +80,9 @@ only surfaces its state), setup wall
 ## Relationships
 
 - The **Landing** is a **Chat**; the window never opens on anything else.
+- The **Brain** is the background of the **Chat** and only the Chat; every chat component
+ floats on **Glass** above it, while the **Rail**, titlebar, and **Panels** stay solid and
+ frame the living pane.
 - Every **Rail surface** opens as a **Panel** beside the **Chat**; the Chat is never
  replaced, only overlaid, and at most one Panel is open at a time.
 - The **Rail** carries navigation surfaces above and the **Account zone** below; both
@@ -100,3 +115,6 @@ reader meeting them in older code or notes should map them forward:
  a possible future, not part of this redesign. See ADR 0001.
 - **Home cards** "Meet Amico" and "Jump back in" → folded into the first-run Landing and
  the Chats surface respectively; they are no longer standalone surfaces.
+- **Inline brain strip** (the Brain as a single row inside the message timeline) → absorbed
+ into the **Brain** background. One render loop, promoted from a timeline row to the room;
+ the timeline keeps only its text shimmer. See ADR 0002.

packages/app/docs/adr/0002-brain-as-chat-background.md

@@ -0,0 +1,125 @@
+# 0002 — The Brain becomes the Chat's living background; components float on legibility-floored glass
+
+- Status: Proposed
+- Date: 2026-07-24
+- Deciders: Kate Bonner (Head of Product)
+- Tracking: harmoniqs/opencode#56
+
+## Context
+
+The amico **Brain** — the data-true, live map of a session's thought — currently renders
+as an inline strip *inside* the message timeline (a native canvas engine; the `/brain.html`
+iframe was already retired). It is one row among many, easy to scroll past, and it competes
+with the messages rather than framing them.
+
+An earlier redesign (`kate/chat-brain-atmosphere`, six commits, never pushed or merged)
+explored promoting the Brain to a chat-wide background behind *tiered per-bubble glass*
+("atmosphere"), reaching a resolved design and a Phase-5 build. Kate reopened the design
+fresh rather than continuing that branch. The open question: does the Brain stay an inline
+timeline element, or become the **permanent background of the Chat** with every component
+floating on frosted glass — and if so, how is legibility *guaranteed* when text sits on a
+translucent surface over a moving graph? "Movement comes through" (translucency) and
+"highly legible" (contrast) are in direct tension, and the earlier build hit exactly one
+legibility failure: muted grey text not clearing WCAG-AA on its most translucent tier.
+
+This decision governs the **Amicode app UI** (the opencode fork). It sits under ADR 0001
+(chat-first shell): the Chat is the one persistent hub, and this ADR settles what the Chat
+*looks like*.
+
+## Decision
+
+**The Brain becomes the permanent, full-bleed background of the Chat** — Landing,
+conversation, and composer all sit on it. There is **one Brain**: the inline timeline strip
+is absorbed into the background (one render loop, promoted from a timeline row to the room).
+The **Rail, titlebar, and Panels stay solid**, framing the living pane. The Brain is a fixed,
+viewport-anchored backdrop; cards scroll over a stationary Brain.
+
+**Every chat component floats as a frosted-glass card, and legibility is guaranteed by
+construction — not by eye:**
+
+- **Blur** — a *high, constant* gaussian blur on every card. It provides perceptual calm and
+ kills the high-frequency motion detail that hurts reading. It is cheap (no per-frame logic)
+ and it is *not* what makes text legible.
+- **Tint** — contrast comes entirely from a tint whose **opacity is pinned to clear the WCAG-AA
+ floor (with a small safety margin, not exactly at the floor) over a fixed reference frame** —
+ a synthetic Brain at peak bloom in which the sampled backdrop is the brightest palette value
+ **the Brain actually paints in that theme** (peak `#fff676` on dark; peak `#8f8000` on light —
+ the light-mode Brain never paints `#fff676`, so validating against it there would derive a
+ near-transparent tint that silently fails over the real light frame). Blur smooths the backdrop but does not darken it; a card over a bright-yellow
+ cluster is still on bright yellow until the tint pulls the composite toward the ink. The tint
+ set is **keyed to the chat's own theme** — the chat runs a theme independent of the app shell
+ (default oc-2) — so a chat-theme change re-keys the Glass and re-validates contrast. **Two
+ tints:** *standard* (prose, bubbles, composer) holds body text ≥4.5:1; *dense* (code, diffs,
+ run-plots) is more opaque and holds code/diff text ≥4.5:1 and graphical marks — syntax colors,
+ diff fills, plot lines — ≥3:1 (WCAG 1.4.11), so colored content stays crisp.
+
+**The Brain is data-true with an adaptive heartbeat.** Nodes light up as amico reads files,
+runs tools, and touches the vault; the graph grows with the conversation, **densifying in place
+within the fixed, viewport-anchored frame** — bounded by a node cap / recency window so a long
+session neither scrolls the Brain nor grows its per-frame cost without limit. Motion follows
+activity: a lively tempo when amico is working, ~8fps breathing at rest, and a **hard pause**
+when the window is hidden or `prefers-reduced-motion` is set. At rest it shows a **sparse
+seed** — a near-empty Landing whose graph grows *only* from the current session's real
+thought, with **no cross-session persistence** (a fresh session opens nearly empty). One Brain
+per window shows the **active** session; switching the titlebar session tab swaps the Brain to
+that session's graph.
+
+**Pre-agreed perf fallback (non-negotiable mid-build):** the frosted look is protected. At
+rest, full blur and full motion always apply. A frame-time monitor trips the ease when p95
+frame time stays over budget for a sustained window (with hysteresis before restoring, to
+avoid oscillation): the Brain's *motion* eases (calm/slow) and returns to full tempo at rest.
+If easing motion to a full stop still misses budget, the Brain **hard-pauses to a static
+blurred field** — the terminal valve — but **blur fidelity never degrades.**
+
+**Build by reuse.** A new, well-named branch off `local/amicode` reuses the de-risked native
+plumbing — the native Brain engine, the glass-token generator, and the mount wiring from the
+earlier branch (design-neutral, already hardened by an adversarial port-fidelity review) — and
+builds the *design* layer fresh. No iframe. Both light and dark themes are first-class from day
+one.
+
+See `packages/app/CONTEXT.md` for the new vocabulary (**Brain**, **Glass**).
+
+## Consequences
+
+- The **inline brain strip** is retired as a timeline row; the timeline keeps only its text
+ shimmer (the existing working indicator — the cycling gerund + live elapsed shown while a turn
+ works). One Brain, one render loop.
+- **No node-level interactivity is lost.** The strip's only affordance was a click-to-expand/
+ collapse toggle (it was a collapsible row); the Brain engine has no node hit-testing. As an
+ always-present background the expand toggle is obviated, not removed from a capable surface.
+- When a solid **Panel** covers the Chat, the Brain need not animate behind the occluded area —
+ a cheap perf saving, since Panels are opaque.
+- The *standard* tint's opacity is **bounded below** by the contrast floor — it cannot be made
+ ultra-transparent, or muted grey text fails AA over the reference frame (the exact
+ failure the earlier build hit). Secondary/muted text therefore rides the *dense* tint or a
+ locally-dimmed zone.
+- The Chat's cost profile changes: high-radius backdrop-blur recomputed per frame over an
+ animating canvas, across many cards, while scrolling, is the expensive combination. It must
+ pass a **laptop-webview 60fps gate** (the real target — headless Chromium is only a proxy);
+ the motion-ease fallback is the release valve.
+- The **Rail, titlebar, and Panels are deliberately excluded.** Nav chrome stays rock-solid,
+ and dense Panel lists (Chats history, Run gallery) don't fight a moving background.
+- The Landing (a fresh, empty Chat per ADR 0001) now opens on the sparse-seed Brain rather
+ than a blank surface — the window's first impression is the living background, calm and
+ nearly empty, not a widget wall.
+
+## Alternatives considered
+
+- **Keep the inline brain strip** (a timeline row). Rejected: it is one element among many;
+ the redesign's intent is the Brain as the room, not a row.
+- **One calm reading lane / unified glass sheet** (Brain only in the margins). Rejected in
+ favor of per-component floating cards — motion lives in the gaps between cards, which is
+ where "movement comes through" *without* the cards themselves having to be see-through to it.
+- **Dynamic contrast glass** (each card samples its backdrop per frame and adapts its tint).
+ Rejected: per-frame readback is the biggest perf risk and AA is hard to prove against every
+ frame; the static AA-floor tint is deterministic and cheap.
+- **Quiet zone behind cards** (dim/slow the Brain under each card's footprint via per-card
+ masking). Rejected: real engine complexity for translucency the gaps already provide.
+- **Breathing skeleton atlas at rest** (the full latent graph, dim, breathing). Rejected in
+ favor of the sparse seed — calmer, more literal, and it lets the floating cards dominate.
+- **Protect motion and ease blur under load**, or **degrade only while scrolling**. Rejected:
+ the frosted look is the aesthetic being protected; motion is the acceptable give under load.
+- **Rebuild the engine clean**, or **evolve the old `kate/chat-brain-atmosphere` branch**.
+ Rejected: reuse the de-risked plumbing (including its already-fixed port defects), but build
+ the design fresh on a new branch rather than inherit the old design structure or re-pay for
+ the engine port.

