Orientation

What this is

A thesis operating system for AI-economy investing, built by one person as an application to Sigil Fund's Supernova cell.

Most investment research dies as prose. Progenitor treats a thesis as something you can run: every claim is written down, and evidence for and against it accumulates automatically from public sources until the claim is confirmed, refuted, or drifting.

One pipeline carries the whole system: a source fact is ingested and stored with its provenance, resolved to a durable identity, rolled up into an investment object, classified as thesis evidence, and surfaced for a human decision. Nothing in that chain is a recommendation, and nothing skips a step.

The division of labour is deliberate: agents propose, tested code decides. Language models extract facts from filings and news and draft the briefs. Every score, gate, and posture is deterministic code with unit tests — so what the fund reads is judgment, and what it counts on is arithmetic.

The five-minute path

In order — each page assumes the one before it

Agent access

Query it from your own tools, not just this dashboard

Progenitor speaks MCP at /api/mcp, so your own agent can ask it about investment objects, claims, cycle state, the ledger and the briefs — with every fact carrying the source URL behind it. The surface is read-only by architecture: an agent can read the decision layer, but no tool can take, change or remove a stance. Those stay human, signed and reasoned. Setup and tool list ↗

Provenance

The constraint the whole system is built around

Every fact here traces to a public source you can open yourself — filings, repositories, package registries, job boards, news feeds — stored with its original URL before anything derives from it. Only sources whose terms permit it feed scores and alerts; nothing login-gated or person-sensitive is ever collected. System shows the coverage, freshness, and running cost of all of it.

Rado Súkala · sukala@me.com