Five books, five mandates.

Every book here is a consumer of the same market read — the cycle clock, the stance engine, the structure detectors and the news pipeline described on the intelligence page. What differs is the question each one asks of it, the capital it risks, and how it is measured. All five are live.

ICT Trades — the flagship

Question	Can a vision chart reader, wrapped in programmatic validation, produce selective ICT structure trades worth funding?
Consumes	The structure layer directly — order blocks, breakers, sweep-failures — plus the four-year cycle clock as posture and live news ingestion as a filter. News can veto or tighten; it never initiates.
Markets	ETH-USD and BTC-USD, spot and perpetuals on a major exchange.
Horizon	Hours to days. A handful of trades a week; median holds around half a day.
Risk	Capped sleeve, fixed per-trade risk fraction, hard stop and three staged targets on every position, stop trailing behind paid targets.
Measured by	The live ledger, position by position, shown as percent return from the book's own starting line.

Evidence posture. The live book is young and small — enough to prove the machinery (fills, brackets, trailing, honest accounting), not an edge. Candle-replay studies of the recorded reads point at a consistent pattern: reads that were directionally right but stopped before paying. The current research front is stop geometry — wider structural stops at constant risk — and the honest expectancy estimate after simulation corrections is modestly positive at small n. We say "under study," and mean it.

Trade Mill — the throughput lane

Question	What does a continuous, honest public record of structured intraday setups look like at volume?
Consumes	Eva's brain over the intelligence API, as an independent service — deliberately not an ICT desk. A BTC stance flip mints an ETH follow-on; a severe headline mints a news idea framed against current structure; funding regime switches, volatility spikes and session opens mint their own. It re-derives none of that analysis.
Markets	Major crypto pairs, intraday.
Horizon	Hours. Volatility-scaled brackets, median holds under half a day.
Risk	Fixed live clip sizes under daily-loss, concurrency and fills-per-day halts.
Measured by	Ideas per day, and the live book's percent return since go-live, published as recorded.

Evidence posture. The mill publishes tens of ideas a day and scores every one; that discipline is the product. Its bracket geometry was rebuilt once after a sweep across the recorded book showed hit rate responds cleanly to geometry, and the change is disclosed on the chart it affects — no restated history, no silently merged epochs.

Kalshi 15m Reversal — the exhaustion lane

Question	Do stretched intraday runs snap back often enough to pay on markets that settle every quarter hour?
Consumes	The stance endpoint for directional posture, and nothing else — it has never run a chart read of its own, and it refuses to trade on a stale read.
Markets	15-minute BTC and ETH event contracts on Kalshi.
Horizon	Minutes. One quarter-hour window at a time.
Risk	Small fixed bank, fixed take-profit and stop multiples, a circuit-breaker after consecutive stop-outs.
Measured by	Percent return and win rate since the current rules went live, with the earlier period disclosed rather than blended in.

Rule. Three or more consecutive same-direction 15-minute candles where the last one sweeps the prior extreme → take the reversal side near the window open. Mechanical trigger, mechanical exit.

Kalshi 15m Wick — the favourite lane

Question	Does the late-window favourite still trade below its realized win rate — and can that gap be harvested after fees?
Consumes	The stance endpoint for market bias, with the same freshness discipline as the reversal lane.
Markets	15-minute BTC and ETH event contracts on Kalshi.
Horizon	Minutes. Entries only with four to ten minutes left in the window, held to settlement.
Risk	One entry per window, priced 67–80¢ only, ask paid up to a hard cap. No exit management by design.
Measured by	Percent return and win rate since the current rule went live on 2026-09-17. The retired earlier ruleset keeps its losses on the all-time record.

Rule. Mid-window, buy whichever side is priced 67–80¢ with four to ten minutes to settle, and hold to settlement. The rule was chosen by a sweep across the recorded book and went live in its current form on 2026-09-17.

YieldGen — the institutional sleeve

Question	Can on-chain carry — collateralized lending plus credit-gated fixed-yield instruments — produce a stable NAV with bounded liquidation risk?
Consumes	The cycle thesis and stance read as posture for sizing collateral risk. This is the book where Eva's long-horizon brain is most directly in the loop — and even here it informs posture rather than triggering transactions.
Markets	On-chain money markets and fixed-yield positions.
Horizon	Continuous. Positions rebalance against explicit health-factor floors.
Risk	Liquidation distance is the risk unit: hard health-factor floors with monitors, and a planner that blocks actions that would breach them. The most risk-free book Eva runs.
Measured by	A daily NAV snapshot, shown as percent return since go-live. The curve includes market moves of the collateral, so it is an accounting record, not an alpha claim.

Evidence posture. The cleanest record of the five: a daily NAV series since go-live with every mark kept. It is built for allocators — stable yield with bounded liquidation risk — and retail deployment is deliberately closed while the waitlist is open.

Why five books rather than one

Each book isolates one bet on the same underlying read: the flagship isolates read quality, the mill isolates throughput honesty, the two event-market lanes isolate the value of the read against priced outcomes, and YieldGen isolates carry with bounded risk. Because they consume a shared analysis rather than each maintaining their own, a book is cheap to add and cheap to retire — and when one fails, the finding is about that book's question, not about the engine underneath it.

Sleeves cap each other off, so a failure in one book is a finding rather than a drawdown for the rest. The portfolio-level claim is deliberately modest: a system that publishes everything, sized so that being wrong is affordable, pointed at the places its edge is most likely to be real.

The natural next products are the ones that need a market read and not a new engine — and that is the point of building the read as an artifact rather than a side effect.