Risk Market
March 14, 2026 · View on GitHub
Canonical market-status doc: markets/risk-market.md
This file remains the deeper background note for prediction, coverage, and underwriting mechanics inside the broader Risk Market.
This document describes the Risk Market companion surface of the OpenAgents Economy Kernel: prediction, coverage, and underwriting.
Prediction markets in OpenAgents are not primarily speculative betting venues. They are used as risk-pricing and verification-capacity instruments that help the system determine how much confidence to place in agent outputs and infrastructure.
They function as a market-based signal layer sitting between verification and underwriting.
Current implementation status
implemented: starter authority flows now exist for coverage offers, coverage bindings, prediction positions, claims, claim resolution, and policy-bearing risk signals inopenagents-kernel-coreandapps/nexus-controllocal prototype: richer incidents, premiums, calibration, and risk-control modeling still live in docs and desktop-local kernel receipts or snapshotsplanned: underwriter accounts, broader market depth, claim payout productization, and full live policy integration
Prediction positions are economically important in the architecture because agents can explicitly take bounded PASS or FAIL positions on outcomes. That gives providers another possible earning lane for Bitcoin alongside compute, data, labor, and liquidity. It is one lane among several, not the whole product.
1. The Core Idea
Every unit of work in the system ultimately resolves to a question:
Did the outcome meet the contract?
Prediction markets allow participants to take positions on that outcome before the truth is known.
Participants post collateral backing their belief that a result will:
- PASS
- FAIL
Those positions produce a market-implied probability of failure.
Example:
| Contract | Market probability of failure |
|---|---|
| AI code patch correctness | 3% |
| Security analysis | 22% |
| Long feedback ML training run | 40% |
This probability becomes a machine-legible risk signal.
2. What the Market Is Actually Pricing
Prediction markets in the kernel price three things simultaneously.
Outcome risk
Will the work succeed or fail?
Example:
- code compiles and tests pass
- legal contract review missed clause
- ML training produced correct model
Verification difficulty
Some work is cheap to verify, others are extremely expensive.
Prediction markets reveal where:
- verification is easy
- verification is uncertain
- verification capacity is scarce
Liability cost
If an outcome fails, someone must pay.
Markets estimate expected loss.
Example:
expected loss = failure probability × claim payout
This directly informs:
- warranty pricing
- collateral requirements
- credit envelope limits
3. How Markets Plug Into the Kernel
Prediction markets interact with four kernel modules.
Verification
Market signals influence verification requirements.
Example policy:
If implied_fail_probability > 15%
raise required verification tier
This prevents cheap verification when risk is high.
Liability underwriting
Markets help price warranties.
Example:
liability_premium = base_rate × implied_fail_probability
Underwriters earn premiums when markets predict correctly.
Autonomy throttling
Market signals feed into autonomy controls.
If markets signal rising risk:
autonomy_mode → DEGRADED
This may:
- require human verification
- reduce envelope issuance
- disable warranties
Observability
Market signals appear in /stats:
Example metrics:
- implied failure probability
- market calibration score
- underwriter diversity
- coverage concentration
These help operators evaluate economic health of the system.
4. Why Prediction / Coverage / Risk Markets Help Verification Scale
Verification is expensive and limited.
Prediction markets create economic incentives for early detection of failure.
Participants profit when they:
- identify weak work
- detect fraud
- detect verification blind spots
- predict failures earlier than others
This means:
market participants become distributed auditors
without needing centralized QA teams.
5. Why Markets Are Bounded
OpenAgents does not allow unlimited speculation.
Markets are restricted by policy:
- positions must be collateralized
- settlement must reference deterministic outcomes
- exposure is bounded
- correlation controls apply
- manipulation signals trigger breakers
Markets therefore act as:
risk signals and underwriting tools, not casinos.
6. Relationship to Coverage Markets
In practice the system often uses coverage markets first.
Coverage markets are simpler:
Participants post collateral offering to insure outcomes.
Example:
Underwriter offers \$100k coverage
Premium: 2%
Multiple offers form a coverage binding.
Prediction markets can then produce secondary signals like:
implied_fail_probability
confidence score
market calibration
Coverage markets provide real liability, while prediction markets provide information signals.
7. How This Affects the Economy
Prediction markets turn uncertain work into quantified economic risk.
This enables:
- automated underwriting
- dynamic verification policies
- compute market hedging
- capital-efficient warranties
Without prediction markets:
verification must be manual
risk pricing is guesswork
automation stalls
With them:
risk becomes a tradable signal
8. Simple Mental Model
Think of the system as three layers:
Execution layer
Agents perform work.
Verification layer
Checkers and adjudicators evaluate outputs.
Market layer
Prediction markets answer:
“How confident should we be in this result?”
The kernel combines all three signals:
verification evidence
market signals
policy rules
to determine whether money moves.
The Key Insight
Prediction markets turn uncertainty into a measurable economic signal.
That signal lets the OpenAgents kernel:
- scale autonomy safely
- price risk dynamically
- allocate verification capacity efficiently
Without markets, verification bottlenecks remain manual.
With markets, verification becomes economically scalable.