DeepNewz vs
SimpleFunctions.

No. SF does not have a curated human news reader. What SF does have is a thesis evaluation heartbeat that scans news as one of its signal sources when evaluating whether a thesis should update its confidence. That is a programmatic signal input, not a consumer news product. If you want to read news with prediction-market odds attached in a readable UI, DeepNewz is the right surface.

POST /api/thesis/create takes a plain-language sentence and decomposes it into a causal tree of testable sub-claims. Each sub-claim is mapped to live contracts on Kalshi and Polymarket. An evaluation heartbeat then runs on a continuous cycle: news scan, price refresh, milestone check, LLM evaluation, confidence update. You can inject external signals via /api/thesis/{id}/signal. Public theses are forkable. No other prediction-market product exposes this.

Portfolio Autopilot is SF's autonomous trading agent. It uses a 1M-context LLM, draws on 13 data sources, and runs a 7-gate risk cascade — including a kill switch, position limits, drawdown gate, and regime check — before any order is placed. A news-based alert tells you something happened; Autopilot decides probabilistically whether that event warrants a trade and executes if gates pass.

Based on publicly available information, DeepNewz does not document a public REST API. Its product is a consumer web and app interface for reading news. If you need programmatic access to prediction-market data, SF exposes a full REST API with an openapi.json spec, a 60-command CLI, and a 56-tool MCP server, with read operations requiring no authentication.

SF publishes its own Brier scores at /api/calibration, broken down by venue, category, and price bucket, refreshed continuously. Current figures: Kalshi 0.20, Polymarket 0.12 on T-24h price over the past 90 days. These are live numbers you can re-verify yourself with a single curl call. Most prediction-market data products claim accuracy without publishing a verifiable, queryable baseline.

SF is designed for AI agents as the primary consumer, not human dashboards. Concretely: a /.well-known/ai-world-state endpoint, an llms.txt and skill.md for LLM crawlers, an /api/agent/world snapshot (~800 tokens) with a delta endpoint for efficient context injection, a 56-tool MCP server compatible with Claude Code and Cursor, and an openapi.json for automated tool discovery. A human can use SF, but every surface is optimised for programmatic and agent access.

Yes, via the CLI: `npm i -g @spfunctions/cli` gives you 60+ commands for prices, screening, thesis management, and orderbook queries. You can also add the MCP server to Claude Code or Cursor in one line and interact conversationally. The /screen page surfaces pre-computed indicators visually. That said, SF's core design is agent-first; if you want an integrated news-reading experience with market odds, DeepNewz is better suited.