Fractal

Fractal is the system I use to study allocation as a process rather than a one-shot optimizer output. The core artifact is the decision journal: thesis version, rationale, confidence, price context, and links between intent and fill reconciliation.

Architecture

• Perception — market and context observations enter a workspace.
• Hypothesis — a thesis proposes what might matter next.
• Allocation — a controller translates intent into weights or trades.
• Risk — gates can reject, shrink, or hold before any simulation.
• Journal — every path (executed, rejected, held) is written with provenance.

Design rationale

If the journal can't explain a trade, the trade doesn't count as research. That standard shows up across my other work: evidence trails in AuditLayer, leakage controls in the spatial evaluation, and ablations in the systems thesis.

What stays private

Implementation details, private prompts, live credentials, and unreleased performance runs are not published. This page describes the shape of the system and the standard it holds itself to.