Skip to content

Summary

Scope note (2026-09-09): within the daily cycle (operating rhythm), this workflow's steps compose as follows — step 1 (intake) = rhythm step 2; step 2 (retrieve) = rhythm step 3; steps 3–5 (generate → EV → stress/risk) = rhythm step 4 (with the decision function's canonical gate order Fit → EV → Risk applying inside them); steps 6–7 (publish/approval) = rhythm steps 6–7. Run ad hoc, it is a standalone path. The canonical gate order of the decision function always governs where the two documents differ.

This playbook defines the only sanctioned path from news to a tradable hypothesis. Every strategy candidate must traverse all seven steps; a candidate without a passing EV run or a baseline comparison is not tradable and stays status: draft.

The Loop

Step Action Consumes Produces
1 Intake Current news / market observation Dated concept in 95-inbox per the intake convention
2 Retrieve Inbox item Relevant wiki concepts: regime state from 60-regimes, strategy vocabulary from 30-strategies, vol context from 20-volatility
3 Generate candidates Retrieved concepts 1–3 candidate structures, each mapped to an abstract template in 30-strategies
4 Compute EV Candidate + market data Receipt from the matching Attested Computation — never improvised arithmetic
5 Edge check EV receipt EV must beat the IV−RV and GARCH/HAR baselines and include explicit costs; else discard (record why in the playbook)
6 Stress test Candidate Walk the tail-failure cases: 2018, 2020, 2024; forbidden structures per regime from regime-dependent delta exposure
7 Publish playbook Steps 4–6 evidence Dated concept in 90-playbooks (status: draft), submitted for human approval

Rules

  1. Steps 1–7 are sequential; a failed gate (step 5 or 6) ends the run with a recorded reason.
  2. The LLM agent may execute steps 1–6; step 7 requires human approval before any playbook becomes tradable (see LLM agent roles).
  3. Realized outcomes are written back onto the playbook concept as # Outcome sections, and repeated failure patterns update the regime mapping — the feedback loop that makes this wiki learn.

Links