Hey {{first_name|there}}.
New video's up. I had Claude Code write an opening range breakout strategy into Sierra Chart from scratch. Real C++, compiled, running on an ES and NQ chart. I didn't write a line of it.
There are seven of you subscribed right now and I want to thank you for joining at the start of this. At seven people I can actually talk to each of you, and I'd rather do that than send you a newsletter.
Now the part I actually want to say.
I think most people trying to use AI on their trading system are stuck between two things. They want the help, and they don't want to introduce bugs they can't see. And every time they go looking for someone to learn from, they find another guru promising numbers.
Here's how I handle that, and you should apply it to me too.
I don't blindly trust anyone or take their word for it. I use people as references and ideas, then I put it to the test on my own. And I don't trust anyone who hypes up numbers or P&L.
That sounds cynical. It's just what works, because the bugs are invisible. The one that got me more than once is look-ahead bias. Your code quietly knows a candle's close before that candle would have finished, so it enters on the favorable bars. The results come back incredible. None of it would have happened.
It bit me through indicators on a different timeframe. ADX runs on volume-based candles, so it isn't a closed-candle timeframe. In the backtest the bar is settled. In real life, at that moment, it wasn't. Nothing throws an error. The curve just looks great.
So take this video the same way. Reference, not gospel. Then go run it on your own machine.
I wrote the whole build up as well, if you'd rather check the specifics than take my word for them. Every input I set, the prompt I used, and the step where my Sierra signals and my Python backtest disagreed with each other: the written version.
What's in it is the real process, button by button. Me pointing Claude at the example studies Sierra ships with, the list of assumptions it hands back before it writes any code, Analysis → Build Custom Studies DLL, setting the inputs, and the study drawing its signals on the chart.
Two things I'd like back from you. Just hit reply, it comes straight to me.
1. Where are you in this, and what would actually move you forward? I want to make the next ones useful to you specifically instead of guessing at it.
2. What should I have Claude build into Sierra next? Give me the real rules, not "a good one." Small and specific beats big and vague, and there's a decent chance I just build it on camera.
Claude Code and Sierra Chart is a cheat code for accelerating your journey to building a system in the market. It is not a cheat code for finding an edge. Nothing is.
— Drew