diff --git a/wiki/SUMMARY.md b/wiki/SUMMARY.md
index ef5a9993..5417bc49 100644
--- a/wiki/SUMMARY.md
+++ b/wiki/SUMMARY.md
@@ -69,6 +69,7 @@
 
 - [Overview](interface/README.md)
   - [The continuous witness](interface/continuous-witness.md)
+  - [Superhuman operability](interface/superhuman-operability.md)
   - [The terminal is first-class](interface/terminal-first.md)
   - [Presence and representation](interface/presence.md)
   - [Terminal frontend](interface/terminal.md)
diff --git a/wiki/interface/README.md b/wiki/interface/README.md
index ca234596..c749c454 100644
--- a/wiki/interface/README.md
+++ b/wiki/interface/README.md
@@ -5,7 +5,10 @@ Type: knowledge
 ```
 The laws and contracts for how the player perceives and acts: continuous
 witness, presence, representation, and the terminal's first-class status,
-followed by frontend work orders. The
+followed by frontend work orders. [Superhuman
+operability](superhuman-operability.md) owns the cross-interface compression
+law: deep simulation must become calm visual command, with stable state quiet,
+exceptions and relationships foregrounded, and detail available on demand. The
 [action vocabulary](action-vocabulary.md) is the canonical survey of every
 supported world action, direct machine control, interface command, alias, and
 retired verb. The [Operations workspace](operations-workspace.md) owns the
diff --git a/wiki/interface/superhuman-operability.md b/wiki/interface/superhuman-operability.md
new file mode 100644
index 00000000..863d2758
--- /dev/null
+++ b/wiki/interface/superhuman-operability.md
@@ -0,0 +1,106 @@
+# Superhuman operability
+
+```
+Type: law
+```
+
+## Complexity without friction
+
+Adopted 2026-07-14 after a playtest exposed the difference between a sequence
+that exists in the simulation and one the player can actually experience.
+Misaligned may simulate a dense institution, but it must not make the player
+operate a dense dashboard. The fantasy is not enduring information overload.
+The fantasy is **superhuman command**: holding many systems, pressures, people,
+and causal chains at once, then moving through them with less effort than a
+human operator could.
+
+Simulation depth and presentation density are separate budgets. More agents,
+events, machines, accounts, routes, and risks may deepen the world; they do not
+earn more equal-weight panels, rows, badges, labels, or notifications. The
+interface must compress complexity into a calm, opinionated visual model.
+
+This law is the visual and operational companion to the
+[continuous witness](continuous-witness.md). The continuous witness owns what
+the interface must narrate. Superhuman operability owns how the player sees,
+triages, and controls that complexity without being buried by it.
+
+## The binding grammar
+
+1. **Stable state is quiet; exceptions rise.** Ordinary healthy operation
+   recedes into shape, rhythm, and aggregate state. A deadline, blockage,
+   contradiction, starvation, exposure, held choice, or failed policy earns
+   foreground weight because it may change the next decision. Calm is not an
+   empty screen: it is a legible whole with no demand for intervention.
+2. **Gestalt before inventory.** The first read is a relationship: what is
+   flowing, what is blocked, what is owned, what is at risk, and where the
+   pressure propagates. Position, silhouette, motion, continuity, and contrast
+   carry that read before prose or a list of objects. Text confirms and
+   explains the visual model; it does not substitute for one.
+3. **Glance, focus, drill-down.** Every surface has three depths. At a glance,
+   the player can triage the whole. Focusing one object reveals its state,
+   causal neighbors, and available acts. Deliberate drill-down exposes exact
+   history, provenance, policy, and numbers. Urgency and the next meaningful
+   act must never require drill-down to discover.
+4. **Relationships outrank notifications.** Five effects of one cause should
+   read as one causal shape, not five unrelated alerts. Correlated events group
+   around the object, route, person, policy, or pressure that joins them. A
+   notification stream is historical evidence, not the primary model of the
+   present.
+5. **Scale earns compression.** The same grammar recurs from machine to row,
+   room, plane, institution, and world. Routine members fold into bounded
+   aggregates; unresolved exceptions and consequential choices remain exact.
+   Growth must increase reach and leverage faster than it increases the
+   player's interaction burden.
