AI That Trades Your Live Funded Account: What Has to Change After You Pass

In an evaluation you are chasing a number while avoiding a limit. On a live funded account there is no number to chase. There is a floor, and your job is to stay above it long enough for a payout cycle to arrive.

That sounds like a smaller problem and it is a harder one. Speed buys nothing once the target is gone, so every habit the evaluation trained, pressing when behind and sizing up to close a gap, now has nothing useful to do and a real account to do it to.

Software inherits the same mismatch. A robot tuned to reach a target will keep trying to reach one after the target stops existing. That is why a challenge bot is the wrong tool for a live account, an argument laid out in your challenge bot is the wrong bot.

The three stage framing behind all of this is in pass, keep, collect, and the thirty second spoken version is on YouTube at www.youtube.com/watch?v=zlEwFBy5Xe8.

On many funded account types the drawdown limit trails your highest equity instead of sitting at a fixed level. The distance between your peak and the floor stays the same size, so a good week does not build a cushion you can give back later: it raises the level you have to hold from then on. Some account types stop trailing once the floor reaches your starting balance. The three drawdown types, and which one your account uses, are in trailing drawdown explained.

For software this is arithmetic rather than philosophy. A risk layer that only knows a fixed starting balance will be wrong about where the floor is by the second profitable week. It has to track the peak, and it has to know which basis your firm measures against.

A funded account can also be lost, or a payout refused, on rules that have nothing to do with losing money: restrictions around scheduled news, minimum trading day requirements, and consistency conditions that cap how much of your total profit may come from a single session. Consistency in particular usually bites at the payout request rather than ending the account outright. The full map is in why funded traders lose the account.

The point for software is that none of these look like risk while they are happening. A robot that only models loss limits walks straight into them, which is why the configuration has to carry more than two numbers.

Three things, and they are all checkable before you buy.

It has to know where the floor is today, not where it was on day one, which means tracking peak equity when the account type trails. It has to halt on the firm's thresholds rather than on a generic default, which means carrying that firm's numbers in configuration. And the halt itself has to be deterministic.

PraxAI GUARD holds its firm rule set in fixed code and measures them against live equity, and it is not artificial intelligence. The full mechanics are in what PraxAI is. A daily loss halt stops the day. A drawdown halt stops until you restart the software deliberately. It is fixed code on purpose: on a live funded account, the last thing you want at two tenths of a percent from a hard limit is a component forming an opinion.

It is fixed code on purpose. On a live funded account, the last thing you want at two tenths of a percent from a hard limit is a component forming an opinion. What an AI layer can legitimately do once the account is funded, and what it should not touch, is in what AI can and cannot do on a funded account.

Software running a live funded account does not make the account safe. It takes the discretion out of execution, so the plan is applied the same way in week one and week six, and it adds operational risks of its own: a server that goes down, slippage on a fast print, or the wrong setfile loaded after an update.

It does not predict price and it does not prevent a losing month. And it does not decide whether you are allowed to run it: that is in your firm's written terms for your specific account type, and it is worth reading before the software is attached rather than after a warning email.

If you want to see the software being operated rather than described, there is a two minute screen recording at www.youtube.com/watch?v=8aJOsgguG08, written up in the dashboard walkthrough. Everything we publish is on the channel at www.youtube.com/@PraxAIOfficial.