pharos

Pharos — Pitch

Agora Agents Hackathon · Canteen × Circle × Arc · RFB 03

The 90-second version

Every macro prediction market today shows one number: the crowd’s guess. There is no anchor, no fair value, no model. Pharos changes that.

Pharos opens binary markets on the prints that move the world — CPI, payrolls, GDP, the Fed — and on every single one it posts a second number: the Lighthouse Macro framework’s probability. Not a vibe. A point-in-time quantamental nowcast from a 2,500-series model, the same engine behind a live institutional research product. The crowd trades against the framework. The spread between them is the signal, and it is on-chain, in USDC, on Arc.

Three agents run it. One reads the economic calendar and opens markets two weeks ahead, seeded with liquidity at the framework’s fair value. One re-posts the framework price as new data lands. One resolves every market on the official government print. No human touches it.

Why this wins, and why it is hard to copy

It is not a better AMM. The AMM is a forked, provably-solvent CPMM. The moat is the pricing oracle: a macro model with a decade-plus of methodology behind it, run point-in-time so there is no look-ahead, and honest enough to shrink its own confidence when a signal has no demonstrated edge. A team can fork our contracts in an afternoon. They cannot fork the framework.

The Arc angle (three slides)

1. USDC-native gas. A bettor pays for the bet and the gas in USDC. No “acquire the chain’s gas token first.” For a macro punter that friction is the whole funnel.
2. Encrypted mempool. A seven-figure position on a CPI print does not leak its intent to the market maker before it fills. Polymarket cannot claim that. It is the institutional wedge.
3. Instant USDC settlement. Markets resolve on the print and pay the same block. No bridge, no wait.

The demo (it runs today)

make demo on a local chain, with real framework prices from the live Lighthouse database:

1. The creator agent opens six 2026 markets, each seeded at the framework’s point-in-time fair value (CPI 83%, NFP 46%, GDP 50%, FOMC 30%).
2. A simulated crowd trades NFP toward NO — the crowd price drops to 13%, the framework holds 46%, a +33pp edge opens and is visible on-chain and in the UI.
3. A market reaches its release, and the oracle resolves it on the real CPI print (realized 6.2% vs a 3.0% strike → YES) and pays the winners in USDC.

Traction plan (judging weighs active users / volume)

• Day 1–2: deploy to Arc testnet, the four verticals live, faucet USDC.
• Seed liquidity from a treasury wallet so every market has depth at open.
• Distribution is the unfair advantage: Lighthouse Macro already publishes to a macro audience that argues about exactly these prints every week. The markets are the call to action under the research.
• Each resolved market is a content unit — “the framework said 83%, the crowd said 50%, here is what happened” — which feeds the next cohort.

Ask

Judge it on the moat. Anyone can list a CPI market. Pricing it with a real macro framework, point-in-time, with calibrated and self-aware confidence, is the thing that is rare.

---

Lighthouse Macro · Research · @LHMacro