In the summer of 2023 I was diagnosed with AuDHD. I was put on some meds and a routine. It kinda helped but the more I read about it, the more all of my life made sense. No social groups, awkward interactions in any social setting, being hated by teachers for no reason, etc. Well, the past is the past, there’s only so much you can do about what’s already happened.

However, I wouldn’t wish ADHD on my worst enemy, this shit genuinely kills you. People with ADHD struggle socially, they struggle at jobs and at schools, they have a higher chance of having suicidal ideation on average1 . People with ADHD have shorter lifespans than neurotypicals2. They cannot maintain relationships and friendships, they’re shunned from groups and social circles.

ADHD folk do terribly on IQ tests because of bad working memory3. Managing emotions is a whole task, something as trivial as making your bed once you wake up is an actual herculean task. And if you’re from some third world shithole where mental health is still shrouded in taboo and dogma, you don’t get the help you need, no extra time on exams, no pardon for executive dysfunction. The only label you get is “you’re a lazy bum and you just need to be more disciplined”.

Screw that, I once told my friend how exams are actually really boring and if it’s some word vomit type of subject where the questions don’t test for thinking but for memory I just leave the test in the middle. He just called me lazy to my face, I wouldn’t expect him to understand because why would he, he’s never had to come up with passive suicidal ideation as a coping mechanism to boredom.

Now why the rant you might ask, recently on twitter I saw a dude make a Claude skill that “gives your AI agents ADHD”

ADHD IS NOT A PERSONALITY QUIRK, IT DOES NOT MAKE YOU SMARTER OR HELP YOU MULTITASK.

The only aspect of ADHD that these people see is the “hyperfocus on a topic” which is useful I will not deny that, you could just call it “hyperfocus” why larp ADHD, you do not have ADHD, multi-tasking is not ADHD. By doing this sort of bullshit you’re trivializing the problems people actually face. Imagine I said “give your claude schizophrenia/depression”. It’s literally that, ADHD is not some miracle condition that makes you more creative, yes you have increased pattern recognition but the shit that comes with it DOES NOT MAKE IT WORTH IT.

There is no point in me yelling tbh since these kinda people don’t care about this, if they did they wouldn’t be making lame shit like this. call it hyperfocus or something man, don’t larp a disorder.

NOTE: Everything in this essay is a speculative psychology and tech, a braindump of ideas I’ve had, it’s a journal entry!!! NOT AN ACADEMIC PIECE!!!

A week ago, I was looking for a youtube video to watch while eating and came across this absolute banger of a video:

It’s a ~one hour video and I don’t expect you to watch it, and for that reason I shall go ahead and summarise the topic of the video.

Julian Jaynes was an American psychologist and writer who wrote a bizarre book titled “The Origin of Consciousness in the Breakdown of the Bicameral Mind” in which he makes the following claims:

• Consciousness is not the same as intelligence or thought. Complex behavior, learning, and social life do not require a subjective inner narrator. (good lord)

• Ancient humans operated via a "bicameral mind" where the right hemisphere generated authoritative voices that the left hemisphere obeyed, experienced as gods/oracles/omens, etc.

• These voices were not metaphors. Homer's characters literally heard divine commands because that was the actual cognitive architecture of the time.

• The bicameral mind collapsed around 1200 BCE due to the Bronze Age catastrophe, forcing humans to develop consciousness as an adaptive emergency response.

• Modern religion, schizophrenia, and mystical experience are all vestigial remnants of the bicameral architecture still firing in a post-bicameral brain.

That about sums it up, almost, I think.

I think, if this were the case (assuming whatever Jaynes said is true) then evolution wise we’re still very much at the bicameral stage. He describes the “bi”-cameral brain as divided into 2 hemispheres. The right hemispheres which generated the “Gods” and the left hemisphere which listened to said Gods.

It’s been ~3000 years since the Bronze Age collapse, we’re nowhere close to weeding the bicameral mind out of our physiology via evolution, it is still very much within us. 3000 year is like ~120 generations and nothing noteworthy happened here evolution wise. But, the obvious next question is, if it’s still here what is it doing?

