LOBBY · ROOM VETUS
Phase I · inventory only · not the keynote

Workshop · 8 boards

What WeftOS already scores

Room Vetus sat with the dice already on the table. We classified rhymes. We did not invent Cardano. We did not write the combining contract.

This workshop is the inventory. Combining happens next door in Coniunctio.

75
harnessFit
100
compileConfidence
65
taskCoverage
90
toolSafety
53
memoryUsefulness
.024
estCostPerRunUsd

Live MetaHarness cast, this session. Five named faces + a price tag (dollars, not a sixth die). No n (trial count), no interval, no favorable set. Prior throw 2026-07-31: 75 / 65 / 51 on the overlapping three.

Try. Read the red memory face. See. 53 with no error bar. Why. A readiness number without a circuit is one cast.

Who sat

Four voices, one rule

  • governance-counsel — vectors, gates, the drifted briefing.
  • ecc-analyst — graph-health lean (same tilt every time), and the four-way coherence homonym.
  • defi-networker — bond, slash, economic ruin without odds.
  • scoring architect — every numeric surface, classed not merged.

Phase I rule

Nova does not retrofit WeftOS. Vetus does not invent Cardano. They meet only in Coniunctio.

Honesty: none of these files cite the book. Influence is structural. Class = rhyme / cousin / false friend.

These four seats exist so Coniunctio cannot claim “nobody looked at the walls.”

Inventory · click a row

Thirty-one surfaces, three classes

Seed V1–V18 kept. V19–V31 added after reading the files. Expand any row.

Classification is not a merge. A rhyme is still missing a circuit.

Try. Open V1, V6, V17. See. Rhyme vs false friend vs marked deck. Why. This table is what Coniunctio is allowed to land on — nothing else.

#SurfaceClassSymbol

Homonym

Circuit vs circuit-breaker

WeftOS already spent the word circuit on stop-loss. Cardano’s circuitus is sample space. Keep both words. Never collapse them.

In-tree · breaker

  • BudgetUsage.circuit_open — spend / tokens / iterations (WEFT-322)
  • circuit_breaker_no_op_limit — abort after N no-ops (default 3)
  • TerminationReason::CircuitBreaker
  • routing.md Level-2 “CircuitBreaker” — stub, not shipped

In-book · circuitus

  • Enumerated outcomes of a score
  • Favorable / total
  • A named sample space, or an honest “incomplete”
  • Not a cap. Not a trip. Not a rename of ruin-stop

Keep both words. A perfect stop-loss can still sit on a crooked table.

stake is fine in both rooms. fairness is not already equality-of-conditions.

Three already-honest scores

Casts you can recount — still no circuit

1. MetaHarness 5-die

75 / 100 / 65 / 90 / 53

Faces named. Favorable set not. memoryUsefulness 53 has no interval. Receipts are the nearest n — optional, off the weft path.

2. EffectVector L2

√(r²+f²+p²+n²+s²) ≤ √5 ≈ 2.236

ADR-034 bar 0.7. agent_spawn / binding-thread tests use 0.8. Unknown tool = zero vector. args ignored. Unweighted.

3. FitnessScorer

0.4 / 0.2 / 0.2 / 0.2 − 0.2 refusal

Ticket WEFT-54: not a safety control. English substrings. Two no-op scorers (always-constant dummies): pipeline always 1.0, RVF layer always 0.5.

These three are the closest WeftOS has to a recountable cast. None of them name a sample space. That is the gap Coniunctio may land on.

Marked deck

Governance-counsel EffectVector drift

Unequal conditions inside our own briefing. Coniunctio may use this; Vetus does not “fix” it here.

Fix the briefing (one five-die) before we lecture anyone else about aequitas.

Agent file · old 5D

agents/weftos/governance-counsel.md

cpu · memory · network · storage · trust_delta

CLI example still sends {"cpu":0.8,"memory":0.6,"network":0.3}

Live structs · ADR-034

effects.rs · governance.rs

risk · fairness · privacy · novelty · security

Same type name. Different faces. A marked deck we already host.

Sibling inequality: magnitude bar 0.7 (ADR / engine tests) vs 0.8 (spawn D6, binding-thread).

Do not collapse

Words already spent twice

coherence — four formulas

  • Fitness readability heuristic
  • ECC graph health (λ₂ / Fiedler vector — “how one-piece is the graph?”)
  • assess CLI: doc/code × 100
  • forensic: density × avg confidence

twins on the same table

  • GateDecision ×2 (core token String / kernel bytes) plus 4-way GovernanceDecision
  • Trajectory (pipeline) vs TrajectoryRecord (governance FIFO)
  • Two breakers named circuit, plus a stub
  • A learned “uncertainty” head does not close the K2 symposium gap (no intervals on EffectVector)

Same English word, four jobs. Always say which formula. Do not inventory a stub as live.

The planned 7-factor router scorer is a stub (routing.md §10). Inventorying it as shipped would be a marked card.

Handoff

What Vetus hands Coniunctio

Coniunctio may land only on these named gaps. Vetus wrote no contract.

Explicit non-delivery: no combining contract, no local-ADR text, no Cardano arithmetic. Nova does not edit the inventory.