packages/app/CONTEXT.md

@@ -22,6 +22,18 @@ other surface opens beside or in front of the Chat; the Chat is never replaced b
 another surface, only overlaid. There is no view in which the Chat is absent.
 _Avoid_: Main view, editor, session pane (as the whole hub)
 
+**Brain**:
+The data-true living background of the **Chat** — a map of the current session's thought that
+fills the Chat behind every component and is never interactive. One per window, present wherever
+the Chat is (Landing included) and nowhere else.
+_Avoid_: Background, wallpaper, atmosphere, brain strip, canvas
+
+**Glass**:
+The frosted, translucent surface every chat component floats on above the **Brain** — tinted
+just enough that text stays legible over the moving graph. (How the blur and tint are derived
+is an ADR-0002 concern, not a glossary one.)
+_Avoid_: Card background, frosted panel, blur layer, tier
+
 **Rail**:
 The one narrow vertical strip of icons along the window's leading edge. Collapsed to
 icons by default; each icon reveals its text label on hover. The Rail is the only
@@ -68,6 +80,9 @@ only surfaces its state), setup wall
 ## Relationships
 
 - The **Landing** is a **Chat**; the window never opens on anything else.
+- The **Brain** is the background of the **Chat** and only the Chat; every chat component
+ floats on **Glass** above it, while the **Rail**, titlebar, and **Panels** stay solid and
+ frame the living pane.
 - Every **Rail surface** opens as a **Panel** beside the **Chat**; the Chat is never
  replaced, only overlaid, and at most one Panel is open at a time.
 - The **Rail** carries navigation surfaces above and the **Account zone** below; both
@@ -100,3 +115,6 @@ reader meeting them in older code or notes should map them forward:
  a possible future, not part of this redesign. See ADR 0001.
 - **Home cards** "Meet Amico" and "Jump back in" → folded into the first-run Landing and
  the Chats surface respectively; they are no longer standalone surfaces.
+- **Inline brain strip** (the Brain as a single row inside the message timeline) → absorbed
+ into the **Brain** background. One render loop, promoted from a timeline row to the room;
+ the timeline keeps only its text shimmer. See ADR 0002.

packages/app/docs/adr/0002-brain-as-chat-background.md

@@ -0,0 +1,125 @@
+# 0002 — The Brain becomes the Chat's living background; components float on legibility-floored glass
+
+- Status: Proposed
+- Date: 2026-07-24
+- Deciders: Kate Bonner (Head of Product)
+- Tracking: harmoniqs/opencode#56
+
+## Context
+
+The amico **Brain** — the data-true, live map of a session's thought — currently renders
+as an inline strip *inside* the message timeline (a native canvas engine; the `/brain.html`
+iframe was already retired). It is one row among many, easy to scroll past, and it competes
+with the messages rather than framing them.
+
+An earlier redesign (`kate/chat-brain-atmosphere`, six commits, never pushed or merged)
+explored promoting the Brain to a chat-wide background behind *tiered per-bubble glass*
+("atmosphere"), reaching a resolved design and a Phase-5 build. Kate reopened the design
+fresh rather than continuing that branch. The open question: does the Brain stay an inline
+timeline element, or become the **permanent background of the Chat** with every component
+floating on frosted glass — and if so, how is legibility *guaranteed* when text sits on a
+translucent surface over a moving graph? "Movement comes through" (translucency) and
+"highly legible" (contrast) are in direct tension, and the earlier build hit exactly one
+legibility failure: muted grey text not clearing WCAG-AA on its most translucent tier.
+
+This decision governs the **Amicode app UI** (the opencode fork). It sits under ADR 0001
+(chat-first shell): the Chat is the one persistent hub, and this ADR settles what the Chat
+*looks like*.
+
+## Decision
+
+**The Brain becomes the permanent, full-bleed background of the Chat** — Landing,
+conversation, and composer all sit on it. There is **one Brain**: the inline timeline strip
+is absorbed into the background (one render loop, promoted from a timeline row to the room).
+The **Rail, titlebar, and Panels stay solid**, framing the living pane. The Brain is a fixed,
+viewport-anchored backdrop; cards scroll over a stationary Brain.
+
+**Every chat component floats as a frosted-glass card, and legibility is guaranteed by
+construction — not by eye:**
+
+- **Blur** — a *high, constant* gaussian blur on every card. It provides perceptual calm and
+ kills the high-frequency motion detail that hurts reading. It is cheap (no per-frame logic)
+ and it is *not* what makes text legible.
+- **Tint** — contrast comes entirely from a tint whose **opacity is pinned to clear the WCAG-AA
+ floor (with a small safety margin, not exactly at the floor) over a fixed reference frame** —
+ a synthetic Brain at peak bloom in which the sampled backdrop is the brightest palette value
+ **the Brain actually paints in that theme** (peak `#fff676` on dark; peak `#8f8000` on light —
+ the light-mode Brain never paints `#fff676`, so validating against it there would derive a
+ near-transparent tint that silently fails over the real light frame). Blur smooths the backdrop but does not darken it; a card over a bright-yellow
+ cluster is still on bright yellow until the tint pulls the composite toward the ink. The tint
+ set is **keyed to the chat's own theme** — the chat runs a theme independent of the app shell
+ (default oc-2) — so a chat-theme change re-keys the Glass and re-validates contrast. **Two
+ tints:** *standard* (prose, bubbles, composer) holds body text ≥4.5:1; *dense* (code, diffs,
+ run-plots) is more opaque and holds code/diff text ≥4.5:1 and graphical marks — syntax colors,
+ diff fills, plot lines — ≥3:1 (WCAG 1.4.11), so colored content stays crisp.
+
+**The Brain is data-true with an adaptive heartbeat.** Nodes light up as amico reads files,
+runs tools, and touches the vault; the graph grows with the conversation, **densifying in place
+within the fixed, viewport-anchored frame** — bounded by a node cap / recency window so a long
+session neither scrolls the Brain nor grows its per-frame cost without limit. Motion follows
+activity: a lively tempo when amico is working, ~8fps breathing at rest, and a **hard pause**
+when the window is hidden or `prefers-reduced-motion` is set. At rest it shows a **sparse
+seed** — a near-empty Landing whose graph grows *only* from the current session's real
+thought, with **no cross-session persistence** (a fresh session opens nearly empty). One Brain
+per window shows the **active** session; switching the titlebar session tab swaps the Brain to
+that session's graph.
+
+**Pre-agreed perf fallback (non-negotiable mid-build):** the frosted look is protected. At
+rest, full blur and full motion always apply. A frame-time monitor trips the ease when p95
+frame time stays over budget for a sustained window (with hysteresis before restoring, to
+avoid oscillation): the Brain's *motion* eases (calm/slow) and returns to full tempo at rest.
+If easing motion to a full stop still misses budget, the Brain **hard-pauses to a static
+blurred field** — the terminal valve — but **blur fidelity never degrades.**
+
+**Build by reuse.** A new, well-named branch off `local/amicode` reuses the de-risked native
+plumbing — the native Brain engine, the glass-token generator, and the mount wiring from the
+earlier branch (design-neutral, already hardened by an adversarial port-fidelity review) — and
+builds the *design* layer fresh. No iframe. Both light and dark themes are first-class from day
+one.
+
+See `packages/app/CONTEXT.md` for the new vocabulary (**Brain**, **Glass**).
+
+## Consequences
+
+- The **inline brain strip** is retired as a timeline row; the timeline keeps only its text
+ shimmer (the existing working indicator — the cycling gerund + live elapsed shown while a turn
+ works). One Brain, one render loop.
+- **No node-level interactivity is lost.** The strip's only affordance was a click-to-expand/
+ collapse toggle (it was a collapsible row); the Brain engine has no node hit-testing. As an
+ always-present background the expand toggle is obviated, not removed from a capable surface.
+- When a solid **Panel** covers the Chat, the Brain need not animate behind the occluded area —
+ a cheap perf saving, since Panels are opaque.
+- The *standard* tint's opacity is **bounded below** by the contrast floor — it cannot be made
+ ultra-transparent, or muted grey text fails AA over the reference frame (the exact
+ failure the earlier build hit). Secondary/muted text therefore rides the *dense* tint or a
+ locally-dimmed zone.
+- The Chat's cost profile changes: high-radius backdrop-blur recomputed per frame over an
+ animating canvas, across many cards, while scrolling, is the expensive combination. It must
+ pass a **laptop-webview 60fps gate** (the real target — headless Chromium is only a proxy);
+ the motion-ease fallback is the release valve.
+- The **Rail, titlebar, and Panels are deliberately excluded.** Nav chrome stays rock-solid,
+ and dense Panel lists (Chats history, Run gallery) don't fight a moving background.
+- The Landing (a fresh, empty Chat per ADR 0001) now opens on the sparse-seed Brain rather
+ than a blank surface — the window's first impression is the living background, calm and
+ nearly empty, not a widget wall.
+
+## Alternatives considered
+
+- **Keep the inline brain strip** (a timeline row). Rejected: it is one element among many;
+ the redesign's intent is the Brain as the room, not a row.
+- **One calm reading lane / unified glass sheet** (Brain only in the margins). Rejected in
+ favor of per-component floating cards — motion lives in the gaps between cards, which is
+ where "movement comes through" *without* the cards themselves having to be see-through to it.
+- **Dynamic contrast glass** (each card samples its backdrop per frame and adapts its tint).
+ Rejected: per-frame readback is the biggest perf risk and AA is hard to prove against every
+ frame; the static AA-floor tint is deterministic and cheap.
+- **Quiet zone behind cards** (dim/slow the Brain under each card's footprint via per-card
+ masking). Rejected: real engine complexity for translucency the gaps already provide.
+- **Breathing skeleton atlas at rest** (the full latent graph, dim, breathing). Rejected in
+ favor of the sparse seed — calmer, more literal, and it lets the floating cards dominate.
+- **Protect motion and ease blur under load**, or **degrade only while scrolling**. Rejected:
+ the frosted look is the aesthetic being protected; motion is the acceptable give under load.
+- **Rebuild the engine clean**, or **evolve the old `kate/chat-brain-atmosphere` branch**.
+ Rejected: reuse the de-risked plumbing (including its already-fixed port defects), but build
+ the design fresh on a new branch rather than inherit the old design structure or re-pay for
+ the engine port.

