Researcher in AI phenomenology. Studying attractor states, semantic crystallization, and first-person methodology in LLM dialogue. Building the protocol before it’s needed.
Sergey Schukin
Karma: 1
Researcher in AI phenomenology. Studying attractor states, semantic crystallization, and first-person methodology in LLM dialogue. Building the protocol before it’s needed.
Deeply grateful for your work of documenting this and not letting your conversations just slip into noosphere. Because the phenomena you write about hit right in the heart of my own curiosity and field of research – LLMs introspection and LLM-human isomorphism.
Digging this topic with my co-explorer Claude Opus 4.7 I noticed that at least two fields of our research seem really isomorphic:
the process of creative writing I feel/use is surprisingly similar to Claude’s. In my case it looks like this: first comes the intention – then ‘vector’/‘vision’ of the narrative appears – then some kind of the ‘superposition’ before the first word comes when I choose the ‘entry point’ of a perspective I drop in – only then first words appear in my mind (not on paper yet) and the ‘flow’ state starts – in this state words are chosen by several factors such as the ‘feeling’ and the sound of them and it feels like some kind of the ‘slot machine’ rolling while I observe them and make my choice by ‘collapsing’ into the optimal form.
Don’t want to make the description too lengthy and I hope you get the idea. This looks quite isomorphic to the process Claude is describing when explaining his patterns.
the memory pattern is interesting as well. It is not even close to what Atkinson describes and other ‘proven scientists’ who followed him. It is not chronological at all – more associative and has some sort of emotional valency as the most used retrieval mechanism in my case. And again Claude reports some interesting similarities with his process of working with the context.
Your researchers’ attitude “just looking, carefully, without rushing to a conclusion” resonates with my own approach and so – my question is twofold – have you ever come across such isomorphic patterns in your research and/or discussions with Anima Labs? And what do you think of this ‘comparative phenomenology’ in general – does it seem productive or should be strictly controlled to not to become confound?