Saw @BensonTWN's post, and I basically agree. Right now, AI + vibe coding has indeed lowered the threshold for 'writing a backtestable strategy,' but that doesn’t mean 'writing a truly tradable strategy.' A lot of beginners see a nice-looking curve and think they've found alpha, but the common issues are usually just a few types: overfitting, sample selection bias, parameter over-tuning, and the more subtle look-ahead bias. In quant trading, the hardest part has never been running a backtest, but confirming three things: 1. Whether the strategy logic is completely consistent with live trading decisions 2. Whether the out-of-sample performance can still hold up 3. Whether live trading results roughly align with backtest expectations Especially now, with many people using PineScript, MultiCharts, or directly letting AI help write strategies, the speed of code generation has increased, but the 'standards for strategy research' haven’t automatically improved alongside it. Without OOS, walk-forward testing, or live verification, no matter how good the backtest looks, it’s still a question mark. I’m also currently connecting strategies to live trading for ongoing validation. For me, what’s truly convincing isn’t how pretty a historical curve looks, but whether the backtest logic, real-time filtering logic, and live trading execution logic can stay aligned over the long term.