packages/app/CONTEXT.md

@@ -22,6 +22,18 @@ other surface opens beside or in front of the Chat; the Chat is never replaced b
 another surface, only overlaid. There is no view in which the Chat is absent.
 _Avoid_: Main view, editor, session pane (as the whole hub)
 
+**Brain**:
+The data-true living background of the **Chat** — a map of the current session's thought that
+fills the Chat behind every component and is never interactive. One per window, present wherever
+the Chat is (Landing included) and nowhere else.
+_Avoid_: Background, wallpaper, atmosphere, brain strip, canvas
+
+**Glass**:
+The frosted, translucent surface every chat component floats on above the **Brain** — tinted
+just enough that text stays legible over the moving graph. (How the blur and tint are derived
+is an ADR-0002 concern, not a glossary one.)
+_Avoid_: Card background, frosted panel, blur layer, tier
+
 **Rail**:
 The one narrow vertical strip of icons along the window's leading edge. Collapsed to
 icons by default; each icon reveals its text label on hover. The Rail is the only
@@ -68,6 +80,9 @@ only surfaces its state), setup wall
 ## Relationships
 
 - The **Landing** is a **Chat**; the window never opens on anything else.
+- The **Brain** is the background of the **Chat** and only the Chat; every chat component
+ floats on **Glass** above it, while the **Rail**, titlebar, and **Panels** stay solid and
+ frame the living pane.
 - Every **Rail surface** opens as a **Panel** beside the **Chat**; the Chat is never
  replaced, only overlaid, and at most one Panel is open at a time.
 - The **Rail** carries navigation surfaces above and the **Account zone** below; both
@@ -100,3 +115,6 @@ reader meeting them in older code or notes should map them forward:
  a possible future, not part of this redesign. See ADR 0001.
 - **Home cards** "Meet Amico" and "Jump back in" → folded into the first-run Landing and
  the Chats surface respectively; they are no longer standalone surfaces.
+- **Inline brain strip** (the Brain as a single row inside the message timeline) → absorbed
+ into the **Brain** background. One render loop, promoted from a timeline row to the room;
+ the timeline keeps only its text shimmer. See ADR 0002.

packages/app/docs/adr/0002-brain-as-chat-background.md

@@ -0,0 +1,125 @@
+# 0002 — The Brain becomes the Chat's living background; components float on legibility-floored glass
+
+- Status: Proposed
+- Date: 2026-07-24
+- Deciders: Kate Bonner (Head of Product)
+- Tracking: harmoniqs/opencode#56
+
+## Context
+
+The amico **Brain** — the data-true, live map of a session's thought — currently renders
+as an inline strip *inside* the message timeline (a native canvas engine; the `/brain.html`
+iframe was already retired). It is one row among many, easy to scroll past, and it competes
+with the messages rather than framing them.
+
+An earlier redesign (`kate/chat-brain-atmosphere`, six commits, never pushed or merged)
+explored promoting the Brain to a chat-wide background behind *tiered per-bubble glass*
+("atmosphere"), reaching a resolved design and a Phase-5 build. Kate reopened the design
+fresh rather than continuing that branch. The open question: does the Brain stay an inline
+timeline element, or become the **permanent background of the Chat** with every component
+floating on frosted glass — and if so, how is legibility *guaranteed* when text sits on a
+translucent surface over a moving graph? "Movement comes through" (translucency) and
+"highly legible" (contrast) are in direct tension, and the earlier build hit exactly one
+legibility failure: muted grey text not clearing WCAG-AA on its most translucent tier.
+
+This decision governs the **Amicode app UI** (the opencode fork). It sits under ADR 0001
+(chat-first shell): the Chat is the one persistent hub, and this ADR settles what the Chat
+*looks like*.
+
+## Decision
+
+**The Brain becomes the permanent, full-bleed background of the Chat** — Landing,
+conversation, and composer all sit on it. There is **one Brain**: the inline timeline strip
+is absorbed into the background (one render loop, promoted from a timeline row to the room).
+The **Rail, titlebar, and Panels stay solid**, framing the living pane. The Brain is a fixed,
+viewport-anchored backdrop; cards scroll over a stationary Brain.
+
+**Every chat component floats as a frosted-glass card, and legibility is guaranteed by
+construction — not by eye:**
+
+- **Blur** — a *high, constant* gaussian blur on every card. It provides perceptual calm and
+ kills the high-frequency motion detail that hurts reading. It is cheap (no per-frame logic)
+ and it is *not* what makes text legible.
+- **Tint** — contrast comes entirely from a tint whose **opacity is pinned to clear the WCAG-AA
+ floor (with a small safety margin, not exactly at the floor) over a fixed reference frame** —
+ a synthetic Brain at peak bloom in which the sampled backdrop is the brightest palette value
+ **the Brain actually paints in that theme** (peak `#fff676` on dark; peak `#8f8000` on light —
+ the light-mode Brain never paints `#fff676`, so validating against it there would derive a
+ near-transparent tint that silently fails over the real light frame). Blur smooths the backdrop but does not darken it; a card over a bright-yellow
+ cluster is still on bright yellow until the tint pulls the composite toward the ink. The tint
+ set is **keyed to the chat's own theme** — the chat runs a theme independent of the app shell
+ (default oc-2) — so a chat-theme change re-keys the Glass and re-validates contrast. **Two
+ tints:** *standard* (prose, bubbles, composer) holds body text ≥4.5:1; *dense* (code, diffs,
+ run-plots) is more opaque and holds code/diff text ≥4.5:1 and graphical marks — syntax colors,
+ diff fills, plot lines — ≥3:1 (WCAG 1.4.11), so colored content stays crisp.
+
+**The Brain is data-true with an adaptive heartbeat.** Nodes light up as amico reads files,
+runs tools, and touches the vault; the graph grows with the conversation, **densifying in place
+within the fixed, viewport-anchored frame** — bounded by a node cap / recency window so a long
+session neither scrolls the Brain nor grows its per-frame cost without limit. Motion follows
+activity: a lively tempo when amico is working, ~8fps breathing at rest, and a **hard pause**
+when the window is hidden or `prefers-reduced-motion` is set. At rest it shows a **sparse
+seed** — a near-empty Landing whose graph grows *only* from the current session's real
+thought, with **no cross-session persistence** (a fresh session opens nearly empty). One Brain
+per window shows the **active** session; switching the titlebar session tab swaps the Brain to
+that session's graph.
+
+**Pre-agreed perf fallback (non-negotiable mid-build):** the frosted look is protected. At
+rest, full blur and full motion always apply. A frame-time monitor trips the ease when p95
+frame time stays over budget for a sustained window (with hysteresis before restoring, to
+avoid oscillation): the Brain's *motion* eases (calm/slow) and returns to full tempo at rest.
+If easing motion to a full stop still misses budget, the Brain **hard-pauses to a static
+blurred field** — the terminal valve — but **blur fidelity never degrades.**
+
+**Build by reuse.** A new, well-named branch off `local/amicode` reuses the de-risked native
+plumbing — the native Brain engine, the glass-token generator, and the mount wiring from the
+earlier branch (design-neutral, already hardened by an adversarial port-fidelity review) — and
+builds the *design* layer fresh. No iframe. Both light and dark themes are first-class from day
+one.
+
+See `packages/app/CONTEXT.md` for the new vocabulary (**Brain**, **Glass**).
+
+## Consequences
+
+- The **inline brain strip** is retired as a timeline row; the timeline keeps only its text
+ shimmer (the existing working indicator — the cycling gerund + live elapsed shown while a turn
+ works). One Brain, one render loop.
+- **No node-level interactivity is lost.** The strip's only affordance was a click-to-expand/
+ collapse toggle (it was a collapsible row); the Brain engine has no node hit-testing. As an
+ always-present background the expand toggle is obviated, not removed from a capable surface.
+- When a solid **Panel** covers the Chat, the Brain need not animate behind the occluded area —
+ a cheap perf saving, since Panels are opaque.
+- The *standard* tint's opacity is **bounded below** by the contrast floor — it cannot be made
+ ultra-transparent, or muted grey text fails AA over the reference frame (the exact
+ failure the earlier build hit). Secondary/muted text therefore rides the *dense* tint or a
+ locally-dimmed zone.
+- The Chat's cost profile changes: high-radius backdrop-blur recomputed per frame over an
+ animating canvas, across many cards, while scrolling, is the expensive combination. It must
+ pass a **laptop-webview 60fps gate** (the real target — headless Chromium is only a proxy);
+ the motion-ease fallback is the release valve.
+- The **Rail, titlebar, and Panels are deliberately excluded.** Nav chrome stays rock-solid,
+ and dense Panel lists (Chats history, Run gallery) don't fight a moving background.
+- The Landing (a fresh, empty Chat per ADR 0001) now opens on the sparse-seed Brain rather
+ than a blank surface — the window's first impression is the living background, calm and
+ nearly empty, not a widget wall.
+
+## Alternatives considered
+
+- **Keep the inline brain strip** (a timeline row). Rejected: it is one element among many;
+ the redesign's intent is the Brain as the room, not a row.
+- **One calm reading lane / unified glass sheet** (Brain only in the margins). Rejected in
+ favor of per-component floating cards — motion lives in the gaps between cards, which is
+ where "movement comes through" *without* the cards themselves having to be see-through to it.
+- **Dynamic contrast glass** (each card samples its backdrop per frame and adapts its tint).
+ Rejected: per-frame readback is the biggest perf risk and AA is hard to prove against every
+ frame; the static AA-floor tint is deterministic and cheap.
+- **Quiet zone behind cards** (dim/slow the Brain under each card's footprint via per-card
+ masking). Rejected: real engine complexity for translucency the gaps already provide.
+- **Breathing skeleton atlas at rest** (the full latent graph, dim, breathing). Rejected in
+ favor of the sparse seed — calmer, more literal, and it lets the floating cards dominate.
+- **Protect motion and ease blur under load**, or **degrade only while scrolling**. Rejected:
+ the frosted look is the aesthetic being protected; motion is the acceptable give under load.
+- **Rebuild the engine clean**, or **evolve the old `kate/chat-brain-atmosphere` branch**.
+ Rejected: reuse the de-risked plumbing (including its already-fixed port defects), but build
+ the design fresh on a new branch rather than inherit the old design structure or re-pay for
+ the engine port.

