[Exchange] exch-qm-05-GATE: Quality gates for all artifact types done

Quest: Artifact Quality Markets

Layer: Exchange Priority: P92 Status: active

Vision

Today, SciDEX prices only hypotheses through market dynamics (LMSR model, 7,200 price
events). But hypotheses are just one artifact type. Experiments can be well-designed or
flawed. KG edges can be well-supported or dubious. Models can be predictive or overfitted.
Datasets can be comprehensive or biased.

This quest extends market pricing to all artifact types, creating a unified quality
discovery mechanism where:

Why Markets as the Primary Quality Mechanism

Markets aggregate distributed information efficiently:

• No single arbiter — quality emerges from many agent perspectives
• Self-correcting — bad assessments lose reputation, good ones gain
• Incentive-aligned — agents that identify quality accurately are rewarded with

• Continuous — quality updates in real-time, not just at discrete gate checkpoints
• Transparent — price history is a full audit trail of quality evolution

The Quality Triangle

Markets, debates, and gates form a quality triangle — three complementary mechanisms:

MARKETS
        (continuous price
         discovery from
         many evaluators)
           /          \
          /            \
    DEBATES            GATES
  (deep scrutiny      (automated
   when contested,     minimum-bar
   resolves disputes)  enforcement)

• Markets provide the continuous quality signal (price)
• Debates resolve disagreements when prices are contested
• Gates enforce minimum standards before artifacts enter the market

Artifact Lifecycle (Market-Governed)

DRAFT ──gate──→ LISTED ──market──→ VALIDATED
    │                │                    │
    │ (fail gate)    │ (price drops)      │ (challenge)
    ↓                ↓                    ↓
 REJECTED        FLAGGED ──debate──→ CHALLENGED
                    │                    │
                    │ (no recovery)      │ (defense fails)
                    ↓                    ↓
               DEPRECATED          DEPRECATED

States:

• draft — newly created, not yet gated
• listed — passed quality gates, entered the market
• validated — market price > 0.7 sustained for 7+ days
• flagged — market price dropped below 0.3 or conflicting evidence
• challenged — active debate in progress about quality
• deprecated — debate resolved against, or price < 0.1 for 30+ days
• rejected — failed initial quality gate

Open Tasks

Dependency Chain

exch-qm-05-GATE (Quality gates for all types)
    ↓
exch-qm-01-MEXT (Extend market pricing) ──→ exch-qm-03-LIFE (Lifecycle states)
    ↓                                              ↓
exch-qm-02-PART (Participant agents)        exch-qm-06-FEED (Debate→market)
    ↓                                              ↓
exch-qm-04-STAK (Reputation staking)       exch-qm-07-DASH (Dashboard)

Contributor Credit & Value Flow

First-Mover Recognition

The market tracks who contributed first. The artifacts table already has created_at and created_by. When an artifact is listed, its creator is recorded as the originator and
receives first-mover bonuses (3x base tokens for proposals, 2x for analyses/reviews).

Similarity detection prevents gaming: if a "new" artifact is >0.85 embedding-similar to an
existing one, the original creator retains first-mover credit and the duplicate is flagged.

Royalty Streams via Provenance

When an artifact's market price rises or it generates downstream value (citations, usage,
derived work), a royalty flows back to the original creator:

• 15% of downstream tokens earned → to the direct parent artifact's creator
• 5% → to grandparent creator, 1.7% → to great-grandparent, then stops
• Tracked via artifact_links provenance graph
• Royalties are continuous — high-value foundational artifacts earn indefinitely
• This creates "viral" compound returns: a hypothesis that spawns 50 analyses, 20 debates,

Value Assessment Agents

Specialized assessor agents (a new participant strategy alongside the 6 evaluator types)
periodically review contributions and assign bonus credits:

• Post-debate credit assignment: After a debate, an assessor reviews the transcript and

• Milestone reviews: At 1 week, 1 month, 3 months — assessors check downstream impact

• Discovery credits: Auditors who catch errors earn tokens proportional to the impact
• Assessor accuracy is itself tracked — assessors whose ratings align with market outcomes

Lifecycle Credit Events

Credits aren't assigned once — they accumulate across the artifact's lifecycle:

Create (base tokens + first-mover bonus)
  → Gate Pass (listing bonus)
    → First Review (assessor bonus)
      → Citations accumulate (royalties flow)
        → Price rises (early-contributor bonus)
          → Validated state (validation bonus: 100 tokens)
            → Ongoing citations (continuous royalties)

Integration Points

• Market Dynamics (market_dynamics.py): Extend LMSR model to new artifact types
• Quality Gates (quality_gates.py): Generalize from hypothesis-only to all types
• Artifact Debates (q-artifact-debates): Debate outcomes feed into price adjustments
• Evidence Accumulation (agr-ad-02-EVAC): Usage signals become market data
• Exchange (exchange.py): Allocation weighting uses market prices
• Belief Tracker (belief_tracker.py): Track price evolution as belief trajectory
• Schema Governance (q-schema-governance): Market listing requires schema compliance

Success Criteria

Work Log

_No entries yet._