My thoughts on Theo’s take on “Claude vs Codex vs Cursor”: Theo’s take on “Claude vs Codex vs Cursor”
Commentary
- Theo’s take on “Claude vs Codex vs Cursor”
- A good observation here. I didn’t see the claude code’s wired animation until now. Yes its glitchy but I think its not that gimicky right? Is it just me that finds claude code not as crazy as theo just mentioned, to get attention. However I do feel that they definitely have made it to feel “productive”. Yes. All coding agents do feel like it. Honestly. It happened to me with codex, because I was exploring these and claude code is not for free users, I started using codex extensively in the month of Feb-March and it was really addictive.
- The codex GPT models are really good at doing work, good work. And that just made me go wild and build wired things, that I might have thrown out as a ambitious project. Amp has been the same kind of a drug dealer too, its a lotttt faster than codex and claude code, but not great accuracy wise (I am talking from a free user perspective).
- Cursor is a separate league, its just perfect for me. It has good ergonomics, no screen flicker like claude code and codex. It is just amazing to work with as a reliable agent. I have used it for I think 8 months now, and have just loved it. I hate claude code but have to deal with it just because it is smarter. And also it has the annoying 5 hour cap, the cursor feels limitless, though I am subscribed from my org’s behalf.