“Jarvis” – Where does this name sound familiar to you? For most, it doesn’t. But I’m sure you’ve probably watched that movie at least twice. Yes, the movie is Iron Man – and Jarvis is the name of his artificial intelligence assistant.
When I first came into contact with ChatGPT, I felt just like in Iron Man, as if I had my own Jarvis by my side.
Over the years, countless ideas, trading strategies, and indicators have been piling up in the archives of my mind. Now, I can finally code and test them. With such vast opportunities, my hopes grew that truly impressive results would eventually come out.
For a while, I wrestled with coding through ChatGPT. In the first versions, my hopes were shaken. When working on heavy tasks with AI, the “chain” would break easily. But then ChatGPT-5, Gemini, and Claude appeared. During a short competitive phase, new versions were released one after another. As a result, the quality of research improved thanks to this competition. At times I work with ChatGPT, when the chain breaks I switch to Gemini, and for technical tasks I use Claude – and even though it’s tough, today it’s possible to reach results for medium-to-hard tasks.
But still, it’s too early. It will take at least 5 more years for publicly available AI to reach the level of Jarvis. Because no matter how strong and logical AI becomes, it will always remain a mirror of your own mind and experience.
What did I mean by that last sentence?
“ChatGPT, build me a bot that makes money in the forex market” – it will build you a bot only as good as the quality of your question. The more intellect shines through your question, the more profitable your bot will be.
So, what is the right question? There’s no such thing as a “right question” – there are only clear, detailed tasks. Tasks based on experience.
For example, if you ask AI: “Is it possible to make money in forex?” – it will give you a positive answer, the one you expect, and dive only shallowly into the subject. If you ask for an opinion about choosing a company, it will first list the most popular websites, the ones that rank companies for money.
For instance, on the Benzinga website, they asked us for $4,000 to list our signal provider in the top spot for just 3 months (as a so-called reliability measure). But they didn’t ask us anything about whether we were actually reliable or whether we provided quality service. If we had paid that money, and you asked ChatGPT, “Which is the best forex signal provider?”, it would have shown you that top spot from Benzinga right away.
The correct question is always filled with doubts. Answers that pass through 3–5 layers of filtering are closer to the truth.
The style of the question absolutely affects the reliability and depth of the answer.
For example, you want to find a reliable prop-trading company. At the very least, you should be aware of the discussions, complaints, and rumors surrounding these firms. The first question must be asked in the form of doubt.
Instead of saying:
“Give me reliable prop firms” –
you should ask:
“Why are prop-trading firms not interested in traders being profitable?” – that kind of question leads to a deeper answer.
But AI will start telling you things like: “They are licensed, they are under strict regulation” and so on. Meanwhile, you were asking specifically about forex prop-trading firms, but the AI spends 10 minutes talking about futures prop firms – and you may not even notice.
The proper question would be:
“Why do forex prop-trading firms prevent traders from making profits, or make it so difficult to pass evaluation tests? They say that even after passing, many times they find excuses not to pay out profits. Compare the conditions of each firm for me, and tell me which ones are really trader-friendly and which ones are not. Please investigate from real, reliable sources. Examine all the feedback about them in detail. Give me both your findings and recommendations.”
That is the first question. Later, based on the answers, you must deepen your questions even further. Otherwise, you’ll just end up with whatever ranking appears on the first site and accept that as “the best.”