α

Every parameter earned. Every result checked. Every trade protected.

Eight gates stand between an idea and capital, and each one can refuse.

Status as of November 2, 2026

The eight gates

A red or unmeasurable result at any gate halts everything after it. Absence of evidence is never a pass.

M7M6M5M4M3M2M1M0S1 EdgeXS1 gTradeS5 gTrade

M2

Population, data, layered search

  • S1 EdgeXIn progress
  • S1 gTradeIn progress
  • S5 gTradeBlocked
What it proves
Every searched parameter stays inside its authority, on a pinned toolchain, layer by layer.
What stops it
A toolchain mismatch, an authority file changing mid-search, or a smoke test that disagrees.

A trade's life: decide, then protect

Every check before the order exists, and every guard while it is open. Each step says whether it is in the engine today.

Decide before the order exists

  1. 1

    Entry engines propose

    Trend, breakout, mean reversion and liquidity sweep, long and short.

    In engine
  2. 2

    Confluence must clear the minimum

    Order flow, momentum and the volatility term structure add evidence.

    In engine
  3. 3

    Conditions can stand it down

    Regime and trend filters, volatility and shock halts, no-trade zones, cooldowns.

    In engine
  4. 4

    HAI vote, switched off

    A bounded vote that never blocks or resizes. It must beat a random-vote control first.

    In buildControl not built
  5. 5

    Leverage and size from the score

    Correlation, tail, regime and stress scalers can only cut size.

    In engineHonest scale in build
  6. 6

    Stop from volatility, the loss law

    If even floor leverage cannot carry the stop, the trade is refused.

    In engine

A trade that cannot fit the loss law is refused before it exists.

Protect while the trade is open

  1. 7

    Order placed, read back

    Venue leverage and isolated margin read back on pinned, graded files.

    M6 gate
  2. 8

    Four-leg exit ladder

    Three targets plus a runner, with ratchet, trail and time exits.

    In engine
  3. 9

    Honest break-even lock

    Placed at the close of the first-target bar, never filled inside it.

    In build
  4. 10

    Halts still apply

    Volatility and shock halts stand new entries down. The loss law holds across every open position.

    In engine
  5. 11

    Book and score

    Live close booking equals engine booking, scored on a scoreboard registered before trading.

    M7

Build status as of October 28, 2026

How Q7 thinks

Four engines propose. Regime and conditions can stand any of them down.

Λ-Volregime routerTRNDTrendBRKBreakoutMRMean reversionSWPLiquidity sweep

Controls across the whole flow

  • Python authority

    The engine and its parameter spec are the only truth.

  • Fail-closed

    A failure, an unknown and a missing artifact all stop the flow.

  • Fixture-first

    Every fix: a test that fails on the old code and passes on the fix.

  • Validate the consumer

    A change counts only when what reads it still works.

  • Data-informed design

    No search or selection rule without a measurement.

  • Producer ≠ Verifier ≠ Armer

    The builder never grades. Only a Verifier closes.

  • The founder arms

    Capital at risk is the founder's act alone.

Held-out data controls

  • Clearance before any read

    Verifier-cleared, mirrored read-only by a separate seat.

  • Pre-registration

    The test is written down before discovery runs.

  • Reachability dry run

    Proves the gates can be reached before any search.

  • Multiple-testing ledger

    Every cell ever tried counts, cumulatively.

  • Unread data is forward evidence

    Only recent data no study has read can prove an edge.

  • Frozen whitelist

    Cells chosen on discovery data only, then frozen.

Every gate, by lane

Swipe to see all three lanes.

Pipeline status by gate and lane
GateS1 EdgeXS1 gTradeS5 gTrade
M0Scope lock & provenance

In progress

Engine seal re-submitted after a scope fix; independent re-grade pending

3d0c6ef79367

In progress

Engine seal re-submitted after a scope fix; independent re-grade pending

3d0c6ef79367

In progress

Engine seal re-submitted after a scope fix; independent re-grade pending

3d0c6ef79367

M1Structural integrity (honest fills)

In progress

Open structural defects are being fixed and graded one by one

b28b138d854e

In progress

Open structural defects are being fixed and graded one by one

b28b138d854e

In progress

Open structural defects are being fixed and graded one by one

b28b138d854e

M2Population, data, layered search

In progress

New entry test cleared for its first stage; no search runs until an entry passes

651154cc5ea5

In progress

New entry test cleared for its first stage; no search runs until an entry passes

651154cc5ea5

Blocked

Entry test stopped at the reachability check; exit and cost design to be revisited

a48c3706ce52

M3Robustness & selection

Blocked

Last cycle: no candidate cleared the selection floors on validation

8aaa20d699b3

Blocked

Last cycle: no candidate cleared the selection floors on validation

8aaa20d699b3

Blocked

Last cycle: no candidate cleared the selection floors on validation

8aaa20d699b3

M4Deliverable package

Not started

Waits for a selected candidate

482acd8263a4

Not started

Waits for a selected candidate

482acd8263a4

Not started

Waits for a selected candidate

482acd8263a4

M5Independent validation

Not started

Waits for a sealed package to grade

482acd8263a4

Not started

Waits for a sealed package to grade

482acd8263a4

Not started

Waits for a sealed package to grade

482acd8263a4

M6Deploy & arming

Blocked

Live-path checks still open; the founder alone arms

b28b138d854e

Blocked

Live-path checks still open; the founder alone arms

b28b138d854e

Blocked

Live-path checks still open; the founder alone arms

b28b138d854e

M7Verdict & ledger

Not started

Starts with the first live fills

482acd8263a4

Not started

Starts with the first live fills

482acd8263a4

Not started

Starts with the first live fills

482acd8263a4

Documents

  • Agentic AI Risk Management: A Top-Down Approach for CROs

    Haneef R. Haqq · 36 pp · October 2026

    Where autonomous agents fit in the enterprise risk taxonomy, and how to govern them in code.

    Download PDF
  • Agentic AI Risk Management: Six Exhibits for the Board

    Haneef R. Haqq · 6 pp · 26 October 2026

    The board-level companion: six exhibits on agentic AI risk and what to ask for.

    Download PDF
  • Calibration pipeline

    The full method, stage by stage.

    Whitepaper

  • Reference framework

    The AI Governance Framework is published and hosted by FINOS, the Fintech Open Source Foundation. The CRO paper discusses it.

algorithmic trading carries substantial risk of loss. past performance does not guarantee future results. Quant7 Alpha, LLC does not custody client assets, does not provide personalized investment advice, and is not a registered investment advisor.