Historical note: this entry is reconstructed from the conversation archive. May 13, 2025 is the first preserved date for this stage; the article itself was edited and published later.
First, find an AI that understands the idea
Before the name AUR existed, there were attempts with different AI tools. The first problem was not choosing a programming language or a crypto exchange. It was finding a model that could follow an unfinished idea: build something that watches the market, analyses a lot of information and helps make better decisions.
After several attempts, Gemini became the most useful partner at that time. Not because it already knew the answer, but because the idea could be developed through a longer conversation, one step at a time.
The first preserved concrete step: crypto and Excel
On May 13, 2025, the archive contains a question about building an Excel tool to monitor the cryptocurrency market, analyse it and produce buy or sell signals. Compared with AUR today, that sounds small. But the core idea is already visible: the computer should do more than display a price. It should help explain what is happening in the market.
Excel was a natural starting point because it was familiar and useful with data. The conversation quickly reached its limits: exchange APIs, automatic updates, larger calculations and the ability to work without constant manual input.
Two minutes later, the direction changed
The next question was simple: what would be better than Excel? The answer pointed toward Python — a language that can collect exchange data, analyse it, backtest ideas on historical prices and eventually automate trading.
That conversation introduced tools and concepts that would keep returning for months: Python for trading, data-analysis libraries, crypto exchange APIs, backtesting and the possibility of building an automated AI trading bot. There was no mature architecture yet. There was only a growing realization that the idea could become much larger than a spreadsheet.
This was not yet today's AUR
The early thinking leaned heavily on technical indicators, buy and sell signals, parameter optimisation and the hope that a good combination could be found. Today's AUR is far more skeptical of that approach.
That is exactly why this stage matters. It shows the distance between asking “how do I build a crypto trading bot?” and asking the harder question: “can we find information that appears before a price move instead of simply describing what already happened?”
The first step was not building the bot. It was finding an AI with which the idea could actually be built.
What came next
The idea soon connected with learning how to work with AI, prompting and writing code without a traditional programming background. Within weeks there would be a VPS, scripts, exchange data and a much bigger ambition. Some early assumptions look naive today. That is why they are worth preserving.
The AUR Journal will reconstruct this road in order — including useful ideas, mistakes, overconfidence, failures and the moments when the project had to change direction.
AUR remains a research project. Historical entries document its development; they are not investment advice or a promise of returns.