Skip to main content
v9.0 — Introduction

Design Principles

Six constraints that are non-negotiable. Every architecture decision is measured against them.


1 · Evidence Before Action​

No planning step may proceed without a confidence-scored ECO. If the knowledge layer cannot produce sufficient evidence, the system stops and says so explicitly.

2 · Determinism Over Magic​

Every score, every classification, every gate result must be reproducible from stored inputs. No probabilistic or LLM-gated decisions in the control path.

3 · Explicit Limits​

The system must communicate what it does not know. Unknown is a valid, first-class answer. Silent failure is not tolerated.

4 · Minimal Footprint​

Lane A tasks (80% of all commits) produce zero LLM calls. The system's default mode is hooks, not inference.

5 · Human Authority Preserved​

The system recommends. Humans decide. Below confidence 60 the system is restricted to suggesting human actions only.

6 · Reversibility​

Every decision is recorded in the Decision Ledger. Any task can be replayed from its stored ECO snapshot to debug or calibrate.