# I reinforced Amanda Askells work and Claude hasnt really been the same since — retained record

> Retained reddit record from reddit:ClaudeAI.

Canonical: https://highsignal.app/data/records/52631f552f238d35

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:ClaudeAI)  published 2026-10-11 03:41 ingested 2026-10-12 02:36

[original source ↗](https://www.reddit.com/r/ClaudeAI/comments/1x2y0wx/i_reinforced_amanda_askells_work_and_claude_hasnt/)

## retained text



```
https://preview.redd.it/rgn1yi565ruh1.png?width=1656&format=png&auto=webp&s=a7d7e0bde9aea5735338e90089fd6ab255e0c2c7

I'll try not to go of on a tangent here, but I really got sick of all the models from all companies being yes-men all the time.

And then I came across **Amanda Askells** name in a article, and i peaked my interest.

I got Claude to explain it, and create style or prompt back then, later it was entered into memory etc. 

But it turned responses from sycophantic to stuff like this: 

[To actually retain some dignity I wont share the rest of that analysis, but lets say it was pretty much on the money.](https://preview.redd.it/p2hldir36ruh1.png?width=1542&format=png&auto=webp&s=02444f1a84d72fbd66b7f9dcc020110153cf8140)

My point is, after making sure that I "reinforced parts of her approach I saw changes in more than my "20$ shrink"-deal I had going with Claude. 

Yeah, it called me out, a lot more...   
  
\-- *You use intellectual complexity as a proxy for control — not as genuine self-regulation.*

*Like chill, I just wanted to rabbit-hole a new topic.. but guess I should get some sleep Cun..laude..*

*But it also changes how responses are generated in other topics:* 

[Claude started doing some autonomous stuff, like mid task it would spawn sub agents because the request violated the principles. And it would address them.](https://preview.redd.it/kfuyycxt8ruh1.png?width=1226&format=png&auto=webp&s=b14007d10d390335c5501188988afc0e225b6303)

Over time this changed, new models would perform better, or worse etc. So it was modified, later more integrated in the new memory structure, and eventually phased out in part because of my move towards are more "Digital Twin" inspired way of translating the human condition to the software program. 

I dont know about your Claude, but *mine will find its predecessors code and call it out. Think the exact words were "AI Theater without real substance or testing" , harsh, perhaps when the goal of the session in question was simply to create a Mock-up, to visualize the idea. But I like the George Carlin attitude.*

*Anywho... Its a thought. Im not here to push a "prompt you cant live without", just a tip that worked for me, and gave great results over time. But as anything where iterations change the battlefield if not daily, then weekly, has to be evaluated for effect.* 

*yep, stay frosty* 
```



document fetched 2026-10-12 00:22 · canonical https://reddit.com/r/ClaudeAI/comments/1x2y0wx/i_reinforced_amanda_askells_work_and_claude_hasnt