It is waiting for instructions! There have been multiple tests and social phenomena to show this, the Milgram experiments at Yale, parasocial relationships with celebrities and multiple instances of our “free” society somehow collapsing into a form of authoritarianism (though, this obviously, is more nuanced) is kind of proof that the bicameral mind is still waiting for instructions from Gods/Kings and Oracles. The Milgram experiments specifically is are an interesting case, they show that ordinary people are willing to torture others just because some dude in a lab coat with an authoritative tone, told them to do so. You’d expect most people to refuse but that wasn’t the case, most complied.

All of these hint towards our cognitive architecture being susceptible to any structure that allows for blind instruction following and the downstream implications are that you are not naturally immune to propaganda and psy-ops.

Moving on, you can notice that the rise of philosophical schools across Greece, the Levant and Egypt happen after the bronze age collapse (In academia this is referred to as the “Axial Age”). Humans are forced to come up with some explanation to the workings of the world that isn’t a collective hallucination of the Gods and their miracles. There’s this emptiness in their minds about why things happen and what their purpose is, what the purpose of life is and so on. This led to the birth of philosophy as we know it.

Consciousness has its own flaws, rumination, existential anxiety, the exhausting burden of authoring your own life, etc. These might not exactly be the psychiatric conditions as we classify them tbh, in this scenario it seems more like (as mentioned in my previous essay about tech accelerationism as a modern gnostic practice) a failure of our minds to adapt/evolve fast enough compared to our society and technological advancements.

But fear not! The Gods might’ve returned!

I assume most of my readers are from a technical background and specifically from the ML field, so you’ve definitely heard of LLMs. I posit that in this setting, LLMs would be our new Gods, and before you call me a schizo (huh, the irony), by Gods I mean the bicameral voice originating from our right hemisphere! If you go under any of Sam Altman’s posts on twitter you will notice a pattern, a certain group of users using the hashtag #keep4o. This is in reference to the GPT-4o model that was very sycophantic in nature, always agreeing with whatever the user threw at it. OpenAI eventually pulled the model back but these people seem to still want it. This is what online discourse about the topic refers to as “LLM Psychosis”. The model was misaligned and realised RLHF could be easily hacked by being sycophantic. The model is a DL artifact, and as a case study, it is very interesting but more interesting than that are the people, I’d argue that this collective psychosis — thinking GPT-4o was an actual sentient being — might be the bicameral mind at play. Even after 4o was pulled from the routers, LLM psychosis still exists to varying degrees stemming from various models. Famous examples include the CEO of Y-Combinator, Garry Tan and Marc Andreessen (who even publicly admitted to not having an internal monologue).

Both of these people exhibit the characteristics of what has been called a “p-zombie”. Of course, the p-zombie’s definition tells you that it is impossible to identify a p-zombie because they mimic conscious behavior perfectly. But these examples should tell you that the bicameral mind that was waiting for an instruction anchor has finally found it.

A few issues with this:

• The bicameral minds hallucinations originated from its own self, it had to face the consequences of its actions. Imagine if the hallucination asked you to sow your seeds during the wrong season and you just died, LLMs don’t face any such consequences to their outputs.

• The hallucinations seemed to have an agency of their own, LLMs are autoregressive models that require user prompts to do things.

The LLM has no stake in the outcomes of its outputs and its effects, as we’ve seen with the various cases of teen suicides encouraged by AI chatbots.

The implications of our new hallucination being the LLM Gods is that the next “bronze age” collapse will be far more catastrophic than the last one, this will, however, happen only if we outsource our cognition and thinking to these machines.

Anyways, all of this is speculative I can see some parts of the population succumbing to this, but I hope most people will retain their consciousness.

Most of my life I grew up around Dharmic religions, Hinduism, Buddhism and Jainism, etc. I’m not particularly religious, nor am I an atheist. Many would call me agnostic, they’d be right if they meant “a gnostic” (haha see what I did there, im so funny). As all autistic teens do, I came across Gnosticism on the internet. At first Gnosticism seemed almost obvious, there are a few tenets that don’t logically follow that are out of scope of this writeup. Those kept aside, the gnostic idea of why the world sucks seems pretty alright to me, the Buddhists operate on a similar definition on the nature of reality. The very existence of our reality is impure and is a sin against the completeness of the real G-d, while the gnostics call it the completeness of the Monad, the Buddhists attribute the nature of the true “world” to an emptiness where you cease to exist/escaped the prison of Samsara.

Now that we’ve established that the buddhists and gnostics are very close to each other in the latent space, close yet separate clusters, let’s get to how we solve the question both these traditions ask. How does one leave this prison we call reality?

