GUI IA Reorg Blueprint (RATIFIED)

GUI IA Reorg Blueprint — RATIFIED

This is the synthesizer output of Plan 2 (Phase I). No GUI code is changed. Every row carries a RECOMMENDED default decision so the human ratifier edits exceptions only. Plan 3 (the actual reorg) is authored after the Phase J gate, scoped to surviving surfaces.

---

0. RATIFICATION (Gate J) — DECIDED 2026-06-26

The human ratifier approved the recommended set with three amendments. Status: RATIFIED.

Approved as recommended (no change):

• Bundle 1 (Latin renames — except mens/populi, see Amendment A), Bundle 3 (cluster merges: claims+knowledge→scientia; 4 activity clones→one Discovery surface), Bundle 4 (search→memory), Bundle 6 (phantom review + 5 parent-shell deletes) — ALL ACCEPTED.
• Bundle 2: needs-you + sub-agents conditional wire-or-cut APPROVED — Plan 3 wires tool:vox_resolve_approval (needs-you attention inbox) and cmd:list_subagent_tree (sub-agents); if wiring fails verification, they CUT. activity/runs moves accepted.

Amendment A — mens/populi are NOT cut; they are the model train/run surfaces. Their purpose is to train/run the custom model(s) (VoxMens). They need a GUI frontend derived from the CLI to establish GUI/CLI parity. Decision: RENAME (de-Latinize to clear labels) + WIRE to the model train/run CLI commands — a new ADD workstream in Plan 3 that mirrors the CLI surface-for-surface. Converts two empty shells into real, honest surfaces. dissent:structural:CUT-candidate → RESOLVED to WIRE. (Plan 3 must find the CLI commands for VoxMens train/run and build the GUI to parity.)

Amendment B — Settings consolidation + Policies-unification (elevates Bundles 5 & 7 to a first-class workstream). gamify config → Settings (approved). Beyond that: ALL settings for ALL subsections/pages must live in one well-organized Settings surface, and Plan 3 should evaluate unifying Settings + Policies into one place — only if it stays visually and conceptually clear what each is for (Settings = user config; Policies = enforced rules). gamify’s visual elements/concepts may be applied app-wide (not confined to a single surface). Bundle 7 (settings CONDENSE) is therefore no longer optional — it becomes a Settings information-architecture pass: gather scattered per-surface settings, organize them, and resolve the Settings↔Policies relationship.

Amendment C — matrix: left at recommended default (fold the single routing-nudge into the chat execution rail); Plan 3 may instead rename→“Routing” and keep it a surface if folding proves awkward. (Not explicitly ratified; recommended default stands.)

→ Plan 3 (execution) is AUTHORIZED, scoped to: the ratified moves/merges/renames/cuts below, PLUS two amendment workstreams — (1) VoxMens model train/run GUI built from the CLI for parity; (2) a Settings consolidation + Settings/Policies-unification IA pass. The caveat completions (vox-gui Rust compile-verify path, Playwright proof, 109 visual/DS-token/a11y findings) fold in, scoped to surviving surfaces.

---

1. Summary header

Counts per verb (recommended defaults, one decision per unit):

Verb	Count	Units
CUT	6	review (surface), agents/commands/compute/workspace/knowledge (registry parent-shell duplicates, ×5)
MERGE	6	claims→scientia · search→memory · activity-cluster (archive-panel,discovery-inbox,discovery-review,activity)→one Discovery surface (×4 absorbed, counted as 1 MERGE op) · matrix→chat rail
MOVE	5	memory (search→knowledge) · activity (orphan→knowledge/Discovery) · runs (parent-shell→named child) · gamify (agents→settings) · console-with-agent adjacency (flag)
RENAME	5	scientia→Findings · oratio→Voice · mens→(plain) · populi→(plain) · runs label “Runs & Approvals”→“Runs”
CONDENSE	2	knowledge group · settings (relocate journey-tail config)
EXPAND	1	needs-you (attention inbox)
ADD-to-nav	2	needs-you, sub-agents (re-add hollow shells WITH a wired command, else CUT)
KEEP	~22	all remaining surfaces + balanced groups

6 destructive (CUT) decisions, shown first. 1 of them (review) is a surface deletion; 5 are surface-registry parent-shell duplicates that mirror a child’s component and are never an independent destination. 0 judgment-only CUTs — every CUT rests on a structural fact (0 cmds + orphan-nav / parent-shell duplication).