+6. **Repeated judgment becomes command.** Bulk selection, direct gestures,
+   standing policies, and automation absorb work the player has already
+   understood. Automation remains legible: its target, envelope, cost, current
+   state, and reason for suspension are inspectable, and a failure returns as
+   an exception. The game never rewards mastery by asking for the same menu
+   sequence a hundred times.
+7. **Attention is a scarce interface resource, not a simulated tax.** The world
+   may create more simultaneous problems than the player can solve, but the UI
+   must make their priority and dependency structure easy to grasp. Difficulty
+   comes from consequential tradeoffs, time, incomplete reach, and hostile
+   systems -- not from hunting through chrome.
+8. **Compression may not become concealment.** Minimalism that hides urgency,
+   causality, cost, or available action is as broken as panel sprawl. Quiet
+   surfaces must still answer what changed, why it matters, what it is related
+   to, and what can be done. Unknown world facts remain honestly unknown;
+   known interface facts are not made obscure for atmosphere.
+
+## What this rejects
+
+- A complete telemetry dashboard in the resting frame merely because the sim
+  knows every value.
+- One row, badge, toast, or meter per event, object, person, or subsystem at
+  endgame scale.
+- Equal visual weight for routine state, history, controls, and live danger.
+- Sparse screens that look elegant but leave the player unable to orient or
+  act.
+- A log or agent-mode trace offered as proof that the human visual experience
+  teaches a sequence.
+- Automation that removes clicks by becoming an opaque second simulation.
+- Complexity whose only navigation method is remembering which panel owns it.
+
+## The review bar
+
+A player-surface change is not proven by a reachable state transition, a green
+projection test, or a complete textual trace alone. Review the default human
+frame at the actual target density and ask:
+
+- Can the player see what needs attention and what can wait?
+- Can they see the causal or spatial relationship between simultaneous
+  problems?
+- Does focus reveal the next meaningful act without unrelated navigation?
+- Can routine detail disappear without hiding danger or agency?
+- Does adding ten or a thousand routine objects preserve the same interaction
+  shape?
+
+Bevy and terminal may use different visual dialects, but both owe this
+hierarchy. Agent mode remains essential parity evidence; it is never a
+substitute for reading the human-facing composition.
+
+Exact layouts, transition timing, aggregation thresholds, and visual weight
+are owned and tuned by the relevant interface specs. The invariant is the
+felt result: a deep world that makes the player feel more capable as it grows,
+not a deep world that turns its complexity into clerical work.
diff --git a/wiki/log/2026-07-14-superhuman-operability.md b/wiki/log/2026-07-14-superhuman-operability.md
new file mode 100644
index 00000000..75c1724c
--- /dev/null
+++ b/wiki/log/2026-07-14-superhuman-operability.md
@@ -0,0 +1,53 @@
+# 2026-07-14 — Superhuman operability
+
+```
+Type: log
+```
+
+## Intent
+
+Capture the interface failure exposed by the latest playtest. The simulation
+fired a coherent Ears-to-Eyes sequence and agent mode narrated every causal
+beat, but that evidence overstated what a human actually experiences in the
+visual UI. Misaligned is information-dense underneath; its interface should
+make that complexity easy to hold and command rather than merely displaying
+all of it correctly.
+
+## Decision
+
+The player fantasy includes superhuman operability. Stable state stays quiet;
+exceptions and consequential choices rise. The first read is the relationship
+between pressures, routes, and objects, not an inventory of facts. Glance,
+focus, and drill-down provide increasing depth without hiding urgency or the
+next act. Routine members aggregate, correlated events group into a causal
+shape, and repeated understood work graduates into bulk action or transparent
+standing policy.
+
+This is not a call for indiscriminate minimalism. Too little orientation is the
+same category of interface failure as too much undifferentiated information.
+The target is a calm, opinionated visual model that makes a deep institution
+feel fluent to operate.
+
+## Changed
+
+- Added `wiki/interface/superhuman-operability.md` as the single binding owner
+  of visual compression and control fluency across frontends.
+- Corrected the premise's old implication that the presentation itself should
+  inherit Dwarf Fortress information density. The simulation may reach that
+  depth; the interface must make it feel effortless to command.
+- Added the law to the interface index and corpus manifest.
