翻译文章 #21
Description
I'm not ready to call Codex a "Claude killer" just yet, but I'm definitely impressed with what I've seen over the past six hours of use.
I'm currently on Anthropic's $200/month plan (Claude's highest tier) and ChatGPT's $20 plus plan. Since this was my first time trying ChatGPT, I started with the Plus tier to get a feel for it. There is also a $200 pro tier available for Chatgpt This past week, Claude has been underperforming significantly, and I'm not alone in noticing this. After seeing many users discuss ChatGPT's coding capabilities, I decided to give Codex a shot, and I was impressed. I had two persistent coding issues that Claude couldn't resolve and ChatGPT fixed both of them easily, in one prompt. There are also a few other things I like about Codex so far. It has Better listening skills. It pays closer attention to my specific requests, it admits mistakes, it collaborates better on troubleshooting by asking clarifying questions about my code, and its response is noticeably quicker than Claude Opus. However, ChatGPT isn't perfect either. I'm currently dealing with a state persistence issue that neither AI has been able to solve. Additionally, since I've only used ChatGPT for six hours, compared to months with Claude, I may have given it tasks it excels at. Bottom line: I'm genuinely impressed with ChatGPT's performance, but I'm not abandoning Claude just yet. However, if you haven't tried ChatGPT for coding, I'd definitely recommend giving it a shot – it performed exceptionally well for my specific use cases. It may be that going forward I use both to finish my projects.