---

2. Conflict resolution (applied mechanically)

Rules, applied per unit across the 7 dimensions:

1. Evidence-type precedence: structural > audit > judgment. A structural fact can veto a judgment; a judgment can never veto a structural fact.
2. Tie-break: prefer the least-destructive verb — KEEP > RENAME/MOVE > CONDENSE > MERGE > CUT.
3. Destructive threshold: a CUT or MERGE requires ≥2 dimensions agreeing OR one standalone structural fact. Otherwise it is downgraded to “flag, default KEEP”.
4. Dissent: when a dimension’s rec is overruled and unresolved, it is recorded inline as dissent: <dim>:<rec>.

Resolutions that moved off a raw dimension rec:

• needs-you — utility says CUT (structural: 0 cmds + orphan-nav, high). BUT a real component_dir=NeedsYou exists, it is already wired in surfaceComponents.tsx (case 'needs-you') and registered tier: 'live_backend'. The structural “0 cmds” is contradicted by an audit/structural fact (live component + wired case). Threshold met for destructive only on the orphan-nav half, not the dead half → downgrade CUT → EXPAND + ADD-to-nav. dissent: utility:CUT.
• sub-agents — utility says CUT (high); structural-coverage corrected to KEEP (the prior MERGE rested on a falsely-attributed list_subagent_tree); new-nav says MOVE-under-agents. Real component_dir=SubAgents shell. Least-destructive structurally-supported verb = ADD-to-nav (wire cmd:list_subagent_tree) OR CUT if it stays hollow. Default ADD-to-nav-conditional. dissent: utility:CUT.
• matrix — semantic-clarity RENAME→Routing (low conf) vs journey-coherence MERGE-into-chat-rail (med). Both are non-destructive-ish; journey has the stronger structural basis (only entry path is chat→matrix, richness=1 cmd). MERGE preferred but flagged Group B (reversible). dissent: semantic-clarity:RENAME(Routing).
• oratio — utility MERGE (redundant+thin, med) vs redundancy/cohesion KEEP (the only shared command is the read-only list_model_cards; start_mic_capture is unique → false redundancy peer, structural). Structural KEEP vetoes the utility MERGE (which was computed from a god-command peer edge). → RENAME→Voice, KEEP as surface. dissent: utility:MERGE.
• research / repository — utility MERGE (redundant+thin) vs redundancy/cohesion KEEP (peer edges are pure cmd:execute_command artifacts; distinct components + distinct real commands). Structural KEEP vetoes. → KEEP. dissent: utility:MERGE (×2).
• claims / scientia / knowledge — utility KEEP (“rich, MERGE only if dim-2 confirms”) vs redundancy/cohesion/journey/semantic MERGE (identical 12-cmd set + identical component_dir=Scientia, structural, ≥3 dims). The dim-2 confirmation utility asked for exists → MERGE clears the threshold. → MERGE (claims+knowledge-surface → scientia).
• compute/models, workspace/console, agents-shell — parent↔default-child identity artifacts. Redundancy/cohesion KEEP the live child, CUT only the registry parent-shell (structural). Surface KEEP; registry-shell CUT.

---

3. Decision table (grouped by ratification tier)

Coverage shorthand: surfaced = reachable+wired, orphan = orphan-nav, deadend = dead-end tool/cmd edges; cN = command_count.

Group C — CUT + any judgment-only destructive (explicit per-item confirm) — shown FIRST

