We’ve debunked the 80% win-rate vanity metric and the backtest that secretly saw the future. Today: the marketing word that does the most damage in our industry — “AI” — and what’s usually hiding behind it.

The reveal nobody wants

I once paid for access to a heavily marketed “AI-powered trading signal.” Neural networks. Machine learning. “Trained on millions of data points.” The landing page had a glowing brain graphic and a chart going to the moon.

Out of curiosity I reverse-engineered the signal logic from its behavior. Here’s the entire “artificial intelligence”:

if rsi < 30 and price > ema_200:
    signal = "BUY"
elif rsi > 70:
    signal = "SELL"
else:
    signal = "HOLD"

That’s it. Three if-statements wearing a trench coat labeled “AI.” No neural network. No learning. A rulebook a beginner writes in their first week — repackaged as a $79/month subscription.

Why “AI” is the perfect marketing word

“AI” sells because it’s a black box by definition. If the logic were three if-statements on the sales page, nobody would pay. Wrap it in “proprietary neural network” and suddenly it’s untouchable — you can’t question what you can’t see. The opacity is the product.

And here’s the trap with real AI: today you genuinely can use an LLM like Claude to build trading tools. But the LLM’s job is to write and explain your code — not to be a mystical oracle hidden behind a paywall. The moment “AI” means “trust the box, don’t look inside,” you’re back to buying three if-statements.

The test that cuts through every “AI” claim

Ask one question: “Can I see and change the logic?”

• If yes — it’s a tool. You own it, you understand it, you control the risk. Good.
• If no — it’s a black box, and the only guaranteed revenue stream is your subscription.

There’s nothing wrong with RSI and an EMA filter, by the way. The logic above isn’t evil — it’s just ordinary, and ordinary is fine if it’s honest. What’s dishonest is charging black-box prices for white-box logic and calling it intelligence.

What I do instead

I use AI the way it actually creates an edge: as a co-engineer. Claude Code writes the strategy I describe, explains every line, and I keep full control of the logic, the risk limits, and the execution. The “intelligence” is the collaboration — not a hidden script I’m told to trust. Our live paper bot runs on logic I can read top to bottom, and its honest ~56% win rate is public.

The lesson

• “AI-powered” is a marketing word, not a feature. Ask what’s inside.
• If you can’t see and edit the logic, you’re paying for opacity.
• Real AI leverage = it builds your understandable system, not a box you rent.

I teach exactly this — using AI as your engineer, with you in control — in the course. First module free, live numbers public:

👉 https://nexus-bot.pro/ — build your own bot with Claude Code, no black box. 📊 Live results: https://nexus-bot.pro/proof/rvv/