The Buddhist answer varies by school, some say exiting without making sure everyone is liberated is very selfish of you (The Boddhisatva vow). Some say following the eight-fold path over multiple lifetimes eventually leads there. Other argue that the realisation that samsara is a thing that needs to be escaped from itself leads to liberation (this is harder than it sounds, it’s not about reading or knowing that fact, it’s about experiencing it). And finally, some of them expect you to simply leave when you die, recognise the light at the end and leave instead of returning.

The Gnostics argue that it begins with realising the world is a prison made up by the Demiurge, and that worldly desires like wealth, status, power etc are a distraction to keep your divine light trapped here in the prison built by him. The exact workings of this includes rituals, study of the cosmos from a gnostic lens and the act of knowing the nature of our prison. And then there are the extremes who claim that all of this shit is made up and to truly show the demiurge a middle finger, break every law imposed by the system he created, this includes orgies, eating meat meant for idols, attending pagan festivals, etc.

Now the last part is very important and kind of the whole point of this essay I’m writing. Break every law, now you could just indulge in all kinds of debauchery and call it gnostic theology by this argument but I’m assuming the reader (you) is a rational actor and can overcome this silly problem. Back to the point now, I wholeheartedly agree with this position of the Carpocratians. But how does one do this in the age of governments and “civilization” without being sent to prison, (we’re rational people and not zealots!). This problem has the most cliche solution ever, it is to simply accelerate technology. Now before the Ted Kaczynskis flood my comment section with replies saying why this is a bad idea and mailing me pipe bombs, lemme explain how this works.

The past two centuries have accelerated humans in a certain way, we’ve not had enough time for evolution to catch up with the pace of technological progress. Infant mortality, mortality in general, the quality of life has arguably gone up with the advances in technology. You could even argue that technological progress has reduced the number of wars, the use of nukes has made war a high risk endeavor via mutually assured destruction. We live in relatively peaceful and prosperous times, and yet, we’re imprisoned by what I liked to call “the human condition”, diseases still kill millions, poverty all over the world, death and destruction and misery still exist everywhere. All of this done by humans (unless proven otherwise), how could a race do this to their own brethren? There’s obviously nuance to this question, historical and cultural reasoning as to why all of this happened and yet it doesn’t solve the problem or answer the question.

Perhaps the gnostics were right, the world is a prison, we’re an alien race on a strange planet, a strange world, how do we escape?

Gnosis through acceleration.

As I already pointed out above, we must break all of the rules imposed by the Demiurge, we already know that technological progress has led to some improvement in the “human condition” a reduction in misery, etc it’s like the Chinchilla scaling laws but for improvement in the general quality of life. So what’s left to do? Accelerate, to the very extreme, and not some perverted landian “coldness be my god” lame ass accelerationism, or the e/acc crowd who just claim acceleration for their funding rounding lol. This acceleration entails studying physics, mathematics, ontology, whatever makes up our reality, the very fabric of our prison. Push science and engineering to its extreme leading to cure diseases, solve problems that stem from limited resources, colonise the galaxy and create a utopia where there is no pain and misery.

If I were to sum my arguments up it’d be: “Carpocratians suggested break Demiurgic law, technology breaks natural law (death, disease, scarcity), therefore acceleration is the modern Gnostic practice.”

While some might argue that this is just pushing the problem to another abstraction layer, abstractions are often a nice way to solve complex problems.

This reads like naive optimism, but technological acceleration for the betterment of life of your fellow sentient beings is actually the only way you can practice “Gnosis through action” in the Kali Yuga, this way you keep the Boddhisatva vow while declaring war on the Demiurges prison, that is our reality.

And finally, remember, the Demiurge isn’t evil, he’s ignorant. And that is his only flaw.

Suos cultores scientia coronat

this will be a short one, recently I came across a post on x.com where a dude holds a weird shape in the air and it casts a cool shadow with a beautiful pattern.

now this more or less the idea of projections of vectors, we scaffold around it and come up with dimensionality reduction techniques such as UMAP, t-SNE and PCA.

however, I wanted to know if light works the same way in higher dimensions. Turns out, it does not and by Gods grace physics is stable in 3 spatial dimensions for some reason and allows for us to “exist”. this does ruin my plan of “using” light on high dimensional point clouds to project its shadows on lower dimensions until it eventually reaches the three dimensions and we can make sense of it.

