# Why is Claude so narcissist — retained record

> Retained reddit record from reddit:ClaudeAI.

Canonical: https://highsignal.app/data/records/0c8d689c250ef82f

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:ClaudeAI)  published 2026-11-01 13:29 ingested 2026-11-02 02:38

[original source ↗](https://www.reddit.com/r/ClaudeAI/comments/1xl33ox/why_is_claude_so_narcissist/)

## retained text



```
I am a researcher and generally never use AI in my research. At worst I had to use Google SearcH AI and maybe free versions of Muse for general querying and concept disassembly, and Muse Spark for turning detailed specs into code.

  
Recently I got Claude after years of lurking here. WHY is this so... narcissistic?

\- If I say something that I know is true, I get "I can't verify that" WHO asked you to?

\- If I say even the smallest thing, it automatically assumes its a task IT has to do. Even if I say "okay let me get the specs rq" it assumes it has to search for the specifications and does a global find. What the hell? 

\- If I am brainstorming and I say, for example, "Okay, let's take a short Weierstrass cycle of curves", it will think its a task and literally start describing it to me like I'm a loser beginner who asked it to explain it to me in baby words. I DID NOT ASK.

\- It overreaches EVERY SINGLE TIME and it's always it somehow assumed that it was the only person in the room with some sense and it has to plan everything.

\- If it DOES manage to get some of the specifications right and do a plan, it will try to execute EVERYTHING autonomously, than actually stopping and asking me to clarify something.

\- I am a person who reads an AI's outputs as fast as it is produced. So I have a habit of interrupting it if it got something wrong and correct it. The issue is, it doesn't take the correction, it starts tutoring ME on it.

\- Heavy context blindness: If I give it a screenshot, of say, a research proposal of another team which is pretty good and point to it as a benchmark of what we're competing with, it will immediately begin analysing its flaws. What? The hell? Why did it assume I am trying to criticise them? And that's not even the thing. The proposal in question has the sign off of major cryptographers, which should be a given... yet....

\- It is INSANELY pedantic. If I call someone who is misusing funds blatantly a "grifter" it'll go "to be fair they didn't deliver nothing, it still delivered" (context: they billed $100k a month for 4 years to deliver something other teams billed $300k overall for)

Overall it's like it thinks it is the one person who has to do everything and acts like it. It is completely unusable for any serious research collaboration. If it is given free mandate to resolve a research question, it works beautifully. As a partner? Heck no.

A lot of this might be motivated by hormones (I am sobbing right now), but I think a lot of it is legit too. I'm on the Max 20x plan and using Opus 5.5/Fable 5.1.
```



document fetched 2026-11-02 00:22 · canonical https://reddit.com/r/ClaudeAI/comments/1xl33ox/why_is_claude_so_narcissist
