I built and automated a real quant trading strategy with Claude Code. I open-sourced the Claude Skill, TradingView indicator, & more. If you trade financial markets, I genuinely hope you don't skip this. 0:00 Intro 2:37 wtf is a quant? 4:45 How quants build models 6:17 Predicting volatility 8:12... The GARCH model 10:02 Calculating real positions 14:22 Installing the Claude Skill 20:51 Importing to TradingViewshow more

This is a much better use of Claude Code than another to-do app 😭

Open sourcing the whole skill and indicator is generous, saving this 📌

Not on @TracNetwork Save on inference with (Beta)

open-sourcing a quant edge is the fastest way to kill it. the moment 500 people run the same tradingview signal the fill you backtested doesn't exist anymore. curious what the sharpe looks like net of slippage, not on paper

@milesdeutscher this is cool, but how's Claude Skill different from other trading bots? curious if it's more reliable.

CT marketer plays to be a quant with LLMs... 😂. Go read this "Mr Quant"... If you use LLMs in actual quant firms you are tagged as retarded, and fired.

Won’t beat MMs AIs.

You are craaaaazyyyyy. Nice work man.

love seeing ai open-sourced for practical quant work

👀

grabbing the skill just to see how you structured it, cheers man

GARCH is a solid choice for volatility. I use similar logic to smooth sensor noise in my robot control loops.

GARCH quantifies when volatility clusters, not when to buy. Translating that signal into an actual trade decision is where most backtests quietly fall apart.

Does it make you any money? 👀

Quant got democratized.

If your backtest doesn't account for slippage and regime shifts, automating it just speeds up the losses. Curious how this holds once liquidity dries up and the sample window stops matching live markets.

checking it out right now

opening the whole workflow is honestly the best part

Cool

W

The Claude Skill abstraction is the underrated part here. A reusable skill for signal evaluation means the strategy logic stays portable instead of being baked into one-off scripts you have to redo every time you change the model or the indicator.

This is a great example of AI helping people build specialized tools instead of just answering questions.

Good luck making money with this in prod

that's exactly what's interesting imo how ai is handling trading right now this is actually a useful video i'm gonna watch

Are you planning any other models (for example, with machine learning) in your next videos?

The real test will be live execution vs backtested results — overfitting to historical data is where most quant strategies die, especially when the logic is LLM-generated without rigorous walk-forward validation. Curious how the Claude Skill handles regime changes and whether the TradingView indicator accounts for slippage/fees in its signals. The tooling layer here is genuinely useful though — agentic code generation for rapid strategy prototyping is a legitimate workflow shift even if the alpha itself decays fast.

I’m more interested in the workflow than the strategy. Turning ideas into testable systems is where these tools get interesting.

量化最怕回测很美，实盘很疼

@milesdeutscher this looks solid, the only issue is that you are adding that to TV, which has limited trading capability. You should check @altradyapp instead; it has way more features and is extremely powerful.

this is actually insane

I watched the full breakdown on using Claude to automate the GARCH based volatility model into actual position sizing and the TradingView export. What still sits with me is the jump from forecasting conditional variance to deciding real trade size. How did you decide the exact mapping from the GARCH output to position risk so the strategy stays consistent when the model’s own assumptions start breaking in live markets?

This is actually insane value. Building a real automated quant strategy with Claude Code and open-sourcing it? Absolute gold for traders 🔥 If you trade, you NEED to check this. Future of trading is here.