An evidence-first autonomous quantitative research lab.
Genesis searches for trading edge, tries to falsify it, measures it after costs, and keeps real execution locked until evidence survives.
Live HQ · Architecture · Research principles · Roadmap · Reproducibility · Latest research run · Contribute
Current safety state:
LIVE_LOCKED/ paper-shadow research only.
Genesis does not claim a proven profitable live edge today.
Most trading repositories optimize for activity:
signal -> order -> backtest chart
Genesis optimizes for evidence:
real market data
↓
edge hypothesis
↓
research / replay / shadow
↓
costs + slippage + fill realism
↓
walk-forward / OOS / holdout
↓
economic truth gates
↓
PAPER / SHADOW
↓
LIVE_LOCKED
The goal is not to make a bot look busy. The goal is to determine whether an edge is real enough to deserve capital.
| Research lane | What it studies | Execution state |
|---|---|---|
| Futures research | systematic trend, breakout, mean-reversion, regime and forward-paper studies | Paper only |
| DEX / MEV arbitrage | same-block route economics, liquidity, simulation, inclusion and failure costs | Shadow only |
| Maker microstructure | queue-aware fills, microprice, depth imbalance, adverse selection and inventory risk | Research only |
| Cross-market research | lead/lag, positioning, funding, taker flow and derivatives context | Research only |
| Validation layer | OOS, walk-forward, holdouts, protocol locks, kill criteria and economic accounting | Enforced before promotion |
| Agent layer | automated research, diagnostics, reporting and orchestration | No autonomous real-capital authority |
Genesis deliberately allows a strategy to fail. A clean rejection is a useful research result.
A candidate is not considered good because it:
- has a high backtest return;
- generated many trades;
- found a large gross spread;
- touched a hypothetical maker order;
- produced an impressive chart.
It becomes interesting only when it survives net economics and forward evidence.
Examples of things Genesis explicitly models or gates:
- trading fees and route costs;
- slippage reserves;
- failed-attempt cost;
- quote freshness;
- liquidity confidence;
- queue coverage;
- empirical fill probability;
- adverse selection;
- inventory risk;
- drawdown;
- walk-forward / out-of-sample behavior;
- holdout discipline.
GENESIS HQ
│
┌─────────┴─────────┐
│ │
Research Surface Backend / Agents
│ │
┌───────────┼───────────┐ │
│ │ │ │
Futures Arbitrage Microstructure
│ │ │
└───────────┴─────┬─────┘
│
ECONOMIC TRUTH LAYER
│
costs · fills · slippage · risk · evidence
│
walk-forward · OOS · holdout
│
PAPER / SHADOW ONLY
│
LIVE_LOCKED
For the deeper module map, data boundaries and extension points, see docs/ARCHITECTURE_PUBLIC.md.
Genesis includes a conservative maker-research lane that avoids one of the most common paper-trading errors: assuming that price touch = fill.
maker_microstructure_shadow_v1 requires measured queue coverage and empirical calibration before it can emit a shadow candidate. The OKX RPI research path captures:
- best bid / ask and best-level queue size;
- microprice and microprice skew;
- depth at 10 / 25 / 50 bps;
- depth imbalance;
- order-count imbalance;
- synchronized aggressive taker flow.
Unknown fill probability, adverse selection, stale evidence or impossible spread capture fails closed to NO_GO.
- Node.js 20+
- npm
- Git
git clone https://github.com/Blackleets/genesis-hq-lab.git
cd genesis-hq-lab
npm install
cp .env.example .env
npm run start:uiOpen the Vite URL shown in the terminal.
Windows PowerShell:
Copy-Item .env.example .env
npm run start:uiNo private key is required for the read-only / paper research surface.
npm run startnpm run test
npm run typecheck
npm run buildFocused MEV / arbitrage checks:
npm run mev:shadow:testGenesis is intentionally difficult to promote to real money.
live_mode = falseis the default project contract.- Current MEV and maker-microstructure lanes have no signing authority.
- Do not commit API keys, private keys, seed phrases or
.env. - Public wallet addresses may be used for read-only observation only where explicitly documented.
- Missing evidence must degrade to
NO_GO, not a guessed value. - A UI must never present fabricated market or trading data as live.
Read AGENTS.md before making changes.
Contributors are welcome in areas where evidence quality improves without weakening the safety boundary.
Good contribution categories:
- new read-only market-data adapters;
- new research hypotheses with predeclared evaluation rules;
- realistic transaction-cost or fill models;
- replay and data-quality tooling;
- statistical validation;
- observability;
- tests for fail-closed behavior;
- documentation and reproducibility.
Start with CONTRIBUTING.md. For evidence conventions and replay discipline, see docs/EVIDENCE_SCHEMA.md and docs/REPRODUCIBILITY.md.
A useful contribution should answer at least one of these questions:
- What market behavior are we measuring?
- What evidence would falsify the hypothesis?
- What costs or execution assumptions could erase the edge?
- How will this be tested forward, OOS or on a sealed holdout?
genesis-hq-lab/
├── AGENTS.md # mandatory AI / contributor safety rules
├── src/ # React/Vite HQ and local research surfaces
├── server/
│ ├── genesis/ # quant, MEV, microstructure and evidence engines
│ ├── crypto/ # crypto/futures research and paper systems
│ ├── research/ # research studies and locked protocols
│ └── tests/ # regression and economic-safety tests
├── supabase/ # persistence / scheduled infrastructure
├── .github/workflows/ # research capture, validation and deployment CI
├── docs/
│ ├── ARCHITECTURE_PUBLIC.md
│ ├── EVIDENCE_SCHEMA.md
│ ├── REPRODUCIBILITY.md
│ ├── RESEARCH_PRINCIPLES.md
│ ├── ROADMAP_PUBLIC.md
│ ├── VISION.md
│ └── DESIGN_DIRECTION.md
└── LICENSE
Evidence > activity.
Genesis should be allowed to say:
INSUFFICIENT_DATANO_GOREJECTEDEDGE_NOT_PROVEN
Those are not failures of the platform. They are how the platform avoids lying to itself.
The project becomes economically interesting only when an edge remains positive after realistic costs, survives forward evidence, and remains stable enough to justify further validation.
The near-term focus is deliberately narrow:
- improve reproducible evidence capture;
- calibrate maker fill probability and adverse selection from observed data;
- strengthen cross-venue arbitrage economics;
- publish machine-readable economic scoreboards;
- make research modules easier for external contributors to extend;
- keep real-capital execution locked until promotion criteria are satisfied.
The scheduled Genesis Institutional Edge Stack publishes a human-readable GitHub Actions summary plus downloadable machine-readable artifacts on each completed research run.
The summary exposes:
- qualified research sleeves and paper cash reserve;
- smart-execution decision (
WAIT,MAKEorTAKE); - maker-market-making expectancy and fill evidence;
- statistical-arbitrage OOS status;
- funding-carry OOS status;
- Solana-liquidity research eligibility.
Open the latest Institutional Edge Stack runs.
These are research/paper measurements. They are not live PnL and are not presented as proof of profitability.
Public HQ: https://genesis-hq-lab.vercel.app
Deployment and optional backend configuration are documented in DEPLOY.md.
Genesis HQ is currently source-available under Business Source License 1.1, with the additional-use grant and Change Date defined in LICENSE.
It is not being described as OSI open source before the Change Date. Personal, educational, research and other non-commercial use is covered by the current grant; commercial use is governed by the license terms.
Build less theater. Measure more truth.