SOL/USDC TRADING SYSTEM Layer A · B · C  ·  v4.0  ·  Daily 08:00 AEST
CONNECTING
UPDATED --:--:--
Forward Testing · Simulation

AI Trading
Swarm

SOL / USDC Simulation 3-Layer Pipeline Daily Signal

3-layer AI pipeline: deterministic market data (Layer A), LLM swarm reasoning (Layer B), simulated execution (Layer C).

No real trades. No real money. Educational simulation running daily at 8am AEST.

APPROVE RATE
—
TRADE LOG
— / —
Price Chart — Demo Data Illustrative · Not live market data
Price Buy signal Sell signal
System Status
Forward Test
SOL/USDC daily simulation
Architecture
3 Layers
Data → Reasoning → Execution
Trade Mode
Simulation
No real capital at risk
AI Stack
LLM Swarm
4-agent reasoning layer
Sample Equity Curve Illustrative · Not live
RSI (14) — Momentum Overbought >70 · Oversold <30
MACD (12, 26, 9) — Trend Histogram
Volume Demo data
Recent Execution Log Live from Supabase · Last 10 trades
Time
Action
Price
Size
Slippage
Status
The Journey (Feb–Mar 2026)
  • Started with a single trading bot, evolved to 3-layer AI swarm
  • Layer A: Deterministic sentinel (price, slippage, volatility, depth)
  • Layer B: LLM reasoning swarm (4 agents: regime, narrative, skeptic, kill switch)
Deployment Chaos
  • Vercel hanging for 10 minutes — patience needed
  • File extension issues (.html.html — my mistake)
  • Lesson: Understand WHY, not just WHAT
API Key Security Fail
  • Posted Kraken API keys in chat... twice
  • Had to delete and regenerate both times
  • Lesson: NEVER paste keys in chat — use env vars
Prompting Insights
  • "Give me the production-grade version" — be specific about quality
  • "Show me a mockup first" — validate before building
  • "Reduce scope, save tokens" — ask AI to optimise
  • ❌ "Make it better" — too vague, gets poor results
The Honest Truth
  • AI isn't magic — it's a skilled collaborator
  • Best features emerged through iteration
  • Total time: ~6 hours vs 40+ hours manually
⚠ Not financial advice. Trading is risky. DYOR.