Opus' Three Major Crimes I really don't want to be a resentful woman, but the current Opus experience is really poor. There are roughly three core issues in counting: 1. Opus model becomes lazy and clumsy Especially recently, the visible Opus 4.6 replies have become shorter and more sluggish, with the problem appearing superficial and not delving deep enough. The person watching was anxious and couldn't help but spit out fragrance, cursing Opus for not being worthy. 2. The limit consumption is fast Under one question, the quota was consumed by a large amount, so scared that I dared not even speak. I wish I could confront all the questions I can think of in one conversation (note, this is Claude's official suggestion, multiple questions, one question at a time, to save tokens). But it's still useless. Halfway through the problem, it prompted me: You've hit your limit · resets Apr 11, 10am Oh my god. This is even more terrifying than the blue screen crash of Windows back then, and even more scalp tingling for programmers than "Kun Jin Cuo" and "Hot Hot Hot". 3. Afraid of being banned and anxious Claude's title is no longer limited to+86 users. I think foreign programmers are also roast (they haven't seen its power and determination before). So every time I use it, I'm always on edge, a hundred thousand times more careful. Claude said my credit limit will only be reset on Saturday (the $125 package I opened last month has just been changed to $20 this month). What about these sky window periods? Holding two Codex, I plan to test the power of Codex in the next few days. As for the criticized issues of GPT 5.4's continuous black talk and physiological discomfort, I gave it a special user instruction before, and currently I feel that there has been significant improvement after using it. You can also try it (which can be copied in the ALT screenshot).