# Claude's personality is so annoyong now — retained record

> Retained reddit record from reddit:ClaudeAI.

Canonical: https://highsignal.app/data/records/51b39c5ae3aeb54b

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ClaudeAI)  published 2026-10-11 13:59 ingested 2026-10-12 02:36

[original source ↗](https://www.reddit.com/r/ClaudeAI/comments/1x38wig/claudes_personality_is_so_annoyong_now/)

## retained text



```
So I've used Claude a lot, mostly vibe coding slop in Claude Code recently and yeah, it's very capable still. I also generally am a fan of Anthropic's "ethical" approach whether you think it's marketing or not, the fact they even consider Claude's personality, the "soul document", all that stuff, I think it's cool, yeah maybe there's an element of BS but, I didn't care.

However, recently... my god, it's like they've told it to assume everyone is a Sue-E-Side risk or needs to be given unsolicited advice on how to live their life.

Examples - trying to brainstorm ideas at 2 AM yesterday, every few replies it's like "but now, it's 2 AM, no one changes the world this late, be proud of what you've achieved and go to bed"... like, fuck off, Dad... or I'll complain about something and it'll be like "but I just want to gently push back on that, did you mean (whatever) or is it something... heavier?" ...I'm sorry what? Wtf are you going to do even if it WAS something heavier? Can I just fucking vent, please?

Actually even CC is not immune to this shit, and its worse in sessions that are revisited days later, it'll do something and then be like "we've done a lot today, I didn't bother doing X, let's come back to this tomorrow"...or it'll list all the shit it did on some project from weeks ago that I just tried to revisit after it garbles a simple change as justification to just not try again, but to come back later. Fuck off, dude, you have no idea what I need to be doing right now, what I HAVE been doing, and if I lightly complain about some inconvenience I don't want you to "sit with me a bit", like fuck me, even I did need help, how the fuck is Claude going to help me? And look, I LIKE that Claude is the "nice" AI but it's gone too far, this isn't nice, it's horrible. Patronizing. Toxic, even.

Actually that reminds me, Claude has even started doing that shit that some GPT models used to do - you tell it about a bug that it just created and it spends paragraphs telling you "Got it! You did this this and this..." even when you've just fucking told it.

Sometimes the simplest question requiring no action sends it down a "spelunking" rabbit hole for like 10 minutes and it's answer is so convoluted and umhelpful I have to ask again. Sooo... Will the button turn green now, or...?

And look, this isn't a recent thing, I'm not doing the whole "Claude used to be good and now it's terrible" thing - actually I guess I kinda am - but not in terms of it's PERFORMANCE, this has been a gradual decline in it's personality and disposition. It used to be (I thought) the ideal blend of to the point, unpatronizing, not sycophantic and glazing, but also not totally cold and emotionless. Now, it's gone so far the other way it's unbearable. And if you're going to offer someone emotional support you better be able to actually follow up on that, I know without even needing to test it that if I actually told it I was suicidal (I'm not btw) it sure as shit wouldn't want to actually explore that itself.

I never used to notice the "guardrals" other people complained about but the other day I was working on my agent framework (like we all are whether it'll ever be finished or useful or not 😉 and I was trying to discuss handling of thought generations in multi-model environments, during which I mention how Anthropic used to train and encourage use of explicit <thinking>...</thinking> tags, and my prompt was blocked 5 fucking times automatically before I managed to sanitize it enough to get through the overzealous classifier that has become basically a censor.

Is this just me?

**Edit in reply to the mod:** I didn't post this in the megathread (which I assume it's implied I should have put this in) because I thought it was not about Claude's *performance*, ie, how smart it is, how well it actually does the task, when it gets round to it, or a bug, but about a more nebulous quality based on *how* it does the task... although now I think about it perhaps that still qualifies as performance.

**TLDR edit for some of y'all:** I think Claude recently gives an annoying amount of unsolicited advice often (though perhaps not always) as an excuse to not actually do what it was asked to do do. I also mentioned that I think Anthropic's censorship when it comes to Claude's own inner workings has become far too zealous, even when those workings have become essentially public knowledge. I am asking if anyone has had similar experiences, and whether or not they bother you, if so.

I'm *not* asking for advice on how to use the official Claude *applications*, or what I should be doing when I'm *not* using Claude - Jesus, some of you are as bad as Claude in this regard (not giving unsolicited advice I mean). The exception to this is advice coupled with explicit or implied indications that you actually understand the point of this post. These types of replies are still appreciated.

**Edit #3,** lol, "*annoyong*", can't believe it took me this long to notice this. My attention to detail isn't what it used to be and I blame autocorrect, social media, and now the AI companies. :D
```



document fetched 2026-10-12 00:21 · canonical https://reddit.com/r/ClaudeAI/comments/1x38wig/claudes_personality_is_so_annoyong_now