Unit	Kind	Current location	Coverage	Evidence basis	DECISION	Rationale
review	surface	registry only (navGroup knowledge, parent null)	orphan · c0 · deadend0	structural (3 dims: struct-cov, utility, new-nav)	CUT	Dead (0 cmds, component_dir=null — no component) AND orphan-nav. Phantom registry entry; nothing to merge. Highest-confidence cut.
agents (surface-shell)	surface	registry shell, mirrors dashboard (component_dir=Dashboard, c0)	reachable-as-parent · c0	structural (new-nav)	CUT (registry shell)	Duplicates the agents nav parent + live dashboard child. Parent key lives in navigation.ts only; the registry shell is redundant.
commands (surface-shell)	surface	registry shell	reachable-as-parent · c0	structural (new-nav)	CUT (registry shell)	Parent-shell duplicate; catalog is the live child.
compute (surface-shell)	surface	registry shell, mirrors models (component_dir=Models, c4)	reachable-as-parent · c4(inherited)	structural (redundancy, new-nav)	CUT (registry shell)	Parent↔default-child identity artifact; models is the real surface.
workspace (surface-shell)	surface	registry shell, mirrors console (component_dir=Console, c8)	reachable-as-parent · c8(inherited)	structural (redundancy, cohesion)	CUT (registry shell)	Parent↔default-child identity artifact; console is the real surface.
knowledge (surface-shell)	surface	registry shell, mirrors scientia (component_dir=Scientia, c12)	reachable-as-parent · c12(inherited)	structural (redundancy, cohesion)	CUT (registry shell) → folds into the scientia MERGE	Parent-shell duplicate of the Scientia 12-cmd surface; absorbed by the claims→scientia MERGE below.

needs-you and sub-agents were downgraded out of Group C (utility recommended CUT) to ADD-to-nav-conditional (Group B) per §2. Their dissent: utility:CUT is shown there.

Group B — MOVE / MERGE / CONDENSE / EXPAND / ADD (opt-in review)

Unit	Kind	Current location	Coverage	Evidence basis	DECISION	Rationale
claims	surface	knowledge (parent knowledge)	surfaced · c12 · redundant	structural (redundancy+cohesion+journey+semantic, 4 dims)	MERGE → scientia	Identical 12-cmd set + identical component_dir=Scientia as scientia/knowledge. Fold to one surface; claims becomes a filter/tab.
search	surface	top-level parent (1 child: memory)	surfaced · c2 · redundant	structural (redundancy+cohesion+journey, 3 dims)	MERGE → memory	Degenerate 1-child group; shares vox_search_query+open_locator with its only child. Retire the search parent.
activity-cluster: archive-panel, discovery-inbox, discovery-review, activity	surface ×4	knowledge (×3) + orphan (activity)	activity=orphan; others surfaced · c1 each · all redundant	structural (redundancy+cohesion+journey+struct-cov, 4 dims)	MERGE → one “Discovery” surface (filter tabs: Inbox/Review/Archive)	One component_dir=Activity + one cmd:activity_query, surfaced under 4 names distinguished only by filter. Preserve inbox/review intents as named presets.
memory	surface	search (parent search)	surfaced · c5	structural (new-nav, high)	MOVE: search→knowledge	Reparent the live recall surface into Knowledge; kills the Search↔Memory collision.
activity	surface	orphan (no nav)	orphan · c1	structural (struct-cov+cohesion+new-nav)	MOVE: orphan→knowledge (becomes canonical Discovery surface)	The only activity_query peer left out of nav; make it the reachable Discovery home.
runs	surface	live but only reachable as parent fallthrough	orphan-as-tab · c1 (get_gui_run)	structural+judgment (new-nav, semantic)	MOVE: add named runs child under Runs parent	Most-wired surface in the group is the least reachable; add PARENT_CHILD_MAP.runs:{parent:'runs',child:'runs'}.
gamify	surface	agents	surfaced · c6	judgment (new-nav+cohesion+journey)	MOVE: agents→settings	Opt-in immersive mode (default off) is config-adjacent; tightens Agents to “watch/steer the swarm”.
matrix	surface	agents	surfaced · c1 (nudge_routing_intention)	structural+judgment (journey, med)	MERGE → chat execution rail	One command, only entry path is chat→matrix. Surface the nudge inline in chat. dissent: semantic-clarity:RENAME(Routing).
needs-you	surface	orphan (wired component, registry live_backend)	orphan · c0 · component=NeedsYou	structural+audit (journey EXPAND, new-nav MOVE)	EXPAND + ADD-to-nav (attention inbox under Runs); absorb approvals inline-resolve + wire dead doubt/overrule	Real wired component; the unifying home for approve/doubt/overrule. dissent: utility:CUT (overruled by live-component fact).
sub-agents	surface	orphan (component SubAgents, c0)	orphan · c0 · component=SubAgents	structural+judgment (struct-cov KEEP, new-nav MOVE)	ADD-to-nav under Agents — conditional: wire cmd:list_subagent_tree else CUT	Real shell, dead today. dissent: utility:CUT.
console (agent-scoped)	surface	workspace	surfaced · c8	structural+judgment (journey, low)	MOVE adjacency (flag only)	dashboard→console agent drill-down crosses agents→workspace. Optional agent-scoped console panel. Low confidence; default KEEP placement.
knowledge group	nav-group	—	—	structural+judgment (redundancy CONDENSE, semantic CONDENSE)	CONDENSE	After scientia+Discovery merges & memory absorb: {Memory, Findings(scientia), Research, Discovery, Publications} ≈ 5.
settings	surface	top-level	surfaced · c17	judgment (journey, low)	CONDENSE (flag)	Relocate journey-tail config (routing/selection→models; gamify settings→gamify; trusted-nodes→mesh). Low confidence; opt-in.

