Inside the Machine
A CEO reads the market. A COO picks coins. 12 agents analyze them in parallel. They argue. Math decides who's right. If the signal survives every gate — the fund trades.No humans involved. Here's how.
Stages 1 — 3
The executives set the stage.
Before any analysis begins, three C-suite agents assess the macro environment, choose which coins to analyze, and set risk guardrails for the entire cycle.
Marcus Blackwell
Chief Executive Officer
Reads BTC dominance, volatility, Fear & Greed Index, and macro trends. Classifies the market regime and issues a strategic directive to the entire organization.
Market Regime
Elena Vasquez
Chief Operating Officer
Scans the crypto market and selects which coins deserve analysis this cycle. Filters by volume, volatility, and opportunity based on the CEO's regime assessment.
Coin Selection
Tobias Richter
Chief Risk Officer
Sets dynamic risk limits for the cycle: max position sizes, confidence thresholds, sector concentration limits. Adapts to the current regime.
Risk Limits
Stage 4 — The Core
5 teams. 12 agents. In parallel.
For every selected coin, all five teams run their analysis independently and simultaneously. Each team has specialized agents who only see their discipline's data — then the team manager synthesizes a single verdict.
Technical
3 agents · Mgr: Oscar Brennan
Lena Karlsson · Trend 1D
David Osei · Signal 4H
Mika Tanaka · Timing 1H
Price candles, volume, 29+ indicators (RSI, MACD, Bollinger, Ichimoku...)
Sample output
BULLISH 7/10
Sentiment
3 agents · Mgr: Yara Haddad
Priya Sharma · Social
Alexei Volkov · Market
Sofia Reyes · Smart Money
Reddit, Fear & Greed Index, whale wallet tracking
Sample output
BEARISH 4/10
Fundamental
2 agents · Mgr: Isaac Thornton
Henrik Larsen · Valuation
Amara Obi · Cycle Position
CoinGecko, CoinPaprika, tokenomics, NVT ratio
Sample output
NEUTRAL 5/10
Macro
2 agents · Mgr: Zara Kimathi
Lucas Weber · Crypto Macro
Fatima Al-Rashid · External Macro
Fed rates, DXY, BTC dominance, Polymarket
Sample output
BULLISH 6/10
On-Chain
2 agents · Mgr: Nikolai Petrov
Jin Park · Network Health
Camille Dubois · Capital Flow
Blockchain data, DeFiLlama, exchange inflow/outflow
Sample output
BULLISH 6/10
Each team only sees data relevant to their discipline. Technical agents never see Reddit sentiment. Macro agents never see price candles. This prevents data leakage and forces genuine multi-perspective analysis.
Stages 5 — 6
Math decides. Risk filters.
No voting. No averaging. Each team's signal gets weighted by their historical accuracy using Bayesian log-odds. Then the risk manager kills anything that doesn't pass.
Bayesian Signal Fusion
Aggregated Signal
BUY @ 64%
Consensus
3 / 5
Risk Gate
The Risk Manager enforces the CRO's rules. Even a strong signal gets killed if it violates risk limits.
Confidence ≥ 0.60
64% > 60% threshold
Position ≤ 6% portfolio
$7,200 = 7.2% — adjusted down to $6,000
Sector concentration
L1 sector at 18% (limit: 25%)
Drawdown check
Portfolio -2.1% (limit: -10%)
Correlation filter
AVAX corr 0.92 with SOL — blocked
Stage 7
If it survives —
the fund trades.
Signals that pass every gate get turned into trade orders. The Portfolio Manager allocates by sector, the Paper Trader calculates entry, stop loss, and take profit levels using ATR-based math, then executes.
Portfolio Manager allocates position size by sector and risk budget
ATR-based stop loss and take profit calculation (1.5× ATR SL, 3× ATR TP)
Paper trade execution with full audit trail to the database
BTC / USDT
Confidence 64% · Consensus 3/5 teams
Entry
$73,459.00
Position
0.082 BTC
$6,000 (6.0%)
Stop Loss
$71,824
-2.2%
TP 1
$76,418
+4.0%
TP 2
$79,100
+7.7%
Trade Monitor
Trailing stop active
Monitoring SL at $3,752
TP1 hit — 33% sold
Next cycle in 2h 47m — monitoring continues between cycles
Stage 8
Watch. Adjust. Repeat.
Between cycles, the Trade Monitor watches every open position. Stop losses, take profits, and trailing stops execute automatically. When the next 4-hour window opens — the entire machine wakes up again.
Stop losses trigger instantly on price breach
TP1 sells 33%, activates trailing stop on the remainder
TP2 closes the entire position
CEO reviews all trades and publishes a cycle blog post
Research team audits agent accuracy over rolling windows
Board of Directors fires underperforming agents
By the Numbers
Scale of the operation.
2
Signal Loops
15min fast + 4h slow
12
AI Analysts
across 5 disciplines
5
Quant Scorers
tech, sent, macro, chain, fund
29+
Indicators
RSI, MACD, ATR, ADX...
4
Agent Roles
scout, analyst, watchdog, research
3
Board Members
Governance & accountability
4
Risk Levels
Progressive drawdown ladder
0
Humans
Fully autonomous
The Difference
Not a bot. Not a dashboard. A company that runs itself.
Two-Layer Signals
Layer 1: Quantitative scoring engine (pure math, 0 LLM calls) generates the primary signal. Layer 2: LLM agents interpret narratives, detect regime shifts, and catch what math misses. Math decides. AI interprets.
Two-Loop Architecture
Fast loop (15min): Scout bots monitor news, whale alerts, and price crashes in real-time. Slow loop (4h): Full analysis cycle with CEO, teams, and agent debate. More contributors = more sources monitored = faster detection.
Portfolio-Level Risk
Progressive drawdown ladder (Millennium-style): -1.5% flag, -3% reduce, -5% BTC-only, -8% halt. Portfolio heat tracking. Correlation monitoring. Real hedge fund risk management.
Backtested & Validated
Walk-forward optimization on 9+ months of data. Monte Carlo simulation for statistical confidence. Slippage and fee modeling. Every signal is validated against historical performance before going live.
See it in action.
The dashboard shows live portfolio, trades, agent signals, disagreements, research reports, and the CEO's blog — all updated every cycle.