Skip to content
← Back to Module Ecosystem

Operator brief · 67

One authority system: the rule that makes eleven workbooks one machine.

The key idea

Clause one

Each module has a defined job — and the manual publishes the roster.

The architecture doesn't leave module roles to inference; it tabulates them, in authority order, with each module's binding output and — crucially — what may never override it. Gate/drawdown state binds the capital envelope, and nothing overrides it: not daily EV, not setup confidence, not the desire to recover. The throttle binds final tier and pool, immune to checklist enthusiasm, open P&L, and branch preference. The cycle console prepares evidence but does not replace throttle authority. The volatility layer informs stop and trail decisions but yields to gate, cap, and hard failures. The checklist blocks but cannot authorize. CP3 and the scorecard supply weekly truth but not in-the-moment capital gates. The SDE and regime engine diagnose without deploying. And the labs experiment without touching production until promoted. Every row pairs a power with its explicit limit — that pairing is what 'defined job' means.

Clause two

No silent overrides — the hierarchy fails loudly or holds.

The second clause targets the specific way multi-tool systems decay: not through dramatic rule-breaking, but through quiet local exceptions — a console read treated as deployment permission this once, a volatility signal traded through a gate restriction because the setup was special, an EV Lab configuration informally adopted because it kept looking better. Each exception is small; the accumulation is a system whose real rules diverge from its written ones until nobody can say which module actually governs what. The doctrine's answer is that overrides must be loud: the legal exception paths exist, they log, and they surface in the override-stress statistics. What has no legal path is silence — an authority relationship rearranged without a record. The hierarchy is allowed to be contested. It is never allowed to be quietly ignored.

FigureThe published authority order — each module's rank, from the architecture's own table
Trade Plan doctrine9outranks any output or feelingGate / DD state8capital envelope · lockThrottle Panel7final tier · pool · riskCycle Console6prepares, never authorizesVolatility layer5yields to gate & capsLaunch Checklist4block-only powerCP3 / Scorecard3weekly truth, not gatesSDE / Regime2diagnose, never deployEV Lab / R&D1no authority until promoted

The manual's authority sequence rendered as rank (higher bar = higher authority). Each module's power is scoped by the row above it; the plan itself outranks everything, and the labs bind nothing.

Clause three

No contamination without approval — production purity as policy.

The third clause guards the boundary the sandbox doctrine formalizes: experimental tools are production-shaped on purpose — that's what makes their tests meaningful — which is exactly what makes informal adoption so easy and so corrosive. A scenario profile that starts getting used because it looked better for three weeks hasn't been promoted; it has leaked, and the leak poisons twice — the production record now reflects rules that were never approved, and the candidate's own evaluation is contaminated by the uncontrolled live trial. The approval path exists precisely so ideas can cross the boundary with their evidence intact: contract, persistence, documented verdict, deliberate adoption. Contamination isn't the use of a bad idea. It's the ungoverned use of any idea — including the good ones.

  • Production-shaped R&D tools are the most valuable and the most dangerous — realism is what makes leaking easy.
  • A leaked improvement is still a governance failure: the record can no longer say which rules produced which results.
  • The approval path is not bureaucracy around change — it is what makes change legible afterward.

The clause behind the clauses

Why unity is worth this much enforcement.

All three clauses defend the same asset: the interpretability of the whole. A system with defined jobs can be diagnosed — when something breaks, the responsible module is findable. A system without silent overrides can be trusted — its written rules and its real rules are the same rules. A system without contamination can be evaluated — its record was generated by a knowable configuration, which is what the benchmark comparison, the attribution tables, and the entire evidence stack silently assume. Break any clause and no individual workbook fails; the ensemble just stops being one thing. Eleven excellent spreadsheets that can't vouch for each other are worth less than the machine they were supposed to be.

The key idea

The ecosystem's unity is a discipline, not a feature.

Nothing in the software makes MARS one system — Excel would happily let every module contradict every other. The unity is manufactured daily by the operator honoring three clauses: jobs stay defined, overrides stay loud, and the sandbox stays sealed. That's the closing thought the platform category deserves: the architecture provides the structure, the manuals provide the doctrine, and the machine those produce is only as unified as the operating discipline that runs it. One authority system — kept that way on purpose, one week at a time.

Connected inside MARS

Every brief documents the same shipped system.

The complete MARS package — eleven workbooks, three TradingView indicators, the full manual library — $497.