AI Trading

A model that trades for you normally needs your balance to trade with. That is what makes it a product on a custodial venue at all: you fund an account, the platform runs something you cannot see, and the only evidence you ever get is a chart it draws about itself. Nothing here can be built that way. A model on GreenX has no key, and no key means no authority - whatever it computes arrives as a proposal and stops at your wallet.

That reads as a limitation until you look at what it fixes. The hard part of automated trading was never the model; it is knowing what you agreed to. A proposal you sign is a thing you can read first - the instrument, the direction, the size, the collateral it draws on. Letting it run while you are away is the actual design problem, and it has the same answer here as it does for copy trading: authority granted in advance has to be narrow, bounded by limits you set, and revocable by you alone.

One difference from following a person is worth stating plainly. A trader you copy is exposed to their own call. A model is exposed to nothing. So the case for trusting it cannot rest on alignment, and it cannot rest on a backtest we publish about ourselves either - it has to rest on a record, of orders that settled against the same book everyone else trades on and can be checked by anyone who cares to. That is what we are building toward: strategies whose history is public because it is on-chain, proposing into a wallet that is always free to refuse.