packages/app/CONTEXT.md

@@ -22,6 +22,18 @@ other surface opens beside or in front of the Chat; the Chat is never replaced b
 another surface, only overlaid. There is no view in which the Chat is absent.
 _Avoid_: Main view, editor, session pane (as the whole hub)
 
+**Brain**:
+The data-true living background of the **Chat** — a map of the current session's thought that
+fills the Chat behind every component and is never interactive. One per window, present wherever
+the Chat is (Landing included) and nowhere else.
+_Avoid_: Background, wallpaper, atmosphere, brain strip, canvas
+
+**Glass**:
+The frosted, translucent surface every chat component floats on above the **Brain** — tinted
+just enough that text stays legible over the moving graph. (How the blur and tint are derived
+is an ADR-0002 concern, not a glossary one.)
+_Avoid_: Card background, frosted panel, blur layer, tier
+
 **Rail**:
 The one narrow vertical strip of icons along the window's leading edge. Collapsed to
 icons by default; each icon reveals its text label on hover. The Rail is the only
@@ -68,6 +80,9 @@ only surfaces its state), setup wall
 ## Relationships
 
 - The **Landing** is a **Chat**; the window never opens on anything else.
+- The **Brain** is the background of the **Chat** and only the Chat; every chat component
+ floats on **Glass** above it, while the **Rail**, titlebar, and **Panels** stay solid and
+ frame the living pane.
 - Every **Rail surface** opens as a **Panel** beside the **Chat**; the Chat is never
  replaced, only overlaid, and at most one Panel is open at a time.
 - The **Rail** carries navigation surfaces above and the **Account zone** below; both
@@ -100,3 +115,6 @@ reader meeting them in older code or notes should map them forward:
  a possible future, not part of this redesign. See ADR 0001.
 - **Home cards** "Meet Amico" and "Jump back in" → folded into the first-run Landing and
  the Chats surface respectively; they are no longer standalone surfaces.
+- **Inline brain strip** (the Brain as a single row inside the message timeline) → absorbed
+ into the **Brain** background. One render loop, promoted from a timeline row to the room;
+ the timeline keeps only its text shimmer. See ADR 0002.

