# Sybil

> The first prediction market built on frequent batch auctions (FBAs).

## Status

Currently in early access. Sign up at [sybil.exchange](https://sybil.exchange).

## Research

- [Deep Dive into Market Maker Losses on Polymarket](https://sybil.exchange/spike-analysis) — Analysis of sniping behavior and market maker losses on Polymarket's CLOB
- [Understanding Sniper's Tax in CLOB Markets](https://sybilpm.substack.com/p/the-snipers-tax) — How snipers exploit market makers in continuous limit order book markets
- [The Math Behind FBAs: Batch Auction Clearing via Eisenberg-Gale Duality](https://github.com/SybilPM/prediction-markets-are-fisher-markets) — Formal analysis of frequent batch auction mechanisms

## Tools

- [Spike Analysis Dashboard](https://sybil.exchange/spike-analysis/dashboard/) — Interactive dashboard for exploring detected sniping incidents on Polymarket
- [Network Visualization](https://sybil.exchange/spike-analysis/viz/network.html) — Network graph of wallet relationships in sniping activity

## Sybil Agentic Research

Research dashboard ranking 27 prediction markets across 5 tiers (A–E) by what an autonomous AI agent can demonstrably do today. Four independent benchmarks — agent accessibility, CLI/MCP, skill, and framework — all run by autonomous coding agents. April 2026 snapshot.

**Fully server-rendered**: the dashboard content — tier list, findings, landscape table, per-PM benchmark grades, and methodology — is all in the initial HTML response. Fetch-only agents (no JS) see the same content a browser sees.

- [summary.json](https://sybil.exchange/agentic-research/data/summary.json): authoritative hand-curated tier assignments + 6 findings (structured JSON)
- [index.md](https://sybil.exchange/agentic-research/index.md): plain markdown snapshot of the entire dashboard
- [Dashboard](https://sybil.exchange/agentic-research): 5-tier ranked list + per-benchmark grades + per-check evidence
- [Summary](https://sybil.exchange/agentic-research/#/summary): 5-tier ranked list + 6 hot-take findings
- [Landscape](https://sybil.exchange/agentic-research/#/landscape): full 27-PM table with category, volume, tools, links
- [Testing](https://sybil.exchange/agentic-research/#/testing): per-benchmark grades for 20 PMs + per-check evidence
- [Methodology](https://sybil.exchange/agentic-research/#/methodology): formulas, floor rules, grade buckets

Dataset (static JSON — no API key, no CORS, no rate limits):

- [pms.json](https://sybil.exchange/agentic-research/data/pms.json): 27 PM entries — name, category, chain, volume, website, twitter, coreTools, aiTools
- [accessibility.json](https://sybil.exchange/agentic-research/data/accessibility.json): per-PM agent-accessibility grades (15 checks, weighted, max 23)
- [tests.json](https://sybil.exchange/agentic-research/data/tests.json): CLI/MCP, skill, and framework benchmark results

Methodology docs:

- [Agent Accessibility](https://sybil.exchange/agentic-research/methodology/agent-accessibility.md) — 15 checks, max 23, weighted
- [CLI/MCP Test](https://sybil.exchange/agentic-research/methodology/cli-mcp-test.md) — 18 checks, max 23, trading 2× weighted
- [Skill Test](https://sybil.exchange/agentic-research/methodology/skill-test.md) — 8 milestones, max 13, M5/M7 weighted 3pt
- [Framework Assessment](https://sybil.exchange/agentic-research/methodology/framework-assessment.md) — 5 quality dimensions × 0–3, max 15

## Links

- Website: https://sybil.exchange
- X / Twitter: https://x.com/sybil_pm
- Substack: https://sybilpm.substack.com
- GitHub: https://github.com/SybilPM