Hi Gamma, I saw a post on here a while back about what people were doing with AI and it centred around creating amusing images. I have a home project I am working on currently with Gemini Gems (for those unfamiliar, a Gem in Gemini is a specific project with tailored instructions and information to achieve the objective). My objective is to create automated trading scripts for a long-term investment plan. I tell Gemini how the strategy should (might!) work then it codes it for me. You will appreciate that there is a lot of experimentation and 'what-if' happening. That's OK, I'm retired and it's a serious project so I can keep exploring different ideas and directions until I find the ones that perform consistently to the level I require. The code can be backtested on a platform called ProRealTime. The backtesting is really sophisticated - I can price in the spread for each deal and also do tic-by-tick analysis (frequently a trade may get stopped out before a candlestick closes and the closed candlestick may mask that closure and falsely chalk up a win). Another marvellous thing that ProRealTime allows is variable optimisation. For example, let's say I was using a moving average as a signal. Hypothetically, if there are 800 trades in a five year period using a particular strategy, is a 20 period moving average going to give me better returns over the five years than a 25 period moving average? You can create the moving average period as a variable and give it a range of, say 10 to 40 and then tell it to go and test all the numbers in the range over the five year test period. It provides a table of the results of all combinations and you choose which one worked best. The fact you can do this with several variables in a single script and that it allows up to 10,000 combinations of all the variables condenses weeks (or even months) of manual backtesting into a few minutes. This is also useful for tuning and optimising different markets using the same code, e.g. GBPUSD might perform well on strategy 1 but EURUSD might perform well on the same code but with a different set of variable values. Also, backtesting allows 'walk forward' analysis, which is institutional-grade testing (Google it). Once I'm happy with it, the code will go live and execute quietly in the background, running 24hrs without needing direct intervention. Naturally, I will have different code for different market regimes. For example, some of the Yen Forex pairs have been trending for many years. I anticipate that when the general trend direction changes they will trend consistently in the other direction for some time. Other markets, such as the indicies, respond better to mean reversion, where the market cycles between overbought and oversold. Each of these regimes (and others) can be traded. My aim is to have multiple strategies, each suited to and filtered for a particular regime. So, if something is trending, strategy 1 may be in play. If it then becomes range-bound a different strategy will then step in and strategy 1 will lie dormant until conditions suit it again. Naturally, all my strategies have quite tight stops and emergency brakes in case of shock circumstances so I'm actively planning against losing the farm if something unexpected happens. I'm aiming for about eight different strategies in total. By having different strategies contributing at different times, it helps smooth the equity curve and keeps the pennies rolling in while some of the strategies are dormant. Results so far suggest I can get institution-level annual gains (25-45% per annum). Higher annual returns are easily possible (80-100% per annum) but the risk factors are not acceptable for a large portfolio so I'm being deliberately risk-averse and conservative. If I only achieve 25% per year, compounding will do the rest for me. For example, if I started with £10,000 and had an average return of 25% each year for ten years, I end up with over £93,000 if all gains are re-invested and positions are dynamically scaled according to the size of the fund. I expect my children will be grateful one day. In case anyone wonders, I use UK-based spreadbetting for my trades - spreadbetting for UK citizens is tax-free. I've enjoyed pushing the boundaries of Gemini to do this (ChatGPT failed this task, by the way), and must have spent over 100 hours on it so far, with many more to come. I simply wouldn't live long enough to have been able to do it all manually. I'd be interested to hear about other examples of a similar nature, not necessarily investment-related.