The Oracle by Polymarket vs
SimpleFunctions.

The Oracle is Polymarket's in-house newsletter and podcast. It publishes curated news, market insights, and analysis from the largest prediction-market platform, targeting human readers who want editorial coverage of what's moving on Polymarket. It is a content product, not an API or data platform. If you need programmatic access to prediction market data, prices, or execution, The Oracle is not the right tool — it serves a different purpose entirely.

SimpleFunctions delivers market intelligence in machine-readable form rather than human-readable editorial. Trade ideas at /api/public/ideas include conviction scores and catalysts. The world snapshot at /api/agent/world gives an ~800-token structured summary of current market state. The thesis evaluation cycle runs a full news scan, price refresh, milestone check, LLM evaluation, and confidence update loop. None of this is a newsletter; all of it is structured data an agent or developer can act on directly.

POST /api/thesis/create takes any sentence — 'ECB cuts before July' or 'Nvidia loses GPU market share' — and decomposes it into a causal tree of testable sub-claims. Each node is mapped to tradeable contracts on Kalshi and Polymarket where edges exist. An evaluation heartbeat runs continuously: news scan, price refresh, milestone check, LLM evaluation, confidence update. New signals can be injected at /api/thesis/{id}/signal. Public theses are forkable. No other prediction market platform publishes anything structurally equivalent.

The CLI (npm i -g @spfunctions/cli) provides 60+ commands for market browsing, thesis creation, and portfolio monitoring with no coding required beyond installation. The MCP server integrates into Claude Code or any MCP-compatible client with a single command. If you want editorial analysis with zero setup — no installation, no accounts — The Oracle by Polymarket is the simpler starting point for that specific need.

Every proposed trade passes through a 7-gate risk cascade before execution: kill switch, position limits, drawdown gate, regime check, and additional guards. The system uses a 1M-context LLM with 13 data sources to reason about each trade candidate. Gates run sequentially — a trade that fails any gate is rejected without reaching the next. The result is an autonomous agent that can operate continuously while respecting pre-defined risk boundaries. This is a fundamentally different capability class from any content or newsletter product.

Hit /api/calibration directly with curl. It returns SF's own Brier scores by venue, category, and price bucket, computed over the past 90 days on resolved markets. Current figures are Kalshi 0.20 and Polymarket 0.12 on T-24h price. All inputs are public Polymarket and Kalshi settlement data. You do not need to trust the claims — you can rerun the computation yourself using public market history. Most prediction market services do not publish equivalent machine-verifiable accuracy figures.

Yes — Polymarket is one of the two primary venues in SF's cross-venue index, alongside Kalshi. All 48K+ active Polymarket contracts are normalised, indexed with computed indicators, and accessible via REST, CLI, and MCP. The thesis system can map sub-claims to Polymarket contracts specifically. Where SF differs from The Oracle is format: SF outputs structured data for programmatic consumption; The Oracle produces human-readable editorial content. The underlying venue overlap is significant; the product surfaces do not overlap.