packages/app/docs/adr/0002-brain-as-chat-background.md

@@ -0,0 +1,125 @@
+# 0002 — The Brain becomes the Chat's living background; components float on legibility-floored glass
+
+- Status: Proposed
+- Date: 2026-07-24
+- Deciders: Kate Bonner (Head of Product)
+- Tracking: harmoniqs/opencode#56
+
+## Context
+
+The amico **Brain** — the data-true, live map of a session's thought — currently renders
+as an inline strip *inside* the message timeline (a native canvas engine; the `/brain.html`
+iframe was already retired). It is one row among many, easy to scroll past, and it competes
+with the messages rather than framing them.
+
+An earlier redesign (`kate/chat-brain-atmosphere`, six commits, never pushed or merged)
+explored promoting the Brain to a chat-wide background behind *tiered per-bubble glass*
+("atmosphere"), reaching a resolved design and a Phase-5 build. Kate reopened the design
+fresh rather than continuing that branch. The open question: does the Brain stay an inline
+timeline element, or become the **permanent background of the Chat** with every component
+floating on frosted glass — and if so, how is legibility *guaranteed* when text sits on a
+translucent surface over a moving graph? "Movement comes through" (translucency) and
+"highly legible" (contrast) are in direct tension, and the earlier build hit exactly one
+legibility failure: muted grey text not clearing WCAG-AA on its most translucent tier.
+
+This decision governs the **Amicode app UI** (the opencode fork). It sits under ADR 0001
+(chat-first shell): the Chat is the one persistent hub, and this ADR settles what the Chat
+*looks like*.
+
+## Decision
+
+**The Brain becomes the permanent, full-bleed background of the Chat** — Landing,
+conversation, and composer all sit on it. There is **one Brain**: the inline timeline strip
+is absorbed into the background (one render loop, promoted from a timeline row to the room).
+The **Rail, titlebar, and Panels stay solid**, framing the living pane. The Brain is a fixed,
+viewport-anchored backdrop; cards scroll over a stationary Brain.
+
+**Every chat component floats as a frosted-glass card, and legibility is guaranteed by
+construction — not by eye:**
+
+- **Blur** — a *high, constant* gaussian blur on every card. It provides perceptual calm and
+ kills the high-frequency motion detail that hurts reading. It is cheap (no per-frame logic)
+ and it is *not* what makes text legible.
+- **Tint** — contrast comes entirely from a tint whose **opacity is pinned to clear the WCAG-AA
+ floor (with a small safety margin, not exactly at the floor) over a fixed reference frame** —
+ a synthetic Brain at peak bloom in which the sampled backdrop is the brightest palette value
+ **the Brain actually paints in that theme** (peak `#fff676` on dark; peak `#8f8000` on light —
+ the light-mode Brain never paints `#fff676`, so validating against it there would derive a
+ near-transparent tint that silently fails over the real light frame). Blur smooths the backdrop but does not darken it; a card over a bright-yellow
+ cluster is still on bright yellow until the tint pulls the composite toward the ink. The tint
+ set is **keyed to the chat's own theme** — the chat runs a theme independent of the app shell
+ (default oc-2) — so a chat-theme change re-keys the Glass and re-validates contrast. **Two
+ tints:** *standard* (prose, bubbles, composer) holds body text ≥4.5:1; *dense* (code, diffs,
+ run-plots) is more opaque and holds code/diff text ≥4.5:1 and graphical marks — syntax colors,
+ diff fills, plot lines — ≥3:1 (WCAG 1.4.11), so colored content stays crisp.
+
+**The Brain is data-true with an adaptive heartbeat.** Nodes light up as amico reads files,
+runs tools, and touches the vault; the graph grows with the conversation, **densifying in place
+within the fixed, viewport-anchored frame** — bounded by a node cap / recency window so a long
+session neither scrolls the Brain nor grows its per-frame cost without limit. Motion follows
+activity: a lively tempo when amico is working, ~8fps breathing at rest, and a **hard pause**
+when the window is hidden or `prefers-reduced-motion` is set. At rest it shows a **sparse
+seed** — a near-empty Landing whose graph grows *only* from the current session's real
+thought, with **no cross-session persistence** (a fresh session opens nearly empty). One Brain
+per window shows the **active** session; switching the titlebar session tab swaps the Brain to
+that session's graph.
+
+**Pre-agreed perf fallback (non-negotiable mid-build):** the frosted look is protected. At
+rest, full blur and full motion always apply. A frame-time monitor trips the ease when p95
+frame time stays over budget for a sustained window (with hysteresis before restoring, to
+avoid oscillation): the Brain's *motion* eases (calm/slow) and returns to full tempo at rest.
+If easing motion to a full stop still misses budget, the Brain **hard-pauses to a static
+blurred field** — the terminal valve — but **blur fidelity never degrades.**
+
+**Build by reuse.** A new, well-named branch off `local/amicode` reuses the de-risked native
+plumbing — the native Brain engine, the glass-token generator, and the mount wiring from the
+earlier branch (design-neutral, already hardened by an adversarial port-fidelity review) — and
+builds the *design* layer fresh. No iframe. Both light and dark themes are first-class from day
+one.
+
+See `packages/app/CONTEXT.md` for the new vocabulary (**Brain**, **Glass**).
+
+## Consequences
+
+- The **inline brain strip** is retired as a timeline row; the timeline keeps only its text
+ shimmer (the existing working indicator — the cycling gerund + live elapsed shown while a turn
+ works). One Brain, one render loop.
+- **No node-level interactivity is lost.** The strip's only affordance was a click-to-expand/
+ collapse toggle (it was a collapsible row); the Brain engine has no node hit-testing. As an
+ always-present background the expand toggle is obviated, not removed from a capable surface.
+- When a solid **Panel** covers the Chat, the Brain need not animate behind the occluded area —
+ a cheap perf saving, since Panels are opaque.
+- The *standard* tint's opacity is **bounded below** by the contrast floor — it cannot be made
+ ultra-transparent, or muted grey text fails AA over the reference frame (the exact
+ failure the earlier build hit). Secondary/muted text therefore rides the *dense* tint or a
+ locally-dimmed zone.
+- The Chat's cost profile changes: high-radius backdrop-blur recomputed per frame over an
+ animating canvas, across many cards, while scrolling, is the expensive combination. It must
+ pass a **laptop-webview 60fps gate** (the real target — headless Chromium is only a proxy);
+ the motion-ease fallback is the release valve.
+- The **Rail, titlebar, and Panels are deliberately excluded.** Nav chrome stays rock-solid,
+ and dense Panel lists (Chats history, Run gallery) don't fight a moving background.
+- The Landing (a fresh, empty Chat per ADR 0001) now opens on the sparse-seed Brain rather
+ than a blank surface — the window's first impression is the living background, calm and
+ nearly empty, not a widget wall.
+
+## Alternatives considered
+
+- **Keep the inline brain strip** (a timeline row). Rejected: it is one element among many;
+ the redesign's intent is the Brain as the room, not a row.
+- **One calm reading lane / unified glass sheet** (Brain only in the margins). Rejected in
+ favor of per-component floating cards — motion lives in the gaps between cards, which is
+ where "movement comes through" *without* the cards themselves having to be see-through to it.
+- **Dynamic contrast glass** (each card samples its backdrop per frame and adapts its tint).
+ Rejected: per-frame readback is the biggest perf risk and AA is hard to prove against every
+ frame; the static AA-floor tint is deterministic and cheap.
+- **Quiet zone behind cards** (dim/slow the Brain under each card's footprint via per-card
+ masking). Rejected: real engine complexity for translucency the gaps already provide.
+- **Breathing skeleton atlas at rest** (the full latent graph, dim, breathing). Rejected in
+ favor of the sparse seed — calmer, more literal, and it lets the floating cards dominate.
+- **Protect motion and ease blur under load**, or **degrade only while scrolling**. Rejected:
+ the frosted look is the aesthetic being protected; motion is the acceptable give under load.
+- **Rebuild the engine clean**, or **evolve the old `kate/chat-brain-atmosphere` branch**.
+ Rejected: reuse the de-risked plumbing (including its already-fixed port defects), but build
+ the design fresh on a new branch rather than inherit the old design structure or re-pay for
+ the engine port.

packages/app/CONTEXT.md

@@ -22,6 +22,18 @@ other surface opens beside or in front of the Chat; the Chat is never replaced b
 another surface, only overlaid. There is no view in which the Chat is absent.
 _Avoid_: Main view, editor, session pane (as the whole hub)
 
+**Brain**:
+The data-true living background of the **Chat** — a map of the current session's thought that
+fills the Chat behind every component and is never interactive. One per window, present wherever
+the Chat is (Landing included) and nowhere else.
+_Avoid_: Background, wallpaper, atmosphere, brain strip, canvas
+
+**Glass**:
+The frosted, translucent surface every chat component floats on above the **Brain** — tinted
+just enough that text stays legible over the moving graph. (How the blur and tint are derived
+is an ADR-0002 concern, not a glossary one.)
+_Avoid_: Card background, frosted panel, blur layer, tier
+
 **Rail**:
 The one narrow vertical strip of icons along the window's leading edge. Collapsed to
 icons by default; each icon reveals its text label on hover. The Rail is the only
@@ -68,6 +80,9 @@ only surfaces its state), setup wall
 ## Relationships
 
 - The **Landing** is a **Chat**; the window never opens on anything else.
+- The **Brain** is the background of the **Chat** and only the Chat; every chat component
+ floats on **Glass** above it, while the **Rail**, titlebar, and **Panels** stay solid and
+ frame the living pane.
 - Every **Rail surface** opens as a **Panel** beside the **Chat**; the Chat is never
  replaced, only overlaid, and at most one Panel is open at a time.
 - The **Rail** carries navigation surfaces above and the **Account zone** below; both
@@ -100,3 +115,6 @@ reader meeting them in older code or notes should map them forward:
  a possible future, not part of this redesign. See ADR 0001.
 - **Home cards** "Meet Amico" and "Jump back in" → folded into the first-run Landing and
  the Chats surface respectively; they are no longer standalone surfaces.
+- **Inline brain strip** (the Brain as a single row inside the message timeline) → absorbed
+ into the **Brain** background. One render loop, promoted from a timeline row to the room;
+ the timeline keeps only its text shimmer. See ADR 0002.

