Phalanx logoPhalanx
LIVE

PHALANX TRADING SYSTEMPHALANXPHALANX [NERD MODE]PHALANX SIGNAL

PhalanxAutonomousAI Trading

03
PHALANX SIGNALLIVE FEED

Intelligence feed · tickers + reasoning · $25/month

Coming soon

A fully autonomous multi-agent trading stack — end-to-end, running live.

PHALANX :: PRODUCTIONLIVE
CHIEFactiveorchestrating
SCANNERscanningmarket open
ANALYSTreadyregime · neutral
RESEARCHERidle
TRADERready
ENGINEERmonitoringall green
BUILDERidleno active build
EVALUATORisolatedaudit @ 16:50 ET
ENGINESWOLF · SLIPSTREAM · TANGO · PHOENIX · RAVEN · CONDOR · DUCK · DOLPHIN · OWL · TERMINATOR · ELLIOTT
GATECLAUDE + GROK + GEMINI + GPT
REGIMENEUTRAL · 0.78 conf
ACCOUNTLIVE · IBKR / ToS

Live status panel · production fleet roster

THE THESIS

Most AI trading is a black box. Phalanx publishes the ledger. Every signal is logged at generation. Every outcome is scored against SPY. The grades feed back into the agents. The ledger is public.

SPECIFICATION

Four things no competitor has all four of.

The "multi-agent LLM trading" narrative is now crowded. What still isn't crowded: the discipline below.

I.

GOVERNANCE

The evaluator runs on isolated infrastructure.

An independent evaluator audits every decision the trading agents make. Separate gateway, separate model, separate audit trail. No execution authority. Reports only to the operator. This is governance — not monitoring.

II.

MEASUREMENT

Every signal scored T+1 against SPY.

An append-only ledger logs every signal at generation. At expiry, each is graded against its directional outcome. Realized alpha versus SPY is computed nightly and fed back into the agent prompts — the system learns from its own measured outcomes.

III.

CONSENSUS

Four models must agree before a trade fires.

Claude, Grok, Gemini, and GPT each reason independently at temperature 0.2. Any single-model disagreement vetoes the trade. Multiple models aren't redundancy — they're cognitive diversity. One model's blind spot is another's signal.

IV.

ARCHITECTURE

Engines execute. Agents think. Signals connect them.

Three layers plus regime governance. Engines are deterministic Python — they execute. Agents are LLMs — they analyze and produce intel. Signals are the shared language between them. Engines trade, agents think, signals carry conviction. They never trade based on free-text — only structured signals with scores, timestamps, and outcomes that get tracked back. Regime governance keeps it all in line.

THE FLEET / XV

Fifteen engines running. From intraday tactical to long-term accumulation.

I. RUNNING

WOLF

Mean Reversion

Fades overextended moves

II. RUNNING

SLIPSTREAM

Momentum / Trend-following

Rides trending stocks

III. RUNNING

TANGO

Pairs Trading

Statistical divergence

IV. RUNNING

PHOENIX

Options Income

Theta strategies

V. RUNNING

RAVEN

Catalyst-Driven Shorts

Event-driven directional

VI. RUNNING

CONDOR

Long-term Accumulation

Daily cadence · patient sizing

VII. RUNNING

DUCK

Defensive Rotation

Risk-off positioning

VIII. RUNNING

DOLPHIN

Offensive Amplification

Risk-on positioning

IX. RUNNING

OWL

Ensemble Consensus

3-model vote: Claude + Gemini + GPT. Routes when all three agree.

X. RUNNING

TERMINATOR

Intraday Momentum

Momentum-continuation daemon. Intraday only.

XI. RUNNING

ELLIOTT

Elliott-Wave Swing

Wave analysis + AI judgment layer. Multi-day holds.

All engines read the strategic regime file · STAND DOWN · BEAR ATTACK · BEAR · NEUTRAL · BULL · STAMPEDE

CLOSE

Coming soon,
worth the wait.

15 engines · 8 agentsindependent evaluator, isolated infra
Phalanx — Autonomous AI Trading