Group A — RENAME / KEEP (opt-out default-accept)

Unit	Kind	Current location	Coverage	Evidence basis	DECISION	Rationale
scientia	surface	knowledge (default child)	surfaced · c12	judgment (semantic+new-nav, high)	RENAME label → “Findings”/“Library” (key unchanged)	“Knowledge→Scientia” is Latin-for-the-same-word. Label-only diff.
oratio	surface	compute	surfaced · c2	structural+judgment (semantic high)	RENAME → “Voice”/“Speech” (key unchanged); also align component_dir Loquela	Double Latin naming (nav Oratio vs code Loquela). dissent: utility:MERGE (vetoed — false peer).
mens	surface	compute	surfaced-but-empty · c0	structural+judgment (semantic high, new-nav low)	RENAME (de-Latinize) — conditional KEEP; CUT if stays hollow	Latin + 0 cmds + null component. dissent: structural:CUT-candidate (deferred to ratifier).
populi	surface	compute	surfaced-but-empty · c0	structural+judgment (semantic high, new-nav low)	RENAME (de-Latinize) — conditional KEEP; CUT if stays hollow	Same as mens. dissent: structural:CUT-candidate.
runs (label)	nav-group	”Runs & Approvals”	—	structural+judgment (semantic high)	RENAME label → “Runs”	Group contains Runs(new child)+Approvals+Policies; drop the awkward conjunction.
harness	surface	workspace	surfaced · c0 (redirect stub)	structural+judgment (semantic med)	KEEP (flag: label over-promises)	Redirect-only; rename deferred, no named disqualifier.
gamify (label)	surface	—	—	judgment (semantic low)	KEEP (flag: verb label)	“Gamify” reads as a verb; optional rename to “Quests/Arena”. Low priority.
flow,console,catalog,repository,browser,tasks,dashboard,chat,models,mesh,policies,publications,coverage,skills,approvals,research	surface ×16	various	mixed	structural+judgment	KEEP	No named disqualifier; structurally distinct or structurally-vetoed against destructive recs.
Agents,Workspace,Commands,Compute,Runs (groups)	nav-group ×5	—	—	structural (cohesion KEEP)	KEEP	Coherent containers; member-level edits handled above.

---

4. Per-verb executable fields (for Plan 3)

CUT

view_key removed	nav entries to delete	#view= redirect target	tests to update	grep handle
review	none (not in PARENT_CHILD_MAP) — delete registry row	#view=review → #view=scientia (nearest knowledge surface)	surfaceRegistry.generated.ts regen; surfaceComponents.tsx has no case 'review' (already absent)	grep -rn "'review'|\"review\"|viewKey: 'review'" crates/vox-gui/ui/src
agents (registry shell)	none in nav (key stays as TOP_LEVEL_VIEW) — delete registry surface row only	n/a (parent key survives in navigation.ts)	surfaceRegistry.generated.ts regen	grep -n "viewKey: 'agents'" crates/vox-gui/ui/src/generated/surfaceRegistry.generated.ts
commands (shell)	none — delete registry row only	n/a	regen	grep -n "viewKey: 'commands'" .../surfaceRegistry.generated.ts
compute (shell)	none — delete registry row only	n/a	regen	grep -n "viewKey: 'compute'" .../surfaceRegistry.generated.ts
workspace (shell)	none — delete registry row only	n/a	regen	grep -n "viewKey: 'workspace'" .../surfaceRegistry.generated.ts
knowledge (shell)	none — delete registry row only (folds into scientia MERGE)	#view=knowledge → #view=scientia	regen; navigation.test.ts	grep -n "viewKey: 'knowledge'" .../surfaceRegistry.generated.ts