packages/app/docs/adr/0002-brain-as-chat-background.md

@@ -0,0 +1,125 @@
+# 0002 — The Brain becomes the Chat's living background; components float on legibility-floored glass
+
+- Status: Proposed
+- Date: 2026-07-24
+- Deciders: Kate Bonner (Head of Product)
+- Tracking: harmoniqs/opencode#56
+
+## Context
+
+The amico **Brain** — the data-true, live map of a session's thought — currently renders
+as an inline strip *inside* the message timeline (a native canvas engine; the `/brain.html`
+iframe was already retired). It is one row among many, easy to scroll past, and it competes
+with the messages rather than framing them.
+
+An earlier redesign (`kate/chat-brain-atmosphere`, six commits, never pushed or merged)
+explored promoting the Brain to a chat-wide background behind *tiered per-bubble glass*
+("atmosphere"), reaching a resolved design and a Phase-5 build. Kate reopened the design
+fresh rather than continuing that branch. The open question: does the Brain stay an inline
+timeline element, or become the **permanent background of the Chat** with every component
+floating on frosted glass — and if so, how is legibility *guaranteed* when text sits on a
+translucent surface over a moving graph? "Movement comes through" (translucency) and
+"highly legible" (contrast) are in direct tension, and the earlier build hit exactly one
+legibility failure: muted grey text not clearing WCAG-AA on its most translucent tier.
+
+This decision governs the **Amicode app UI** (the opencode fork). It sits under ADR 0001
+(chat-first shell): the Chat is the one persistent hub, and this ADR settles what the Chat
+*looks like*.
+
+## Decision
+
+**The Brain becomes the permanent, full-bleed background of the Chat** — Landing,
+conversation, and composer all sit on it. There is **one Brain**: the inline timeline strip
+is absorbed into the background (one render loop, promoted from a timeline row to the room).
+The **Rail, titlebar, and Panels stay solid**, framing the living pane. The Brain is a fixed,
+viewport-anchored backdrop; cards scroll over a stationary Brain.
+
+**Every chat component floats as a frosted-glass card, and legibility is guaranteed by
+construction — not by eye:**
+
+- **Blur** — a *high, constant* gaussian blur on every card. It provides perceptual calm and
+ kills the high-frequency motion detail that hurts reading. It is cheap (no per-frame logic)
+ and it is *not* what makes text legible.
+- **Tint** — contrast comes entirely from a tint whose **opacity is pinned to clear the WCAG-AA
+ floor (with a small safety margin, not exactly at the floor) over a fixed reference frame** —
+ a synthetic Brain at peak bloom in which the sampled backdrop is the brightest palette value
+ **the Brain actually paints in that theme** (peak `#fff676` on dark; peak `#8f8000` on light —
+ the light-mode Brain never paints `#fff676`, so validating against it there would derive a
+ near-transparent tint that silently fails over the real light frame). Blur smooths the backdrop but does not darken it; a card over a bright-yellow
+ cluster is still on bright yellow until the tint pulls the composite toward the ink. The tint
+ set is **keyed to the chat's own theme** — the chat runs a theme independent of the app shell
+ (default oc-2) — so a chat-theme change re-keys the Glass and re-validates contrast. **Two
+ tints:** *standard* (prose, bubbles, composer) holds body text ≥4.5:1; *dense* (code, diffs,
+ run-plots) is more opaque and holds code/diff text ≥4.5:1 and graphical marks — syntax colors,
+ diff fills, plot lines — ≥3:1 (WCAG 1.4.11), so colored content stays crisp.
+
+**The Brain is data-true with an adaptive heartbeat.** Nodes light up as amico reads files,
+runs tools, and touches the vault; the graph grows with the conversation, **densifying in place
+within the fixed, viewport-anchored frame** — bounded by a node cap / recency window so a long
+session neither scrolls the Brain nor grows its per-frame cost without limit. Motion follows
+activity: a lively tempo when amico is working, ~8fps breathing at rest, and a **hard pause**
+when the window is hidden or `prefers-reduced-motion` is set. At rest it shows a **sparse
+seed** — a near-empty Landing whose graph grows *only* from the current session's real
+thought, with **no cross-session persistence** (a fresh session opens nearly empty). One Brain
+per window shows the **active** session; switching the titlebar session tab swaps the Brain to
+that session's graph.
+
+**Pre-agreed perf fallback (non-negotiable mid-build):** the frosted look is protected. At
+rest, full blur and full motion always apply. A frame-time monitor trips the ease when p95
+frame time stays over budget for a sustained window (with hysteresis before restoring, to
+avoid oscillation): the Brain's *motion* eases (calm/slow) and returns to full tempo at rest.
+If easing motion to a full stop still misses budget, the Brain **hard-pauses to a static
+blurred field** — the terminal valve — but **blur fidelity never degrades.**
+
+**Build by reuse.** A new, well-named branch off `local/amicode` reuses the de-risked native
+plumbing — the native Brain engine, the glass-token generator, and the mount wiring from the
+earlier branch (design-neutral, already hardened by an adversarial port-fidelity review) — and
+builds the *design* layer fresh. No iframe. Both light and dark themes are first-class from day
+one.
+
+See `packages/app/CONTEXT.md` for the new vocabulary (**Brain**, **Glass**).
+
+## Consequences
+
+- The **inline brain strip** is retired as a timeline row; the timeline keeps only its text
+ shimmer (the existing working indicator — the cycling gerund + live elapsed shown while a turn
+ works). One Brain, one render loop.
+- **No node-level interactivity is lost.** The strip's only affordance was a click-to-expand/
+ collapse toggle (it was a collapsible row); the Brain engine has no node hit-testing. As an
+ always-present background the expand toggle is obviated, not removed from a capable surface.
+- When a solid **Panel** covers the Chat, the Brain need not animate behind the occluded area —
+ a cheap perf saving, since Panels are opaque.
+- The *standard* tint's opacity is **bounded below** by the contrast floor — it cannot be made
+ ultra-transparent, or muted grey text fails AA over the reference frame (the exact
+ failure the earlier build hit). Secondary/muted text therefore rides the *dense* tint or a
+ locally-dimmed zone.
+- The Chat's cost profile changes: high-radius backdrop-blur recomputed per frame over an
+ animating canvas, across many cards, while scrolling, is the expensive combination. It must
+ pass a **laptop-webview 60fps gate** (the real target — headless Chromium is only a proxy);
+ the motion-ease fallback is the release valve.
+- The **Rail, titlebar, and Panels are deliberately excluded.** Nav chrome stays rock-solid,
+ and dense Panel lists (Chats history, Run gallery) don't fight a moving background.
+- The Landing (a fresh, empty Chat per ADR 0001) now opens on the sparse-seed Brain rather
+ than a blank surface — the window's first impression is the living background, calm and
+ nearly empty, not a widget wall.
+
+## Alternatives considered
+
+- **Keep the inline brain strip** (a timeline row). Rejected: it is one element among many;
+ the redesign's intent is the Brain as the room, not a row.
+- **One calm reading lane / unified glass sheet** (Brain only in the margins). Rejected in
+ favor of per-component floating cards — motion lives in the gaps between cards, which is
+ where "movement comes through" *without* the cards themselves having to be see-through to it.
+- **Dynamic contrast glass** (each card samples its backdrop per frame and adapts its tint).
+ Rejected: per-frame readback is the biggest perf risk and AA is hard to prove against every
+ frame; the static AA-floor tint is deterministic and cheap.
+- **Quiet zone behind cards** (dim/slow the Brain under each card's footprint via per-card
+ masking). Rejected: real engine complexity for translucency the gaps already provide.
+- **Breathing skeleton atlas at rest** (the full latent graph, dim, breathing). Rejected in
+ favor of the sparse seed — calmer, more literal, and it lets the floating cards dominate.
+- **Protect motion and ease blur under load**, or **degrade only while scrolling**. Rejected:
+ the frosted look is the aesthetic being protected; motion is the acceptable give under load.
+- **Rebuild the engine clean**, or **evolve the old `kate/chat-brain-atmosphere` branch**.
+ Rejected: reuse the de-risked plumbing (including its already-fixed port defects), but build
+ the design fresh on a new branch rather than inherit the old design structure or re-pay for
+ the engine port.

packages/app/CONTEXT.md

@@ -22,6 +22,18 @@ other surface opens beside or in front of the Chat; the Chat is never replaced b
 another surface, only overlaid. There is no view in which the Chat is absent.
 _Avoid_: Main view, editor, session pane (as the whole hub)
 
+**Brain**:
+The data-true living background of the **Chat** — a map of the current session's thought that
+fills the Chat behind every component and is never interactive. One per window, present wherever
+the Chat is (Landing included) and nowhere else.
+_Avoid_: Background, wallpaper, atmosphere, brain strip, canvas
+
+**Glass**:
+The frosted, translucent surface every chat component floats on above the **Brain** — tinted
+just enough that text stays legible over the moving graph. (How the blur and tint are derived
+is an ADR-0002 concern, not a glossary one.)
+_Avoid_: Card background, frosted panel, blur layer, tier
+
 **Rail**:
 The one narrow vertical strip of icons along the window's leading edge. Collapsed to
 icons by default; each icon reveals its text label on hover. The Rail is the only
@@ -68,6 +80,9 @@ only surfaces its state), setup wall
 ## Relationships
 
 - The **Landing** is a **Chat**; the window never opens on anything else.
+- The **Brain** is the background of the **Chat** and only the Chat; every chat component
+ floats on **Glass** above it, while the **Rail**, titlebar, and **Panels** stay solid and
+ frame the living pane.
 - Every **Rail surface** opens as a **Panel** beside the **Chat**; the Chat is never
  replaced, only overlaid, and at most one Panel is open at a time.
 - The **Rail** carries navigation surfaces above and the **Account zone** below; both
@@ -100,3 +115,6 @@ reader meeting them in older code or notes should map them forward:
  a possible future, not part of this redesign. See ADR 0001.
 - **Home cards** "Meet Amico" and "Jump back in" → folded into the first-run Landing and
  the Chats surface respectively; they are no longer standalone surfaces.
+- **Inline brain strip** (the Brain as a single row inside the message timeline) → absorbed
+ into the **Brain** background. One render loop, promoted from a timeline row to the room;
+ the timeline keeps only its text shimmer. See ADR 0002.

