pharos

Pharos

Macro prediction markets on Arc, priced by the Lighthouse Macro framework.

Working name. Pharos = the Lighthouse of Alexandria. Ties to Lighthouse Macro without colliding with the publication brands (The Beacon / The Beam / The Horizon). The name lives in exactly one place per layer — see docs/ARCHITECTURE.md — rename is a find/replace.

Built for the Agora Agents Hackathon (Canteen × Circle × Arc), RFB 03 — macro prediction-market verticals.

---

The one-liner

Every Polymarket / Kalshi macro contract has one price: what the crowd thinks. Pharos contracts have two — the crowd price and the framework price. The Lighthouse Macro 12-pillar model (2,500 series, real database) posts a fair-value probability on every market it creates. The spread between the two is the product.

Binary outcomes on CPI, NFP, GDP, FOMC. Short duration. Auto-resolving on official BLS / BEA / Federal Reserve data. Settled in USDC on Arc, with USDC-denominated gas (no “go buy ETH first” friction) and an encrypted mempool so a 7-figure macro bet does not leak intent to the market maker.

Why this is defensible

It is not a better AMM. The AMM is a forked, well-trodden CPMM. The moat is the pricing oracle: an autonomous agent that reads Lighthouse_Master.db, runs the same pillar models that power the Lighthouse Macro research product, and opens every market at the framework’s fair value. The framework is the market maker’s prior. Nobody else in the field has a quant macro model behind the quote.

Architecture (one screen)

  Lighthouse_Master.db ──► pricing engine ──► /price  (FastAPI, :6910)
   (2,500 real series)      (pillar models)      │
                                                 ▼
  econ calendar ──► market-creator agent ──► PharosFactory.createMarket()
   (BLS/BEA/Fed)     (deploys 14d ahead,        │   seeds liquidity AT
                       seeds at framework p)     │   framework probability
                                                 ▼
                                          PharosMarket  (Arc, USDC)
                                          binary CPMM + complete-set escrow
                                                 ▲
  official print ──► oracle agent ──► PharosMarket.resolve(outcome)
   (FRED/BLS at        (auto-resolve on
    release time)       the data feed)
                                                 ▼
                                       Next.js front-end — dual-price card

Four layers, three of them agents:

Layer	Path	Stack
Smart contracts	contracts/	Solidity 0.8.24, Hardhat, OpenZeppelin
Pricing engine (the moat)	pricing/	Python, FastAPI, reads the real LHM DB
Agents	agents/	Python + web3 (creator + oracle + runner)
Front-end	web/	Next.js 14, wagmi/viem, LHM 23/89/BB palette

Full design: docs/ARCHITECTURE.md. Pitch outline: docs/PITCH.md.

Quick start

make setup            # installs hardhat deps, python venv, web3, web deps
make test             # solidity test suite (market lifecycle, solvency)
make price            # sanity-check the pricing engine against the live DB
make demo             # full end-to-end on a local chain (see below)

make demo runs the whole thing with real framework prices from the live database: spins a local chain, deploys, the market-creator agent reads the 2026 econ calendar and opens markets seeded at the LHM fair value, simulates trades that move the crowd price away from fair value, then the oracle agent resolves a market on a real historical print and pays the winners in USDC.

Status

See docs/ARCHITECTURE.md#build-status for the day-by-day sprint board and exactly what is wired vs. stubbed.

---

Lighthouse Macro · MACRO, ILLUMINATED.