As an anthropic fan boy(check my prev. comments), this is the first opus release where I feel like the model is just not pleasant to talk to not to mention untrustworthy.

The two examples for me where I lost confidence in it is once where it started with 2 random echo commands: https://snipboard.io/tpqfP2.jpg Another where I asked it to create a new landing page and it just deleted an existing app page as a replacement.

I'm not exactly sure if this is a harness problem with the claude code updates(maybe system prompt changed) or if its just the model itself that has gotten too "safety-pilled" as I've been seeing similar opinions where devs are complaining about the fact that the model seems to distrust the user's intentions.

Either way, this is the first model release where I've downgraded to previous model since I was already pretty happy with it before. Should make it clear if its a model problem or the harness