A nightly positioning instrument for active traders and small fund managers. Every position traces to one machine-stamped Radar source, carries a break condition inherited from that source rather than invented, and is sized by deterministic code. The model never touches a number that moves money.
The active trader's problem is not lack of information. It is the opposite. There are ten thousand words written every morning about every thematic you care about, most of them rephrasings of the same news cycle, none of them telling you what a trading position should look like against the actual structural state of the system.
The structural state lives underneath the news. It is readable, but only if you are watching many categories at once, only if you apply a real test to every headline, and only if you are honest about what a signal does and does not imply for a position.
Most systems that attempt this share one flaw: the model that writes the analysis also picks the trade and decides the size. Sextant is built the other way round. Radar produces the structure. Sextant translates it. A calculator, not the model, decides every weight. The model never invents a thesis, a family, a break condition, or a number that moves money.
Radar publishes inspected theses, measured divergences, situations and register state every night. Sextant runs against that substrate and produces one portfolio. Positions come from two lanes with different speeds and different risk grammars, sized by a single governor.
Every position traces to exactly one Radar-made source: an inspected thesis, or a measured divergence from the night’s snapshot. The model’s job is translation only. A position with no traceable source is a validation failure, not a judgement call.
Thesis-lane positions run on the thesis’s own horizon. Their break conditions are that thesis’s own falsifiers, translated into checkable register levels. The code rejects any exit that does not anchor on the registers in that thesis’s falsifier set.
Divergence-lane positions come from the measured gap between physical state and market price. A divergence proves two sides disagree. It does not say which wins. So the fast lane trades the fact both sides agree on, names the inference it is deliberately not taking, and exits mechanically on closure or absence.
A named instrument that cannot be priced tonight is rejected before sizing. It never reaches the book, the counters, or a proposal. A position already held whose feed goes dark is not silently dropped: it holds its weight and is named in the brief.
Deterministic, logged, replayable code computes every weight, haircut, cap and deployment level. An unmeasured buffer takes the maximum haircut rather than full weight. Unknown is never treated as calm. Weight freed by a cap is released, never pushed into the next-largest bet.
Two records run from the same prices: each night’s book scored on its own, and one continuously-run portfolio charged for every trade and marked against the S&P 500. Records are never rewritten. Corrections are new records carrying their own dates.
Real positions from the book of 27 August 2026. Each names its lane, its single source, the computed weight, and the condition that would end it. Not the theme. The current night’s book is in the terminal.
Data centre power and thermal capacity is the binding constraint on AI compute build-out, not chips. The position expresses the bottleneck through the equipment supplier rather than the compute buyer.
A measured gap between physical state and sovereign credit pricing. The fast lane takes the fact both sides agree on and states plainly the disputed inference it is not taking.
The price is anchored to the value of one good trade per month. It is not negotiable. The 50-subscriber early access cap is honest — not artificial scarcity. When it is full, it closes.
Every position traces to exactly one Radar-made source, stamped by code from that source’s machine identity and never taken from the model. A position that cannot name its source does not ship. Code computes every number that moves money: weights, haircuts, caps, deployment, turnover, scores.
Each thesis names the specific mechanism — not the theme, not the general condition. "Energy prices will rise due to supply disruption" is not a Sextant thesis. "Hormuz blockade achieved 100% transit denial on day one with SPR at IEA 90-day floor" is.
Every position states a one-sentence, observable break condition. What would make this thesis wrong. Not "monitor developments." A specific, checkable condition. Risk management is built into the format.
The brief serves both the human subscriber and automated downstream consumers. Every position carries machine-readable fields sufficient for an execution product to resolve it to exchange, symbol, contract type, and sizing range without guessing.
Quantitative claims in theses are grounded in the signal record or they do not appear. Thesis family labels are reviewed for narrative drift before validation. The product does not manufacture certainty it does not have.
Sextant delivers structural analysis in trading language. The subscriber makes all decisions and executes independently. Sextant does not place trades, connect to brokerages, or act on positions. The architectural separation is preserved at every layer.
Twenty-seven years in uniform, much of it at sea, some of it in command. The ocean does not wait for your reporting cycle. The threat does not announce itself on a schedule that suits your staff process. You learn, in that environment, that the quality of your decisions is largely determined before the moment of decision — by the quality of your reading of what is building toward it.
After the Navy: a startup in the cryptocurrency market of the mid-2010s, where the people who navigated it best were reading the system beneath the market — the leverage concentrations, the structural fragilities accumulating invisibly behind the performance. Then consulting to Navy capability, where an institution I knew intimately was not reading its own system, not because the people making decisions lacked intelligence, but because they had no systematic way of reading the system at the level where the real information lived.
Sextant is the trading form of that discipline. The Radar pipeline reads the system state. Sextant translates it into positions — every one of them traceable to the signals that ground it.