I've heard recently that many people are considering a complete switch to AI programming assistants. But I personally plan to continue subscribing to my current main tools. The reason is actually quite simple—my quantitative arbitrage system code can currently only be stably reconstructed and maintained by specific advanced models. Although other tools can do small patches and fixes, when it comes to large-scale architecture optimization, the probability of introducing hidden bugs significantly increases. What's more concerning is that only that high-level model can reliably diagnose and fix such deep-seated issues. In scenarios where the stability of financial trading code is so critical, this difference translates into real costs and risks. So, in the short term, I will stick with my current tech stack.