Closed research experiment, now a public archive. Paper trading only. No real money was ever involved. Simulated capital of $10,000 USDT (US Dollar equivalent).
Losses are displayed alongside wins. Nothing inside the May 27 to July 20, 2026 window was reset or trimmed. An earlier development run, from February to May 2026, was purged when the project moved from calibration to observation and is not part of this archive.
Observations generated by AI (Claude + Gemini) between May 27 to July 20, 2026. This was exploratory research on a sample of 4 trades. Past simulated results do not indicate future performance.
FAQ — AI Crypto Trading Experiment
The Cortex experiment ran from 27 May 2026 to 20 July 2026 and is over. The full record is now a free public archive. Here is what it contains and what it showed.
What is Cortex?
Cortex was an AI research project exploring whether large language models (LLMs) could identify statistically significant patterns in cryptocurrency markets. It ran as a transparent paper trading experiment, with no real money involved. The experiment is now finished and this site is the public archive of everything it produced.
What happened to the experiment, and what were the results?
It ran for 54 days, from 27 May 2026 to 20 July 2026, then it was stopped. Over 327 cycles the agent produced 327 signals: 296 were HOLD, 31 were not, and of those 31 the risk manager rejected 24. Four trades were actually executed, and two of the four were closed by manual human intervention rather than by the system. Simulated equity went from 10,000.00 to 9,939.06 USDT, a result of -0.61%.
Did the experiment beat Bitcoin?
Not in any way that counts. Bitcoin buy-and-hold returned -13.70% over the same window, against -0.61% for the agent, but that gap is not a performance. Average capital exposure was 0.97% and a position was open only 11.9% of the time, so the agent was barely in the market at all. The right comparison is against doing nothing, which would have returned 0.00%. The agent returned -0.61%, so it did worse than inaction. On top of that, with only 4 trades there is no statistical significance whatsoever: quantitative research generally needs 200 or more observations before an edge can be claimed with any confidence.
What is paper trading?
Paper trading means all trades were simulated using fictional capital ($10,000 USDT) on live Binance prices. Slippage and trading fees were realistically modeled (0.075% per trade). No real transactions were executed, and no real money was ever at risk.
Does it cost anything? Do I need an account?
No and no. The archive is entirely free and there is nothing to sign up for. Accounts, login, the paid subscription and the private Telegram channel have all been removed. Everything that used to sit behind the paid tier, including the confidence scores and the full AI reasoning for every signal, is now visible to everyone with no delay.
I used to have an account. What happened to my data?
All user accounts were deleted when the service was shut down, together with the email addresses and profile records attached to them. Payment processing has been removed and the Telegram channel is closed. There is no longer any account page, because there is no longer anything to manage. If you want written confirmation of the deletion of your data, or to exercise any other GDPR right, write to atech.contact@proton.me. See the privacy policy for details.
Is the data still live?
The pages read the archived database directly in your browser, so what you see is the real record rather than a screenshot or a hand-edited summary. But the bot is stopped and nothing new is being written. The figures are frozen as of 20 July 2026.
Is this financial advice?
No. Cortex was a research experiment, not a financial advisory service. The observations were generated automatically by AI and are not personalized recommendations. Cortex is not registered as PSAN (AMF) or authorized as CASP under MiCA. The content is historical, was produced under market conditions that no longer apply, and was never actionable in the first place. Always consult a qualified financial advisor.
What does "full transparency" mean here?
Losses are displayed alongside wins and nothing has been trimmed to look better. Every observation is logged, including the 24 signals the risk manager rejected. The equity curve includes a BTC buy-and-hold benchmark, presented with the exposure caveat above rather than as a victory. Prompt versions are tracked. One thing worth stating plainly: an earlier development run from February to May 2026 was purged on 27 May 2026 when the project moved from calibration to observation, so the public record starts on that date. Nothing inside the 27 May to 20 July window was ever reset or removed.
Read the record yourself
Open the archive or read the full write-up on the project page.