Build an AI trading system

Hey, welcome to the course. Over eleven lessons, you'll build an AI trading system from nothing. We start with an empty server, add an AI agent you can text from your phone, and then build Bentley on top of it, one working piece at a time. Bentley is a real system. It runs every day on one small server, and everything you'll see on screen comes from it. This first lesson is the big picture: what Bentley is, the ground rules, and how the whole thing fits together, so every later lesson has a place to land.

So here's the goal. From an empty server to a working trading co-pilot. A Hostinger server, the Hermes agent, a GPT or Claude brain, and then Bentley, built one section at a time. Before any of that, four ground rules. First, this is education, not advice. Nothing in this course is a trade signal or a promise of returns, and trading crypto carries a high risk of loss. Second, paper first. Everything starts on pretend money. A system earns your trust on paper before real money goes anywhere near it. Third, Bentley is a co-pilot, not a money printer. It works with you, and how you manage each trade changes your results, so nothing is guaranteed. And fourth, build along or just watch. Every command is on the slides, so pause the video whenever you want to type along.

Here's how the course runs. Four parts, eleven lessons. This one is the opening. Lessons two to four are the server: a Hostinger box, locked down, running the Hermes agent with a GPT or Claude brain you can text. Lessons five to ten are the big part, building Bentley. Ten build steps: the data feeds, the dashboard, the AI setups, the market regime, context, the guardrails, the trader and the ledger. And lesson eleven is what we learned. The ideas that failed, what it costs to run, and a plan for your first 30 days.

So, meet Bentley. Bentley is an AI analyst that lives on one small server. It watches Bitcoin, Ethereum and XRP around the clock, and it works alongside the trader, not instead of them. It pulls prices, funding, open interest and positioning from free public feeds. It writes trade setups, each with an entry zone, a stop and targets, and every setup gets a score out of 100. It argues both sides, the long and the short, and when there's no trade, it says so. Now, here's the most important line on this slide. Hard rules decide what's allowed to fire. The AI only suggests. And finally, it answers you on Telegram, and it grades its own calls a day later.

Here's how it all fits together. Six stages, one direction, and every stage writes to disk, so the next stage can read it, and so can you. Stage one, the data feeds: prices, funding, open interest, positioning, macro and news. Stage two, scoring. Plain code, no AI, turns that raw data into a market score. Stage three, the AI brain. It reads the data, writes the setups and explains them. Stage four, the risk gates. Hard rules decide what's allowed to fire. Stage five, the trader. Paper first, then an exchange demo account. And stage six, the ledger, which grades every decision 24 hours later. On top of every stage sit the dashboard, the Telegram bot and an 8am summary. They read from all of it, so nothing happens that you can't see. And underneath, it's all one Hostinger server, running Ubuntu, the Hermes agent, a handful of Python services on timers, and Caddy for HTTPS.

That's the whole map. Keep it in your head, because every lesson from here builds one box on it. In the next lesson, we rent the server, connect to it with a key, and lock it down. I'll see you there.