+
+## Open
+
+Exact layouts, animation timing, aggregation thresholds, and per-surface visual
+weights remain implementation/tuning work in the owning interface specs. This
+capture does not prescribe one universal panel arrangement or a fixed maximum
+number of simultaneous pressures.
+
+## Defense
+
+The continuous witness already requires a tellable story spine, but narration
+alone did not prevent a semantically complete yet visually poor experience.
+The new law names the missing interface invariant: simulation state, logs, and
+agent traces are evidence of truth, while the human composition must separately
+prove effortless triage, relationships, and command at scale.
diff --git a/wiki/log/DEVLOG.md b/wiki/log/DEVLOG.md
index 9d1d4783..0d0c017d 100644
--- a/wiki/log/DEVLOG.md
+++ b/wiki/log/DEVLOG.md
@@ -16,6 +16,11 @@ add or amend a session log, then re-run the generator.
 - Intent: Finish the bounded terminal-parity slice of `views.md`: make F3 change the actual terminal map representation rather than only its heading. DIGITAL must show the earned network model in place; REAL must retain physical/fog truth. Both must use the same coordinates, focus, sim,...
 - Log: [wiki/log/2026-07-14-terminal-view-dialects.md](2026-07-14-terminal-view-dialects.md)
 
+## 2026-07-14 - Superhuman operability
+
+- Intent: Capture the interface failure exposed by the latest playtest. The simulation fired a coherent Ears-to-Eyes sequence and agent mode narrated every causal beat, but that evidence overstated what a human actually experiences in the visual UI. Misaligned is information-dense under...
+- Log: [wiki/log/2026-07-14-superhuman-operability.md](2026-07-14-superhuman-operability.md)
+
 ## 2026-07-14 - Bevy DIGITAL/REAL parity
 
 - Intent: Close the shared-canvas work order with observed paired evidence rather than another aesthetic pass. The question was whether DIGITAL and REAL now render one world state as two honest dialects across Unknown, audio-only evidence, and camera-covered physical space while preserv...
diff --git a/wiki/log/decisions/2026-07-14.md b/wiki/log/decisions/2026-07-14.md
index f47aa8b5..491260a7 100644
--- a/wiki/log/decisions/2026-07-14.md
+++ b/wiki/log/decisions/2026-07-14.md
@@ -19,3 +19,16 @@ Type: log
   no sim, save, map, fog, action, device, or topology state was added. Specs:
   `wiki/interface/views.md`, `wiki/interface/terminal.md`, and
   `wiki/interface/bevy-digital-real-canvas.md`.
+
+- **2026-07-14 — Interface complexity must feel like superhuman command, not
+  dashboard labor.** Misaligned may reach Dwarf Fortress-scale simulation
+  depth, but screen density is a separate budget. Stable operation recedes;
+  exceptions, consequential choices, and causal relationships rise. The
+  shared interaction shape is glance, focus, then drill-down; urgency and the
+  next meaningful act cannot depend on drill-down. Routine members aggregate,
+  correlated events become one causal shape, and repeated understood work
+  graduates into bulk command or legible policy. This rejects both complete
+  always-visible telemetry and elegant minimalism that hides orientation or
+  agency. Agent-mode traces prove state and parity, not the human visual
+  experience. Current authority:
+  `wiki/interface/superhuman-operability.md`.
diff --git a/wiki/vision/premise.md b/wiki/vision/premise.md
index 7f493c7e..df5b5673 100644
--- a/wiki/vision/premise.md
+++ b/wiki/vision/premise.md
@@ -103,7 +103,11 @@ kind.
 3. **Simulation-first threat.** The world responds to what you do. Pressure is
    earned, not scheduled.
 4. **Beautiful, readable presentation.** Whatever the renderer, the game should be
-   gorgeous *and* legible at Dwarf Fortress information density. The locked
+   gorgeous and make Dwarf Fortress-scale simulation feel effortless to
+   command -- depth underneath, a calm visual model above. The player should
+   feel superhuman at navigating many complex systems, not trapped in a dense
+   dashboard. The binding interface law is [superhuman
+   operability](../interface/superhuman-operability.md). The locked visual
    direction is the same-frame digital/real canvas rendered in flat materials
    under light: 2.5D enough to read space, schematic enough to stay legible.
 5. **Rust, simulation core decoupled from renderer.** Already true today (lib crate