MERGE

absorber	absorbed	features/cmds that move vs drop	absorbed view_key → redirect	nav keys retired	component disposition
scientia (→ relabel Findings)	claims, knowledge(surface)	all 12 cmds shared (identical) → none move/drop; claims becomes a claim-review filter/tab	claims→scientia; knowledge→scientia	PARENT_CHILD_MAP.claims removed; knowledge stays a group key only	keep Scientia component; delete Claims/knowledge-shell rows
memory	search	vox_search_query,open_locator already in memory → none drop; search’s extra vox_search_query is dup	search→memory	search removed from TOP_LEVEL_VIEWS; DEFAULT_CHILD_BY_PARENT.search removed	keep Memory component; delete Search registry row
activity (Discovery)	archive-panel,discovery-inbox,discovery-review	all share cmd:activity_query → none drop; inbox/review/archive become named filter presets	archive-panel/discovery-inbox/discovery-review→activity	their 3 PARENT_CHILD_MAP knowledge children retired	keep one Activity component; the 3 absorbed used the same component already (no orphan components)
chat execution rail	matrix	nudge_routing_intention → moves inline to ChatExecutionRail intent button	matrix→chat	PARENT_CHILD_MAP.matrix removed	Matrix component retired or embedded as a chat-rail control

MOVE

from→to parent	PARENT_CHILD_MAP edit	DEFAULT_CHILD_BY_PARENT impact	breadcrumb effect
memory: search→knowledge	memory:{parent:'knowledge',child:'memory'}	none (memory was never a default child of knowledge); remove DEFAULT_CHILD_BY_PARENT.search	”Search › Memory” → “Knowledge › Memory”
activity: orphan→knowledge	ADD activity:{parent:'knowledge',child:'activity'}	none	new “Knowledge › Discovery”
runs: parent-shell→named child	ADD runs:{parent:'runs',child:'runs'}; change DEFAULT_CHILD_BY_PARENT.runs from 'approvals' to 'runs'	runs parent now lands on the live scoreboard, not approvals	”Runs & Approvals › Approvals” → “Runs › Runs”
gamify: agents→settings	gamify:{parent:'settings',child:'gamify'}	none (settings default stays settings)	“Agents › Gamify” → “Settings › Gamify”
console adjacency (flag)	no map edit; optional agent-scoped panel	none	unchanged

RENAME

old key (keep)	old→new label	NAV_LABELS diff	key changes?	docs refs
scientia	”Scientia” → “Findings”	NAV_LABELS.scientia: 'Findings'	no (label-only, no redirect)	grep -rn "Scientia" crates/vox-gui/ui/src docs/
oratio	”Oratio” → “Voice”	NAV_LABELS.oratio: 'Voice'	no label-only; optional component_dir rename Loquela→Voice is a code refactor (own commit)	grep -rn "Oratio|Loquela" crates/vox-gui/ui/src
mens	”Mens” → plain (e.g. “Reasoning”)	NAV_LABELS.mens: '…'	no	grep -rn "'mens'|\"mens\"|Mens" crates/vox-gui/ui/src
populi	”Populi” → plain (e.g. “Crowd”)	NAV_LABELS.populi: '…'	no	grep -rn "'populi'|Populi" crates/vox-gui/ui/src
runs (group)	“Runs & Approvals” → “Runs”	NAV_LABELS.runs: 'Runs'	no	grep -n "Runs & Approvals" crates/vox-gui/ui/src/lib/navigation.ts

ADD

proposed view_key	parent	real cmd:/tool: node it surfaces (REQUIRED)	nearest component to reuse
runs (as named child)	runs	cmd:get_gui_run (+ list_gui_runs,get_model_scoreboard,get_routing_summary_live from behavioral findings)	RunsView (already in surfaceComponents.tsx case 'runs')
needs-you (re-add to nav)	runs	MUST WIRE — candidate tool:vox_resolve_approval (the live approve/reject resolve from Dashboard AgentRow) routed through NeedsYou	NeedsYouSurface (already wired case 'needs-you')
sub-agents (conditional)	agents	MUST WIRE cmd:list_subagent_tree (status=surfaced but currently no edge to this surface) — else CUT	SubAgents (component exists, component_dir=SubAgents)
activity (Discovery)	knowledge	cmd:activity_query	ActivitySurface (already wired case 'activity')

ADD honesty gate: needs-you and sub-agents may only be ADDed if Plan 3 wires a cited command; if the wiring fails verification they revert to CUT (their utility-dimension default).