packages/app/docs/adr/0002-brain-as-chat-background.md

@@ -0,0 +1,125 @@
+# 0002 — The Brain becomes the Chat's living background; components float on legibility-floored glass
+
+- Status: Proposed
+- Date: 2026-07-24
+- Deciders: Kate Bonner (Head of Product)
+- Tracking: harmoniqs/opencode#56
+
+## Context
+
+The amico **Brain** — the data-true, live map of a session's thought — currently renders
+as an inline strip *inside* the message timeline (a native canvas engine; the `/brain.html`
+iframe was already retired). It is one row among many, easy to scroll past, and it competes
+with the messages rather than framing them.
+
+An earlier redesign (`kate/chat-brain-atmosphere`, six commits, never pushed or merged)
+explored promoting the Brain to a chat-wide background behind *tiered per-bubble glass*
+("atmosphere"), reaching a resolved design and a Phase-5 build. Kate reopened the design
+fresh rather than continuing that branch. The open question: does the Brain stay an inline
+timeline element, or become the **permanent background of the Chat** with every component
+floating on frosted glass — and if so, how is legibility *guaranteed* when text sits on a
+translucent surface over a moving graph? "Movement comes through" (translucency) and
+"highly legible" (contrast) are in direct tension, and the earlier build hit exactly one
+legibility failure: muted grey text not clearing WCAG-AA on its most translucent tier.
+
+This decision governs the **Amicode app UI** (the opencode fork). It sits under ADR 0001
+(chat-first shell): the Chat is the one persistent hub, and this ADR settles what the Chat
+*looks like*.
+
+## Decision
+
+**The Brain becomes the permanent, full-bleed background of the Chat** — Landing,
+conversation, and composer all sit on it. There is **one Brain**: the inline timeline strip
+is absorbed into the background (one render loop, promoted from a timeline row to the room).
+The **Rail, titlebar, and Panels stay solid**, framing the living pane. The Brain is a fixed,
+viewport-anchored backdrop; cards scroll over a stationary Brain.
+
+**Every chat component floats as a frosted-glass card, and legibility is guaranteed by
+construction — not by eye:**
+
+- **Blur** — a *high, constant* gaussian blur on every card. It provides perceptual calm and
+ kills the high-frequency motion detail that hurts reading. It is cheap (no per-frame logic)
+ and it is *not* what makes text legible.
+- **Tint** — contrast comes entirely from a tint whose **opacity is pinned to clear the WCAG-AA
+ floor (with a small safety margin, not exactly at the floor) over a fixed reference frame** —
+ a synthetic Brain at peak bloom in which the sampled backdrop is the brightest palette value
+ **the Brain actually paints in that theme** (peak `#fff676` on dark; peak `#8f8000` on light —
+ the light-mode Brain never paints `#fff676`, so validating against it there would derive a
+ near-transparent tint that silently fails over the real light frame). Blur smooths the backdrop but does not darken it; a card over a bright-yellow
+ cluster is still on bright yellow until the tint pulls the composite toward the ink. The tint
+ set is **keyed to the chat's own theme** — the chat runs a theme independent of the app shell
+ (default oc-2) — so a chat-theme change re-keys the Glass and re-validates contrast. **Two
+ tints:** *standard* (prose, bubbles, composer) holds body text ≥4.5:1; *dense* (code, diffs,
+ run-plots) is more opaque and holds code/diff text ≥4.5:1 and graphical marks — syntax colors,
+ diff fills, plot lines — ≥3:1 (WCAG 1.4.11), so colored content stays crisp.
+
+**The Brain is data-true with an adaptive heartbeat.** Nodes light up as amico reads files,
+runs tools, and touches the vault; the graph grows with the conversation, **densifying in place
+within the fixed, viewport-anchored frame** — bounded by a node cap / recency window so a long
+session neither scrolls the Brain nor grows its per-frame cost without limit. Motion follows
+activity: a lively tempo when amico is working, ~8fps breathing at rest, and a **hard pause**
+when the window is hidden or `prefers-reduced-motion` is set. At rest it shows a **sparse
+seed** — a near-empty Landing whose graph grows *only* from the current session's real
+thought, with **no cross-session persistence** (a fresh session opens nearly empty). One Brain
+per window shows the **active** session; switching the titlebar session tab swaps the Brain to
+that session's graph.
+
+**Pre-agreed perf fallback (non-negotiable mid-build):** the frosted look is protected. At
+rest, full blur and full motion always apply. A frame-time monitor trips the ease when p95
+frame time stays over budget for a sustained window (with hysteresis before restoring, to
+avoid oscillation): the Brain's *motion* eases (calm/slow) and returns to full tempo at rest.
+If easing motion to a full stop still misses budget, the Brain **hard-pauses to a static
+blurred field** — the terminal valve — but **blur fidelity never degrades.**
+
+**Build by reuse.** A new, well-named branch off `local/amicode` reuses the de-risked native
+plumbing — the native Brain engine, the glass-token generator, and the mount wiring from the
+earlier branch (design-neutral, already hardened by an adversarial port-fidelity review) — and
+builds the *design* layer fresh. No iframe. Both light and dark themes are first-class from day
+one.
+
+See `packages/app/CONTEXT.md` for the new vocabulary (**Brain**, **Glass**).
+
+## Consequences
+
+- The **inline brain strip** is retired as a timeline row; the timeline keeps only its text
+ shimmer (the existing working indicator — the cycling gerund + live elapsed shown while a turn
+ works). One Brain, one render loop.
+- **No node-level interactivity is lost.** The strip's only affordance was a click-to-expand/
+ collapse toggle (it was a collapsible row); the Brain engine has no node hit-testing. As an
+ always-present background the expand toggle is obviated, not removed from a capable surface.
+- When a solid **Panel** covers the Chat, the Brain need not animate behind the occluded area —
+ a cheap perf saving, since Panels are opaque.
+- The *standard* tint's opacity is **bounded below** by the contrast floor — it cannot be made
+ ultra-transparent, or muted grey text fails AA over the reference frame (the exact
+ failure the earlier build hit). Secondary/muted text therefore rides the *dense* tint or a
+ locally-dimmed zone.
+- The Chat's cost profile changes: high-radius backdrop-blur recomputed per frame over an
+ animating canvas, across many cards, while scrolling, is the expensive combination. It must
+ pass a **laptop-webview 60fps gate** (the real target — headless Chromium is only a proxy);
+ the motion-ease fallback is the release valve.
+- The **Rail, titlebar, and Panels are deliberately excluded.** Nav chrome stays rock-solid,
+ and dense Panel lists (Chats history, Run gallery) don't fight a moving background.
+- The Landing (a fresh, empty Chat per ADR 0001) now opens on the sparse-seed Brain rather
+ than a blank surface — the window's first impression is the living background, calm and
+ nearly empty, not a widget wall.
+
+## Alternatives considered
+
+- **Keep the inline brain strip** (a timeline row). Rejected: it is one element among many;
+ the redesign's intent is the Brain as the room, not a row.
+- **One calm reading lane / unified glass sheet** (Brain only in the margins). Rejected in
+ favor of per-component floating cards — motion lives in the gaps between cards, which is
+ where "movement comes through" *without* the cards themselves having to be see-through to it.
+- **Dynamic contrast glass** (each card samples its backdrop per frame and adapts its tint).
+ Rejected: per-frame readback is the biggest perf risk and AA is hard to prove against every
+ frame; the static AA-floor tint is deterministic and cheap.
+- **Quiet zone behind cards** (dim/slow the Brain under each card's footprint via per-card
+ masking). Rejected: real engine complexity for translucency the gaps already provide.
+- **Breathing skeleton atlas at rest** (the full latent graph, dim, breathing). Rejected in
+ favor of the sparse seed — calmer, more literal, and it lets the floating cards dominate.
+- **Protect motion and ease blur under load**, or **degrade only while scrolling**. Rejected:
+ the frosted look is the aesthetic being protected; motion is the acceptable give under load.
+- **Rebuild the engine clean**, or **evolve the old `kate/chat-brain-atmosphere` branch**.
+ Rejected: reuse the de-risked plumbing (including its already-fixed port defects), but build
+ the design fresh on a new branch rather than inherit the old design structure or re-pay for
+ the engine port.

packages/app/CONTEXT.md

@@ -22,6 +22,18 @@ other surface opens beside or in front of the Chat; the Chat is never replaced b
 another surface, only overlaid. There is no view in which the Chat is absent.
 _Avoid_: Main view, editor, session pane (as the whole hub)
 
+**Brain**:
+The data-true living background of the **Chat** — a map of the current session's thought that
+fills the Chat behind every component and is never interactive. One per window, present wherever
+the Chat is (Landing included) and nowhere else.
+_Avoid_: Background, wallpaper, atmosphere, brain strip, canvas
+
+**Glass**:
+The frosted, translucent surface every chat component floats on above the **Brain** — tinted
+just enough that text stays legible over the moving graph. (How the blur and tint are derived
+is an ADR-0002 concern, not a glossary one.)
+_Avoid_: Card background, frosted panel, blur layer, tier
+
 **Rail**:
 The one narrow vertical strip of icons along the window's leading edge. Collapsed to
 icons by default; each icon reveals its text label on hover. The Rail is the only
@@ -68,6 +80,9 @@ only surfaces its state), setup wall
 ## Relationships
 
