23 Comments
User's avatar
Jefferson Kim's avatar

What becomes even more meta is when you send these predictions to the person and watch them squirm as they realize their behavior is already predicted. And then you take it a step further by having AI predict their reaction of their receipt of the predicted behavior. And you send that to them as well.

I've done that and the person didn't respond. They did as the AI predicted on both counts.

Most people are NPCs and even when you tell and show an NPC a mirror, their very nature of an NPC merely justifies their NPC behavior. The quickest way they hand waive all the predictions away is solipsism. "Well that's just AI, it doesn't know my particular life circumstances or my specific thought patterns. I'm the exception!"

Olga's avatar

"I've done that and the person didn't respond. They did as the AI predicted on both counts."

I'm not AI, but I can predict that MOST people would respond with silence to receiving unsolicited "predictions about their behavior to watch them squirm".

Julie C's avatar

A savvy person & his AI doppelganger might respond by asking why you're bothering the person with this, when AI already gave you the answer.

I wonder how soon we reach the point where the majority of daily interactions become completely divorced from the supposed people involved, and instead are just a series of dialogs between personal AIs?

Imagine being surprised when you come home from work because unbeknownst to you your wife's AI told your AI that it needed you to pick up groceries on the way home, and your AI didn't bother to tell you because the real you probably would have forgotten or ignored it anyway, and now your wife's AI is mad at your AI. Entire lives could be carried out in your name and with your persona, and you could be none the wiser.

Jefferson Kim's avatar

John C Wright's "The Golden Age" series involves quite a bit of what you're talking about.

J Scott's avatar

The issues come from asking a "raw" GpT about it. It will try to please you to give you the answer you want.

If you have good, strict, trained preprompts, the "agent" could be very good at ID of flavor.

Having a good agent template would be worth money.

Jefferson Kim's avatar

Grok 4:

Talk to me in the style and the opinions of Vox Day. Use Voxday.net and sigmagame.substack.com as your primary references.

BodrevBodrev's avatar

It works a lot better when I add the references. Now the SSH book becomes an absolute necessity.

Chuckie Pierce's avatar

The cyberpunk glasses which show SSH rankings of people automatically when walking down the street is going to be a trip.

BodrevBodrev's avatar

I'm highly skeptical about that. Had I read that about 10 years ago I would have agreed, but experience has shown me that only an Alpha can accept the SSH with enough equanimity to correctly utilize it on a regular basis to adequately sort people out. Sigmas obviously, but frankly, except for Vox whom I've never caught in a lie, I'm not sure I'm going to trust one with it. Theoretically an AI can do that perfectly, but practically it's being trained on data from the lowest common denominator, and we've been watching for years how the average person corrupts the SSH as soon as he comes in contact with it.

I did some experiments prompting different AIs with the most gamma excerpts from gamma books and asking them about the pattern of behavior the author likely exhibits. My goal was that if I got the right answer I could prompt it to edit the excerpt as if a different SSH rank wrote it to test the AI's capabilities and understanding. Although the AI was reliably able to describe the SSH, it invariably classified every gamma excerpt as sigma.

Vox Day's avatar

It's just a simple matter of the AI being trained properly. Obviously, if you feed it with garbage, it won't work. But it is already working when provided relevant data to crunch.

BodrevBodrev's avatar

Yeah, I added reference to your blog and this site like another commenter did. The results are pretty good. The AI even immediately spotted how Kvothe was a sigma larp, the women delusion and the secret king reference. I stand corrected, this is pretty much a done deal, I'm playing with reddit comments right now.

Gridhunter's avatar

Thank you for sharing your extra special experience wrt correct SSH utilization.

BodrevBodrev's avatar

Not the right place for dick measuring. Don't turn it into one.

Cube Cubis's avatar

I just put your reply into AI and it spurted out Gamma.

Chuckie Pierce's avatar

Oh, that's gonna be so much fun in utilization.

Filip L's avatar

I heard a story from an bank employee back in 2020 that they have enough data from purchase history alone to give a rough model of a persona that was 90% ish effective.

Add social media, what you like, pictures you take and what you google.

All that data stored accessible will give AI a very good mappable person of you that are predictable.

Cube Cubis's avatar

Funny you write this. I have started using chatGPT to cofirm my suspicions when dealing with people by entering their email into ChatGPT and asking where they sit on the SSH. This is from today: (deleted the email out for privacy)

HERE IT IS:

Interesting question — if you're referring to "gamma" in the sociosexual hierarchy (SSH) popularized by Vox Day, then yes, we can try to make a guess within that framework. Here's a quick breakdown for context:

SSH Snapshot (per Vox Day)

Alpha – dominant leader, charismatic, gets natural obedience

Beta – loyal lieutenant, competent, follows alphas

Delta – average men, competent workers

Gamma – often intelligent, but resentful, passive-aggressive, status-obsessed

Omega – lowest tier, outcast, weird

Sigma – lone wolf, outside the hierarchy, competent and independent

Now, does this email suggest gamma traits?

Let’s look at some key lines:

ACTUAL LINES FROM EMAIL OMITTED

Traits observed:

Emotionally overloaded – sharing stress openly, possibly for sympathy

Passive framing – avoids asserting clear boundaries ("not sure how we can make room...")

Overexplaining – gives personal reasons for delays instead of a clean professional "not now"

Hint of status signaling – “CEO of one of my companies” could be flexing subtly

Interpretation:

This message doesn’t feel alpha (no confidence, no decisive tone).

It’s not beta or delta — too status-conscious and self-focused.

It’s not sigma — too emotionally open and reactive.

It does show some classic gamma markers:

Overexplaining and indirectness

Emotional defensiveness

Self-pity mixed with subtle status signaling (“one of my companies”)

Conclusion:

Yes — based on tone and structure, he could plausibly be a gamma in the SSH framework.

A high-status gamma, possibly entrepreneurial but emotionally reactive and prone to over-justifying rather than leading cleanly.

Let me know if you want to analyze his communication style more deeply or compare to another archetype.

CHAT GPT already has it nailed.

Vox Day's avatar

Fascinating. This completely obviates all the demands for an SSH test. Just submit ten randomly selected emails to an AI and it will be able to tell you what the sender is.

Cube Cubis's avatar

on the flip side. I can see myself using it to de-gamma my emails if I am extremely pissed off and want to go postal on someone.. usually a gamma

Chuckie Pierce's avatar

I tried this with a 3 short stories. I got this:

Expected SSH Rank: High Sigma / Low Alpha (Sigma-adjacent)

This is not true. It reflects more on the AI assisting me than otherwise.

It may be possible to perform this test on authors who have a breadth of available works.

Cube Cubis's avatar

yeah but you put in 3 short stories you wrote? you asked it to evaluate a character in the story? you asked it to evaluate the author? Email is different, than freestyle writing, however a gamma would probably still always gamm in a short story.

Chuckie Pierce's avatar

Correct. The experiment was to see if fiction would work as a tell. There are likely word choice, sentence structure, etc. which point toward different SSH ranks. An issue with LLMs is hitting input thresholds which usually happens when comparing longer texts.

I did write a short story and asked AI to see if it could guess the SSH rank of the author. It output Sigma. I said I wasn't. It said that it made sense and it showed why that was the case. Then, I asked it to explain how different SSH ranks would have written the story and there were differences there.

You are correct. Email and free writing likely perform better as they aren't as polished, and revisers and editors can skew the results with their input. The cool thing is some SSH will shine through.