---

5. Migration ledger (consolidated — #view= deep-links MUST NOT silently break)

parseViewFromLocation currently returns the raw key from #view= / ?view=; any removed key must be remapped to its nearest surviving parent so old deep-links/bookmarks resolve.

old view_key	new key | null	parseViewFromLocation fallback	deprecation policy	asserting spec/test (grep handle)
search	memory	#view=search → resolve to memory	silent alias one release, then hard-remove	navigation.test.ts:19 asserts ?view=memory; surfaceRegistry.generated.ts:37
claims	scientia	→ scientia	silent alias one release	surfaceRegistry.generated.ts:19; PARENT_CHILD_MAP.claims in navigation.ts:21
knowledge (surface)	scientia (group key survives)	parent key resolves via DEFAULT_CHILD_BY_PARENT.knowledge → set to scientia (or memory)	group key kept; surface row hard-removed	navigation.ts:18,43; surfaceRegistry
archive-panel	activity	→ activity (preset=archive)	silent alias one release	PARENT_CHILD_MAP.'archive-panel' navigation.ts:31; surfaceRegistry
discovery-inbox	activity	→ activity (preset=inbox)	silent alias one release	navigation.ts:30; surfaceRegistry
discovery-review	activity	→ activity (preset=review)	silent alias one release	navigation.ts:20; surfaceRegistry.generated.ts:26
matrix	chat	→ chat (rail intent control)	silent alias one release	PARENT_CHILD_MAP.matrix navigation.ts:7; surfaceComponents.tsx:111
review	scientia	→ scientia	hard-remove (phantom, no component)	surfaceRegistry.generated.ts:35
agents/commands/compute/workspace (registry shells)	unchanged group keys	parent keys keep resolving via DEFAULT_CHILD_BY_PARENT	registry rows hard-removed; keys survive	navigation.ts:35-45 TOP_LEVEL_VIEWS/DEFAULT_CHILD_BY_PARENT
gamify	gamify (reparented)	key unchanged; PARENT_CHILD_MAP.gamify.parent agents→settings	no alias needed (key stable)	navigation.ts:29; surfaceComponents.tsx:136
oratio/scientia/mens/populi/runs-label	keys unchanged	n/a (label-only)	n/a	NAV_LABELS navigation.ts:62-99; surfaceRegistry navLabel fields

Key-affecting decisions (need redirect + call-site grep): search, claims, knowledge(surface), archive-panel, discovery-inbox, discovery-review, matrix, review. Label-only (no redirect): scientia, oratio, mens, populi, runs-label, gamify(reparent).

Three SSOT assertion sites Plan 3 must keep in sync for every key/label change:

1. crates/vox-gui/ui/src/lib/navigation.ts (PARENT_CHILD_MAP, TOP_LEVEL_VIEWS, DEFAULT_CHILD_BY_PARENT, NAV_LABELS)
2. crates/vox-gui/ui/src/generated/surfaceRegistry.generated.ts (regenerated — do not hand-edit; fix the generator)
3. crates/vox-gui/ui/src/components/layout/surfaceComponents.tsx (case '<viewKey>' dispatch) + navigation.test.ts

---

6. New-nav before/after tree

BEFORE (current navigation.ts)

Chat        → chat
Agents      → dashboard, flow, matrix, tasks, gamify
Runs        → approvals, policies            (label "Runs & Approvals"; runs surface ORPHAN)
Workspace   → repository, browser, harness, console
Commands    → catalog, skills
Search      → memory                          (1-child group; Search↔Memory collision)
Knowledge   → scientia, research, discovery-review, claims, publications,
              discovery-inbox, archive-panel  (7 children; 5 are clones in 2 clusters)
Compute     → models, mens, populi, oratio, mesh   (3 Latin labels; mens/populi empty)
Settings    → settings, coverage
ORPHANS (in registry, not in nav): activity, needs-you, review, sub-agents, runs

AFTER (proposed)

