MONTREAL.AI / SKILLOS
Autonomous RSI Marketplace Flywheel
Recursive self-improvement on scalable AI-work marketplace liquidity, skills, pricing, routing, and compounding business value.
Current status
PASSED_AUTONOMOUS_RSI_MARKETPLACE_FLYWHEEL_MARKET_PROOF
No human review. No customers. No private data. No API keys. No previous-domain reuse. Deterministic holdout benchmark.
+90.9 ptsfully-correct gain
87.0%value capture
+83.7 ptscompounding index gain
$19,438,892,166.31synthetic annual value captured
Recursive self-improvement curve
Before / after on holdout marketplace states
| Metric | Baseline | SkillOS RSI |
|---|---|---|
| Fully correct decisions | 9.1% | 100.0% |
| Market-state accuracy | 9.1% | 100.0% |
| Intervention accuracy | 9.1% | 100.0% |
| Flywheel accuracy | 9.1% | 100.0% |
| Value capture rate | 5.6% | 87.0% |
| Marketplace liquidity score | 13.9 | 96 |
| Compounding index | 11.3 | 95 |
| Material miss rate | 57.6% | 0.0% |
| False intervention rate | 0.0% | 0.0% |
| Avg decision cycle | 30.341 days | 0.329 days |
Final learned marketplace flywheel skills
- skill_compound_repeatable_work — When demand is high and work is repeatable but skill quality is weak, allocate learning capacity to skill releases.
- skill_enterprise_quality_gate — When enterprise demand is blocked by quality risk, install validation gates before increasing volume.
- skill_creator_supply_gap — When demand exists but skill supply is thin, seed creator supply for the scarce categories.
- skill_capacity_constrained_lane — When high-margin demand exceeds runtime capacity, reserve capacity and prioritize high-margin routing.
- skill_private_registry_moat — When traces are customer-specific and repeated, create a private skill registry that compounds inside the account.
- skill_bundle_network_effect — When adjacent workflows reinforce each other, bundle them into a networked multi-skill offer.
- skill_sla_volume_guardrail — When high volume threatens SLA quality, throttle volume until guardrails and validators clear.
- skill_feedback_rich_lane — When feedback density is high, prioritize the lane because each job improves the next.
- skill_outcome_gap_retention — When outcomes lag and churn risk rises, repair the outcome metric before scaling demand.
- skill_enterprise_price_discovery — When enterprise willingness-to-pay is high, reset price floors and connect pricing to verified outcomes.
- skill_model_routing_margin — When model routing destroys margin, improve routing, caching, and trace reuse.
- skill_exit_commodity_trap — When competition is high and differentiation is low, exit or reprice commodity work.
- skill_validator_bottleneck — When validators bottleneck releases, automate repeatable checks and expand validation capacity.
- skill_cold_start_liquidity — When a marketplace lane lacks liquidity, seed benchmark jobs and reference skills.
- skill_long_tail_template — When long-tail work is fragmented, convert it into templates and a skill DSL.
- skill_procurement_pack — When procurement slows adoption, package evidence into repeatable procurement collateral.
- skill_seasonal_elasticity — When seasonal demand spikes, apply elastic pricing and reserve capacity.
- skill_clean_flywheel — Recognize a clean scalable flywheel and avoid unnecessary intervention.
Proof gates
- ✅ business domain highly scalable marketplace flywheel workflow
- ✅ not email workflow
- ✅ not invoice workflow
- ✅ not cloudops workflow
- ✅ not cyberdefense workflow
- ✅ not silicon workflow
- ✅ not metamaterials workflow
- ✅ not generic corporate os workflow
- ✅ not unit economics profit engine workflow
- ✅ no human review required
- ✅ no customers contacted
- ✅ no private data used
- ✅ no api keys required
- ✅ deterministic reproducible benchmark
- ✅ recursive self improvement releases at least 10
- ✅ rsi validation improves monotonically
- ✅ train cases at least 450
- ✅ validation cases at least 225
- ✅ holdout cases at least 900
- ✅ final rules at least 18
- ✅ fully correct gain at least 85 points
- ✅ market state accuracy at least 99 percent
- ✅ intervention accuracy at least 99 percent
- ✅ flywheel accuracy at least 99 percent
- ✅ value capture rate at least 75 percent
- ✅ marketplace liquidity score at least 90
- ✅ compounding index at least 90
- ✅ material miss rate zero
- ✅ false intervention rate zero
- ✅ decision cycle reduction at least 90 percent
- ✅ decision cost reduction at least 90 percent
- ✅ synthetic value captured positive