Every trading product on the market has “AI” in the description now. Ours does too – and I want to explain what that word is actually doing in our case, because it’s doing much less than people assume.

In our system the model does not decide to open a trade. The strategy does that, out of indicators, the way it has for years. The model sits on top and answers a narrower question: does this agree with what everything else is saying right now. That’s it – a second opinion with a veto, not an oracle. And it isn’t one general model with a clever prompt behind it; it’s assembled from several sources, deliberately, because a single model has a single way of being wrong, and you find out which one at the worst possible moment.

Here’s how I’d read anyone’s AI claim, including ours. Ask what the model is allowed to do. If the answer is “it finds opportunities” or “it predicts the move,” that’s a prediction claim – and prediction has the worst public record in this industry. If the answer is “it filters, it confirms, it blocks,” that’s a much smaller claim, and much easier to actually deliver.

Smaller claims age better. That isn’t modesty – it’s what survives contact with a real market.

#crypto #ai #tradingbots #algotrading #buildinpublic