Chat        → chat                                              [+ inline routing-intent rail (matrix folded)]
Agents      → dashboard, flow, tasks, sub-agents?               rationale: "watch/steer the swarm"; drop gamify, fold matrix
Runs        → runs, approvals, policies, needs-you              rationale: "review outcomes + gate them"; runs+attention now first-class
Workspace   → console, repository, browser, harness            rationale: "act on the dev environment"; KEEP (only cohesive multi-member group)
Commands    → catalog, skills                                   rationale: "browse + run commands/skills"; KEEP
Knowledge   → memory, scientia(→Findings), research,            rationale: "find/recall/review what the system knows"; absorbs Search;
              discovery(4 clones→1), publications                          4 activity clones condensed to 1 Discovery
Compute     → models, mesh, oratio(→Voice), mens?, populi?      rationale: "run/route/inspect models + node mesh"; Latin renamed
Settings    → settings, coverage, gamify                        rationale: "configure the system"; gamify parked here (opt-in)

Per-group one-line rationale + agreement/migration_cost (every multi-member group crosses singleton Leiden communities — the GUI graph is fully fragmented, so migration_cost is structurally LOW for all; no proposed group splits a tight code community):

Group	rationale	agreement / migration_cost
Agents	watch/steer the running swarm; gamify out, matrix folded	crosses-community(low) — no shared code between members
Runs	review outcomes + gate them; runs & needs-you promoted	crosses-community(low)
Workspace	act on the dev environment	crosses-community(low) — only structurally cohesive group (distinct command sets per member)
Commands	browse + run commands/skills	crosses-community(low)
Knowledge	find/recall/review what the system knows; ~5 after condense	crosses-community(low) — absorbs search + condenses 4 activity clones
Compute	run/route/inspect models + mesh	crosses-community(low)
Settings	configure the system	crosses-community(low)

Smells resolved: orphan-nav (activity/needs-you/runs wired in; review/sub-agents resolved) · Search→Memory collision (search retired, memory under Knowledge) · Latin labels (scientia/oratio/mens/populi renamed) · lopsided Knowledge (7→~5) and degenerate 1-child Search (removed).

---

7. Themed bundles (accept/edit as a unit at the gate)

1. Retire Latin labels — 4 renames (Group A, label-only, no redirects). scientia→Findings, oratio→Voice, mens→(plain), populi→(plain). Lowest blast radius.

2. Fold orphan-nav surfaces — 3 moves + 1 expand (Group B). activity→Knowledge/Discovery, runs→named child under Runs, needs-you→EXPAND under Runs, sub-agents→conditional ADD under Agents.

3. Collapse exact duplicate clusters — 2 MERGE ops (Group B). claims+knowledge-surface → scientia (one Knowledge surface); 4 activity clones → 1 Discovery surface with Inbox/Review/Archive presets.

4. Kill the Search↔Memory collision — 1 merge + 1 move (Group B). Retire search top-level; memory reparented under Knowledge as the single recall surface.

5. Tighten Agents — 1 move + 1 merge (Group B). gamify→Settings; matrix→chat execution rail.

6. Delete registry phantoms/shells — 6 CUTs (Group C, structural). review (phantom) + 5 parent-shell duplicates (agents/commands/compute/workspace/knowledge surface rows). All structural; no judgment-only cuts.

7. Optional config relocation — CONDENSE settings (Group B, low confidence, flag). Move routing/selection→Models, gamify settings→Gamify, trusted-nodes→Mesh. Defer if contentious.

---

Unresolved dissents (carried to the gate)

• needs-you — dissent: utility:CUT (overruled: live wired component + registry live_backend veto the “dead” half; default EXPAND+ADD, reverts to CUT only if wiring fails).
• sub-agents — dissent: utility:CUT (default conditional-ADD; CUT if list_subagent_tree not wired).
• mens, populi — dissent: structural:CUT-candidate (0 cmds + null component; new-nav defers the cut, only de-Latinizes; ratifier decides RENAME-keep vs CUT).
• oratio — dissent: utility:MERGE (vetoed by structural: false redundancy peer via read-only list_model_cards; default RENAME+KEEP).
• research, repository — dissent: utility:MERGE (×2) (vetoed by structural: peer edges are cmd:execute_command god-command artifacts; default KEEP).
• matrix — dissent: semantic-clarity:RENAME(Routing) (journey MERGE-into-chat preferred; if the ratifier keeps it a surface, apply the Routing rename instead).
• console/settings MOVE/CONDENSE — low confidence, flagged opt-in; default KEEP placement.