Status: pre-launch. This page describes target design; this document does not establish a deployed service.
RealityFutures (COULD) — Scenarios and Forecasts
Domain: realityfutures.com | Pipeline stage: COULD (predict) | Topology: AXIOM (N:1 convergence)
Current position: RealityFutures is a later prediction layer, not a current opening commercial offer.
What RealityFutures Does
RealityFutures explores possible futures through scenarios and uncertainty-bearing forecasts. Sources, assumptions, method, time horizon and limitations remain visible. A forecast is not a promise, a recommendation or permission to act. Cortex is a memory capability within Agentz's intended Cloud Computer, not a separate Menexus product. “Probabilistic cortex” in earlier descriptions was an analogy, not an assignment of memory ownership to ReFu or Menexus. Menexus owns the proposed ensurance, insurance and assurance response to risk; a ReFu forecast does not create coverage.
Prediction markets are one researched mechanism, not a requirement for every forecast. A market price expresses a market estimate under its rules; it does not establish the event's true probability or prove calibration.
Historical market-first presentation
Historical market-first design — preserved, not current performance evidence
RealityFutures takes Circle's verified facts and converts them into probabilities. Not opinions. Not forecasts. Prices. Every claim about the future becomes a market. The market price IS the probability. Skin in the game replaces credentialism.
Current Evidence Note (2026-04-19)
- code-only register:
P = 0.59 - contracts-plus-web-app lane is real in code
- Forge test lane passes
- organism-local proofs still rely on a statistical F2 fallback rather than live deployed market-backed emission
The Price IS the Probability
ReFu uses Logarithmic Market Scoring Rules (LMSR) — the same mechanism behind prediction markets that consistently outperform expert panels, polls, and institutional forecasts. LMSR provides:
- Infinite liquidity: The automated AMM always offers a price. No order book needed.
- Bounded loss: The AMM's maximum loss is known in advance (the liquidity parameter b).
- Information aggregation: Each trade moves the price toward the true probability. Traders who know more profit. Traders who don't, lose. The market learns.
When Circle publishes a signal — "OPEC to cut production Q3" with Watchman Score 0.78 — the intended ReFu path is to spin a market: "OPEC cuts production by >500K bbl/day in Q3 2026." The opening price reflects the Watchman Score. Traders then push it toward truth.
The Ensemble Engine
ReFu doesn't rely on markets alone. Behind each market, a model ensemble provides baseline forecasts:
- TimesFM: Google's foundation model for time-series forecasting (pretrained on 100B+ real-world time points)
- Chronos: Amazon's probabilistic time-series model (T5-based, zero-shot capable)
- Custom specialists: Domain-specific models trained on relevant historical data
The ensemble generates a prior probability. The market adjusts it. The final output is a ProbabilityBundle — a structured object containing the current market price, model ensemble estimate, confidence interval, volume, and time decay.
AXIOM: a recorded market outcome
In the fully live market path, a resolved ReFu market would be recorded as an AXIOM: many predictions converge on one outcome under the market's rules. That is a money-weighted resolution, not consensus or proof of truth.
Financial accountability adds context, not automatic evidential superiority. Provenance, resolution quality and independent outcome evidence remain necessary. Where live settlement is not evidenced, this page describes intended COULD architecture, not runtime proof.
AXIOMs may feed back into Circle's source-reputation model and into Agentz's decision context after a valid outcome receipt. This is target coupling, not a claim of live market settlement.
The F2 Interface
ReFu's external F2 output to Agentz is the canonical ForecastPacket. ProbabilityPacket remains a compatibility representation and RFContext remains internal. A forecast preserves its question/scenario, source bundle, timing, uncertainty and warranted estimates. Market identifiers, prices, liquidity and resolution metadata apply only to market-backed forecasts. This explanatory boundary does not add mandatory fields to a schema.
Agentz consumes valid forecasts to compile unsigned proposals. Missing
probability or warrant yields UNKNOWN/HOLD, never 0.5. ReFu never
recommends, authorizes, or executes; it evaluates possibilities with uncertainty.
Retained revenue proposals
Rates, fee bases, recipients and instrument identities remain unresolved; these proposals are not available commercial terms. The LMSR liquidity parameter is not itself a trading fee.
- Trading fee or spread: A separate proposed charge, not implied by the LMSR liquidity parameter
- Market creation fees: Paid by entities requesting custom markets
- Data licensing: ProbabilityBundles sold as API feeds to institutional clients
- AXIOM archive: Historical resolved markets as structured datasets
Architecture
- Frontend: Next.js
- Contracts: 18 Solidity contracts (Foundry), migrating to Hedera native services
- AMM: LMSR engine (contract lane exists; live deployment is a separate proof question)
- LP-100: Compatibility research only; this page does not establish adopted market-pool instruments, supply, rights or distributions. See LP-100 status.
- Settlement: Target path: on-chain or externally anchored resolution with oracle verification
This page describes the intended COULD architecture. Treat live deployed market claims separately unless a dated runtime/public proof is cited.
ReFu works from sourced observations. It explores possible futures; it does not decide, authorize or execute.
Related pages
Historical predecessor wording — retained on 2026-09-27, not active rules
The displaced passages are preserved verbatim for lineage. This retention date is not their original authorship date.
Predecessor passage 1
ReFu's external F2 output to Agentz is the canonical ForecastPacket.
ProbabilityPacket remains a compatibility/concrete representation and
RFContext remains internal. A valid forecast names:
- Market ID and question text
- Current LMSR price (a market estimate, not proof of the event's true probability)
- Ensemble model estimate
- 95% confidence interval
- Market depth and volume
- Time to resolution
- Related AXIOM history
Historical predecessor wording — retained on 2026-09-28, not active rules
Editorial reconciliation of this public source only; no upstream contract is amended.
Predecessor passage 1
AXIOMs are stronger than Circle signals because they carry financial accountability. Someone paid for being wrong. Where live settlement is not yet present, this page should be read as intended COULD architecture rather than as blanket runtime proof.
Predecessor passage 2
Revenue Model
- Trading spread: Built into LMSR liquidity parameter
Predecessor passage 3
- LP-100: 100 equity tokens per market pool, conservation enforced (total always = 100)