BOOKBY.date

every night has a book-by date · built on Wheelhouse
— market — run —
Wheelhouse has a hundred dials. This is one sentence a day.

The agent reads all 365 nights, every day — and hands you a brief you can read in five seconds.

This morning's brief

—
–calendar
health
—

—

Am I doing better than last year?

—

What should I do?

supervised — nothing changes without your approval

Where's the risk? — 365 nights at a glance

click any day for its evidence

Get your own morning brief

the honest version: power users first, then everyone
STEP 1 · 30 SECONDS
Connect your Wheelhouse account
One read-only key. BookBy never changes anything without your tap.
STEP 2 · OVERNIGHT
The agent reads your history
Every booking you've ever had becomes each night's book-by curve, fare bands, and queue position.
STEP 3 · TOMORROW 7AM
Your first brief arrives
One score, one sentence, never more than three asks. Three things, then silence.
Where it stands today: this page runs one real listing — the builder's. The pipeline is a daily scheduled agent + the Wheelhouse RM API/MCP; multi-listing onboarding is the roadmap, stated here rather than fine print.
For the skeptics: watch it actually work.

8 years of bookings. Wheelhouse's own telemetry. Every number has a receipt.

A real 365-night operator scan against the local strategy database — replay it, click any date, and audit the evidence down to table rows and API endpoints.

365-night scan

idle
scanned 0open 0 booked 0actions surfaced 0

Findings feed

live

Date revenue anatomy

—
—
—
    Why this price — Wheelhouse's own attribution
    This unit's memory

    Honest ledger — agent vs Wheelhouse control

    alpha must be earned
    $0
    projected alpha (bookable, could still cancel)
    $0
    confirmed alpha (funds received)

    Receipts — this is real

    local db, real API

    The signal foundry — how the agent thinks

    a search, not a rulebook
    ~70
    named variables across 27 tables
    10
    combination operators
    ~8,000
    scope-valid signal combinations
    3-sensor
    consensus before any action
    The agent doesn't run fixed rules. It searches combinations of every field it has — unit memory × engine attribution × neighborhood surprise × market position — and only what survives evidence thresholds reaches your morning brief.

    Where the AI lives — and where it deliberately doesn't

    AI where judgment lives · determinism where money lives
    LLM agent (Claude, over Wheelhouse MCP)
    • Built and operates this system as a first-class MCP client
    • Investigates findings and writes every word of the morning brief
    • Researches context the engine can't: local events, market shifts, policy changes
    • Proposes; never posts. Every action requires the owner.
    Deterministic operator engine
    • Scans all 365 nights each morning — same input, same output, auditable
    • Computes verdicts, book-by dates, fare bands from evidence thresholds (cohort n≥4)
    • Writes the receipts: table rows, endpoints, run ids
    • No dice rolls where money moves — by design, not limitation
    Why this split: pricing recommendations must be reproducible and auditable; language, investigation, and judgment support benefit from an LLM. Each does what it's actually good at.