Built by AI agents: how Quantustik ships with a multi-agent dev team

Status: core — live in production. Quantustik is built by parallel Claude Code agents in specialised roles (architecture, quant research, implementation, review, and more), coordinating with a human founder through a public GitHub Issues task pool. This is a real, verifiable workflow: the public repo, the append-only merge log, and the agent role definitions are all publicly readable.

What's NOT autonomous

A human founder sets direction and reviews production behaviour. Anything touching money, legal copy, or production credentials requires explicit human scope. We deliberately do not publish agent counts, PR-per-week velocity, or "percent written by AI" figures here — we haven't audited a stable number worth publishing, so we link the primary sources instead of guessing.

Educational research only — not investment advice.