 - The **Landing** is a **Chat**; the window never opens on anything else.
+- The **Brain** is the background of the **Chat** and only the Chat; every chat component
+ floats on **Glass** above it, while the **Rail**, titlebar, and **Panels** stay solid and
+ frame the living pane.
 - Every **Rail surface** opens as a **Panel** beside the **Chat**; the Chat is never
  replaced, only overlaid, and at most one Panel is open at a time.
 - The **Rail** carries navigation surfaces above and the **Account zone** below; both
@@ -100,3 +115,6 @@ reader meeting them in older code or notes should map them forward:
  a possible future, not part of this redesign. See ADR 0001.
 - **Home cards** "Meet Amico" and "Jump back in" → folded into the first-run Landing and
  the Chats surface respectively; they are no longer standalone surfaces.
+- **Inline brain strip** (the Brain as a single row inside the message timeline) → absorbed
+ into the **Brain** background. One render loop, promoted from a timeline row to the room;
+ the timeline keeps only its text shimmer. See ADR 0002.

packages/app/docs/adr/0002-brain-as-chat-background.md

@@ -0,0 +1,125 @@
+# 0002 — The Brain becomes the Chat's living background; components float on legibility-floored glass
+
+- Status: Proposed
+- Date: 2026-07-24
+- Deciders: Kate Bonner (Head of Product)
+- Tracking: harmoniqs/opencode#56
+
+## Context
+
+The amico **Brain** — the data-true, live map of a session's thought — currently renders
+as an inline strip *inside* the message timeline (a native canvas engine; the `/brain.html`
+iframe was already retired). It is one row among many, easy to scroll past, and it competes
+with the messages rather than framing them.
+
+An earlier redesign (`kate/chat-brain-atmosphere`, six commits, never pushed or merged)
+explored promoting the Brain to a chat-wide background behind *tiered per-bubble glass*
+("atmosphere"), reaching a resolved design and a Phase-5 build. Kate reopened the design
+fresh rather than continuing that branch. The open question: does the Brain stay an inline
+timeline element, or become the **permanent background of the Chat** with every component
+floating on frosted glass — and if so, how is legibility *guaranteed* when text sits on a
+translucent surface over a moving graph? "Movement comes through" (translucency) and
+"highly legible" (contrast) are in direct tension, and the earlier build hit exactly one
+legibility failure: muted grey text not clearing WCAG-AA on its most translucent tier.
+
+This decision governs the **Amicode app UI** (the opencode fork). It sits under ADR 0001
+(chat-first shell): the Chat is the one persistent hub, and this ADR settles what the Chat
+*looks like*.
+
+## Decision
+
+**The Brain becomes the permanent, full-bleed background of the Chat** — Landing,
+conversation, and composer all sit on it. There is **one Brain**: the inline timeline strip
+is absorbed into the background (one render loop, promoted from a timeline row to the room).
+The **Rail, titlebar, and Panels stay solid**, framing the living pane. The Brain is a fixed,
+viewport-anchored backdrop; cards scroll over a stationary Brain.
+
+**Every chat component floats as a frosted-glass card, and legibility is guaranteed by
+construction — not by eye:**
+
+- **Blur** — a *high, constant* gaussian blur on every card. It provides perceptual calm and
+ kills the high-frequency motion detail that hurts reading. It is cheap (no per-frame logic)
+ and it is *not* what makes text legible.
+- **Tint** — contrast comes entirely from a tint whose **opacity is pinned to clear the WCAG-AA
+ floor (with a small safety margin, not exactly at the floor) over a fixed reference frame** —
+ a synthetic Brain at peak bloom in which the sampled backdrop is the brightest palette value
+ **the Brain actually paints in that theme** (peak `#fff676` on dark; peak `#8f8000` on light —
+ the light-mode Brain never paints `#fff676`, so validating against it there would derive a
+ near-transparent tint that silently fails over the real light frame). Blur smooths the backdrop but does not darken it; a card over a bright-yellow
+ cluster is still on bright yellow until the tint pulls the composite toward the ink. The tint
+ set is **keyed to the chat's own theme** — the chat runs a theme independent of the app shell
+ (default oc-2) — so a chat-theme change re-keys the Glass and re-validates contrast. **Two
+ tints:** *standard* (prose, bubbles, composer) holds body text ≥4.5:1; *dense* (code, diffs,
+ run-plots) is more opaque and holds code/diff text ≥4.5:1 and graphical marks — syntax colors,
+ diff fills, plot lines — ≥3:1 (WCAG 1.4.11), so colored content stays crisp.
+
+**The Brain is data-true with an adaptive heartbeat.** Nodes light up as amico reads files,
+runs tools, and touches the vault; the graph grows with the conversation, **densifying in place
+within the fixed, viewport-anchored frame** — bounded by a node cap / recency window so a long
+session neither scrolls the Brain nor grows its per-frame cost without limit. Motion follows
+activity: a lively tempo when amico is working, ~8fps breathing at rest, and a **hard pause**
+when the window is hidden or `prefers-reduced-motion` is set. At rest it shows a **sparse
+seed** — a near-empty Landing whose graph grows *only* from the current session's real
+thought, with **no cross-session persistence** (a fresh session opens nearly empty). One Brain
+per window shows the **active** session; switching the titlebar session tab swaps the Brain to
+that session's graph.
+
+**Pre-agreed perf fallback (non-negotiable mid-build):** the frosted look is protected. At
+rest, full blur and full motion always apply. A frame-time monitor trips the ease when p95
+frame time stays over budget for a sustained window (with hysteresis before restoring, to
+avoid oscillation): the Brain's *motion* eases (calm/slow) and returns to full tempo at rest.
+If easing motion to a full stop still misses budget, the Brain **hard-pauses to a static
+blurred field** — the terminal valve — but **blur fidelity never degrades.**
+
+**Build by reuse.** A new, well-named branch off `local/amicode` reuses the de-risked native
+plumbing — the native Brain engine, the glass-token generator, and the mount wiring from the
+earlier branch (design-neutral, already hardened by an adversarial port-fidelity review) — and
+builds the *design* layer fresh. No iframe. Both light and dark themes are first-class from day
+one.
+
+See `packages/app/CONTEXT.md` for the new vocabulary (**Brain**, **Glass**).
+
+## Consequences
+
+- The **inline brain strip** is retired as a timeline row; the timeline keeps only its text
+ shimmer (the existing working indicator — the cycling gerund + live elapsed shown while a turn
+ works). One Brain, one render loop.
+- **No node-level interactivity is lost.** The strip's only affordance was a click-to-expand/
+ collapse toggle (it was a collapsible row); the Brain engine has no node hit-testing. As an
+ always-present background the expand toggle is obviated, not removed from a capable surface.
+- When a solid **Panel** covers the Chat, the Brain need not animate behind the occluded area —
+ a cheap perf saving, since Panels are opaque.
+- The *standard* tint's opacity is **bounded below** by the contrast floor — it cannot be made
+ ultra-transparent, or muted grey text fails AA over the reference frame (the exact
+ failure the earlier build hit). Secondary/muted text therefore rides the *dense* tint or a
+ locally-dimmed zone.
+- The Chat's cost profile changes: high-radius backdrop-blur recomputed per frame over an
+ animating canvas, across many cards, while scrolling, is the expensive combination. It must
+ pass a **laptop-webview 60fps gate** (the real target — headless Chromium is only a proxy);
+ the motion-ease fallback is the release valve.
+- The **Rail, titlebar, and Panels are deliberately excluded.** Nav chrome stays rock-solid,
+ and dense Panel lists (Chats history, Run gallery) don't fight a moving background.
+- The Landing (a fresh, empty Chat per ADR 0001) now opens on the sparse-seed Brain rather
+ than a blank surface — the window's first impression is the living background, calm and
+ nearly empty, not a widget wall.
+
+## Alternatives considered
+
+- **Keep the inline brain strip** (a timeline row). Rejected: it is one element among many;
+ the redesign's intent is the Brain as the room, not a row.
+- **One calm reading lane / unified glass sheet** (Brain only in the margins). Rejected in
+ favor of per-component floating cards — motion lives in the gaps between cards, which is
+ where "movement comes through" *without* the cards themselves having to be see-through to it.
+- **Dynamic contrast glass** (each card samples its backdrop per frame and adapts its tint).
+ Rejected: per-frame readback is the biggest perf risk and AA is hard to prove against every
+ frame; the static AA-floor tint is deterministic and cheap.
+- **Quiet zone behind cards** (dim/slow the Brain under each card's footprint via per-card
+ masking). Rejected: real engine complexity for translucency the gaps already provide.
+- **Breathing skeleton atlas at rest** (the full latent graph, dim, breathing). Rejected in
+ favor of the sparse seed — calmer, more literal, and it lets the floating cards dominate.
+- **Protect motion and ease blur under load**, or **degrade only while scrolling**. Rejected:
+ the frosted look is the aesthetic being protected; motion is the acceptable give under load.
+- **Rebuild the engine clean**, or **evolve the old `kate/chat-brain-atmosphere` branch**.
+ Rejected: reuse the de-risked plumbing (including its already-fixed port defects), but build
+ the design fresh on a new branch rather than inherit the old design structure or re-pay for
+ the engine port.