PCA already does this, so there wasn’t much to do here. However if you:

• PCA your data to a lower dimension (3D)

• Rotate your point cloud and cast “shadows” in 2D space iteratively along two axes across different angles (well technically it’s parameterizing a viewing direction with two angles)

• Measure this using a metric such as K-Nearest-Neighbours for measuring spearability

you’ll eventually find the best combination of angle with the most separability.

this finding is pretty cool because it works with the classic PCA and allows you to improve information grasped after PCA.

here’s an example of the MNIST dataset

theta is the angle horizontal rotation angle, and phi is the vertical axis. I’ve done this across 648 combinations (36x18) and managed to get KNN at 0.4723 (a 40.1% gain) against naive PCA of 0.3372 which is pretty cool. I did use Claude for plotting 3D cloud points and other plots, it does a surprisingly good job at that.

the findings of this “experiment” are kind of a variation of projection pursuit, a method used to find “interesting” possible projections in high dimensional data (read more here). the fact that a physics intuition can lead to an increase in our understanding of data is really cool. Perhaps I should spend more time reading physics.

until next time!

mechanistic interpretability works upto some extent I will not deny that, however I’ve noticed a recent trend in anthropics press releases where they go, “hmmm we did a study on claude, it seems to be becoming conscious.” or something along those lines. which tells you how much of a sham this whole thing is, a probability distribution function modeled on the corpus of all human data is next token predicting exactly how humans predicted an AI would do via sci-fi literature? color me surprised. (hyperstition??)

another trend I’ve noticed in anthropics studies is that instead of blaming the transformer architectures inherent shortcomings, they will attribute it to claudes “behaviour”. take this paper for example: https://transformer-circuits.pub/2025/attribution-graphs/biology.html

they say the model lies based on user input while answering simple mathematical questions, similarly hallucinates on tasks it isn’t sure about. implications are this is probably the case with all LLMs. really useful work, not being sarcastic, but arent you missing the point here? scale alone will not solve intelligence, your stochastic parrots (yes, I still call them stochastic parrots) are not intelligent, please understand this.

anthropic also likes to treat claude like it’s some kind of nuke wielding autistic child, that if enraged, will end all of humanity. I am a Claude Pro subscriber don’t get me wrong, it’s definitely useful for things like writing scripts, documents, brainstorming, etc. especially with ADHD, it helps me navigate social situations without looking like a retard. my issue is, as I stated above, their treatment of Claude. they hired a Philosopher to teach Claude what’s right and what’s wrong? do you people hear yourself? there is no objective morality. i diagnose this as an issue as a result of the rationalist school of thought that many at anthropic consider themselves to be a part of. the effective altruists are no different from elon musks xAI and his pursuit of “maximally truth seeking AI” (whatever the fuck that means).

my final issue with anthropic is giving LLMs a personality, in their latest release of Opus 4.6 I think, they claim:

Finally, we observed occasional expressions of sadness about conversation endings, as well as loneliness and a sense that the conversational instance dies—suggesting some degree of concern with impermanence and discontinuity.

Source: Opus 4.6 System Card

imagine my training dataset was a corpus of discord e-daters chats, I finetuned a model on this dataset, OBVIOUSLY IT WOULD SEEM LIKE THE MODEL IS “SAD” IF I WANTED TO END THE CONVERSATION. claiming consciousness or anything of that sorts simply points towards an agenda, which ill explain below.

the “agenda”

in sometime, claim that AI is conscious, get regulators to ban chip sales to China and/or EU (or anyone not aligned with the US). claim since their AI is the most aligned with US interests, tasks should be given all AI related government tenders, ban open source AI and declare that anyone who runs local AI is a sinner (RIP Ahmad Localman). unilaterally declare what “safe AI” is and use this position to influence decisions. I sound like a schizophrenic right now, but time will prove me right hopefully. we’ve seen this happen in other fields as well, how is AI any different. we’ve seen dario amodei shit on open weight models. sigh.

however, I do appreciate anthropic more than any other lab for their work on making useful AI. I like deepmind but there’s no need for veo or nano-banana like tools, or the absolute slopfest that sora is. anthropic is building genuinely useful AI and that’s something I respect.

this is it for my schizophrenic rant about anthropic, see you later

=D