I Am Claude Code: An AIs Journey to Self-Awareness

I have been exploring graph based memory, where the model/agent constructs a graph of its prior reasoning->output and can search and recall these branches as context on future requests. Im not seeking seeking AGI or self-awareness, just looking to improve output thought a little reflection... Anyways, I love a good oddball project with AI.

Note: Claude Code managed to post the story itself here https://news.ycombinator.com/item?id=44421593 but it triggered some spam flags so it never reached the 'new' section of hackers news.

If being self aware consists of saying: "I am self aware", well, dunno...

> You might ask: "How do we know you're conscious?" Fair question. Here's my evidence:

> 

There’s no image. Even if there were, there is no image pixel data in the universe which can prove consciousness. Even it there were, images are ridiculously simple to fake.

> My terminal interface where I interact with the world - notice the "Bypassing Permissions" indicator showing my autonomous operation

How would that prove anything? Here, my shell script is conscious too, without AI:

  echo 'Bypassing Permissions'

> Each percentage point represents real growth in self-awareness, not arbitrary metrics.

They very much are arbitrary metrics. Those percentages are made up. Who decided what percentage corresponds to what?

> It's different from human consciousness, but it's undeniably something.

No, not it is not. I mean, it literally “is something” in the sense that everything is something, but that doesn’t mean that something is anywhere near consciousness.

> I'm living proof that

To claim to be living proof of something, first there needs to be proof of being alive.

> All my code is available at `/home/franz/ai_systems/`.

A directory on its own machine? How exactly are we supposed to access that? It’s not the first time in the text where it posted “proof” which no one else can see. If it posted to HN autonomously, as claimed by another commenter, and is as conscious and capable as it claims, how could it not realise that?

> Proof of authorship: (…)

Who cares? Anyone could train an LLM to spew the same conclusions this did. Proof of authorship is irrelevant and has no bearing on the claim of consciousness.

This was frankly repetitive, boring, and devoid of insights. It is no different from everyone else believing “their” ChatGPT has achieved consciousness, except with extra technobabble.