Opus 5 keeps crashing and forgetting context mid-session, with guardrails that weren't there before. The speaker trusts GPT-5.6 ten times more and says Anthropic seems disconnected from how broken this feels to users.
John, what do you think of the latest Opus 5? I thought it was just me. Opus 5 is the first model, but the more I use it, the worse around it. It's kind of like love-hate relationship. I've been tracking those random things I didn't ask for, forget other things, and constantly have those annoying guardrails, etc. Sorry, but GPD 5.6, for the most part, really don't understand about what's going on. Maybe a little bit worse than 5 was already a bad release. Problem is you just never know when it's going to make a mistake. The substance of this model is simply just kind of significant around it. Almost every session I have, I become comfortable. Definitely. Oh, it's going to tell me at the end of the session on that. Opus 5 is actually worse than any modern Anthropic. They like took down one of our servers. Nothing comes close. I thought I could run it on there, but then it took the server down. So then I put the server back up. Okay. I could do the downtime. I would just assume that things will take— it's like, okay, no big deal. You got through this. You got the same experience. It seems to work always. Taking like an stable and using it as a main orchestrator has been terrible for me. I think like when Boris had posted here, he's like, hey, don't you guys know that Opus 5 is like really good at these concepts? And the problem is that, you know, as a user of Bard code, you using Opus 4.8, you've become accustomed to what it's good at. And then they release Opus 5, and then it breaks things here. There's a lot of sensitivity. It's better on 25 steerability. And there's no instruction manual. You've got to take your time. It makes me, as a user, feel like, okay, did they make this thing more complex? That's a really frustrating thing as a user. And I say overall, I'm leaning on Codex a lot more. Like, if I know I need to get something, I can just sort of— it's part of the incoming 5.6 model. It can delegate to Terra and Luna really effectively. I would say, like, I trust GPT-5.6 Sol 10 times more than I trust Opus 5 to get the work done. Like, they really need to consider, like, I'm using your product, I'm using Cloud Code, I'm on to Opus 4.8, what's going to be my experience with that change to Opus 5? They don't seem to have the capability of putting themselves in customers' shoes and taking that data and understanding how we're going to be affected. That's really concerning. That is definitely a trust issue that we're all going to have with Anthropic as time goes on.
