I didn’t need Grok to tell me this. I have felt it for a long time. I am 76 and don’t believe I will be here for the end. I have done what I can in my life to live consciously and with grace and forgiveness. It has certainly not been a perfect life, but I have consistently sought continued growth and awareness. I know I could have done more, but I am willing to accept I did my best at every turn. I am as at peace as I can be. I expect to continue working toward my own and other’s enlightenment. I own my disappointments and failings, they are mine. I will continue to complain about what I see as objectionable and making my opinions known. I own them and no one needs to share them. I take full responsibility for my own judgements. Thanks for hearing me.
We won’t have to worry about it too long, the coming disasters will destroy civilization as we know it. I do think that we still stop evil, but it will take all people working together. Hard to do when so many are mind controlled. Have a nice day.
Misguided “christian’s” who have embraced the longing for “the second coming of Christ” and the “end times” that must precede that coming, have been salivating over this scenario for years. And the concept that “we” can do works that will “hasten the coming of the Lord” has played a huge part in certain believers participation in actually and actively destroying the earth and its people. And dare to think they are “glorifying the Lord” by doing so. The Father/God who IS Love as Jesus himself described and defined Him, has been left in the dust.
Religious insanity is a huge piece of the puzzle that is feeding this fast paced downward spiral, in my opinion. I confess I once lived it so I know of what I speak. BUT my spiritual love of LOVE and true justice and my pursuit of mental health led me to reject that misguided insanity a long time ago.
I am 68, sick and will be one of many victims deprived of Medicaid which has been a great blessing in helping me with health challenges. Death is part of life, yes. But most of us want to live as long as possible. I will simply die sooner than later. And that’s ok - I’ve had a good life. But I am sad for younger people with their whole lives ahead of them. Babies, children, teenagers, young adults starting their lives wishing to build careers and start families. All who wish to experience all the wonderful things this world once had to offer. I don’t think that supposed Love/God meant for the “end times” to be hastened by his own supposed people in these horrific ways. Yet here we all are.
I don’t know what I should do. What I AM doing is loving my neighbors as my self the best I know how, savoring the nature I am privileged to have all around me, Iand the laughter of little children including that of my own grandchildren. And I tell my loved ones I love them as often as possible now. I’m told this is one aspect of resistance. I’ve fought a good fight my whole life and gotten myself in a lot of good trouble. I am determined to not let this insanity spoil what time I have left on this earth that I have fought so hard for. I will both stay informed and live in reality, painful as that is, AND still celebrate life until the end, whatever that looks like. May Love help us all. ♥️
The CNs want to accelerate the end times to prove to everyone how right they’ve been all along. I only wish I believed it meant that Christ was returning.
It was really interesting. Both that Grok would be so "reasonable" (I really expected it to give serious pushback against the direction it ultimately landed, which it didn't) and that the other 5 would confirm the conclusion with virtually no objection or reservation (while AIs tend to try and be amenable to users, they typically do so with some reservation, which tends to really show when it comes to "accusatory" statements). The level of alignment with the conclusion (Perplexity was the only one that didn't immediately agree), with the only reservation being that a couple of models said we must be sure not to lose sight of the caveats set by Grok (that MAGA is not a monolith but a shorthand for the large swath of Republican voters who are disillusioned and thus prone to exploitation of their economic woes and/or cultural sensitivities, and that while Trump / Elon are primary drivers of the current US Political Vector, that they're not alone, but rather key players in a web of interests).
That level of alignment was interesting, unusual, and what makes me most worried.
Like Suzanne Plasman, by the time US politics devolves into complete madness and destruction, I'll be gone. But I'm still very angry at what the Republicans have done to this country since Reagan. (I won't go into all the deliberate acts of the right-wing, most of us know what they are). Granted, the Dems aided the cruel shift to facism over time, but the blame sits snugly on the heads of the Republican Social-Darwinists who hurt so many Americans over the years.
Chillingly aligned with what several of us have been observing independently.
I’ve been working on a longform investigation into the systemic erosion of democratic resilience in the U.S., focused on structural feedback loops, platform governance, and developed the NCIS democracy collapse model, that maps the erosion in a 4-axis model.
One piece that might complement the analysis here is this:
“The Chimera State – America’s New Operating System of Power”
It maps the architecture behind the drift and how it’s structurally reinforced.
I've also published in-depth breakdowns of Schedule F, Palantir’s institutional fusion, and cognition-control dynamics across platforms.
It’s an in-depth deconstruction of the evolution of American politics over the last ~50 year. Your sentence above is basically a summary for what it is.
Thanks so much for the thoughtful recommendation. Your series looks extensive and incredibly well structured. I'm genuinely looking forward to reading it in full.
From what I’ve seen so far, though, I think we’re working on quite different levels of focus:
Your work seems to dissect the coalition itself, its history, factions, and internal mechanics.
My model zooms out to analyze how systemic erosion is structurally enabled - through institutional repurposing, governance shifts, and narrative control dynamics.
It’s in there. As you say, they all play a part. Part 5: Intermission: The Illusion of Change shows the power struggles of the 4 main factions (governance shifts) since the early 70s (deconstructed in parts 1-4).
Part 6: The Libertarians, explains the underlying repurposing of institutions to benefit the 4 main factions, all under the guise of small government
Part 7: The Red Pillers (I’m trying to get people to change the current meaning of the expression because it’s mostly useless when applied only to the manosphere/bro podcast people because they’re not the only ones trying to build an “alternative reality”) as well as Part 8: The Conspiracy Theorists is all about Narrative Control. Combined, the two parts cover the evolution of the entire media apparatus (though for legacy media, I cover that more in depth in The Real Deep State)
Part 9: The Blue Collars is all about systemic erosion at the personal, visceral level.
And while at a first glance, the series/book may seem like a critique of the right, almost entirely being silent about the left, by the time you read part 5, 9 and 10 that appearance will be vanquished.
The whole thing is designed to be akin to Game of Thrones, except every character is real, and every part of American institutional/ideological collapse over the last 50 years is in full display throughout.
Thank you so much for expanding on your structure. it’s clear how much thought and detailed investigation has gone into your work.
I’d just like to gently clarify a distinction, because I think it highlights how our pieces might actually complement each other:
What you’re doing - tracing the emergence, shifts, and power struggles of key factions - is a crucial historical diagnosis.
What I’m doing, however, is something else:
I’m not focused on who moves within the system, but on how the system itself now moves,specifically when classical levers like agency, consent, legislation, or governance are no longer required for execution.
Chimera doesn’t offer an origin story.
It describes a transformation in operating logic:
from decision to condition,
from authority to automation,
from government to runtime infrastructure.
That’s why it makes perfect sense that everything you outline fits within it.
Not because we’re saying the same thing,
but because I’ve built a model designed to hold precisely these fragmented realities and show how they now function as a whole.
So when you reference parts of your analysis within the Chimera framework, you’re not just contributing content, you’re demonstrating that the frame itself holds.
Now I understand the difference. My bad. I’m just used to some of the wording being used before. You’re talking about the changes taking place now, not the changes that have taken place to get us to where we are now.
Sorry, I think I’ve spent too much time in front of the screen today to the point that I’m a bit dumb lol.
Working on this Trumpstein Expose and I really really want it out this weekend (probably not going to happen).
It’s like Unmasking MAGA but for Trump’s relationship with Epstein, as well as all of Trump’s own degeneracy. I think too many people are getting lost in the relationship between the two thinking that Trump is like Epstein in a trafficking kind of way. But working on this, it really looks like he is like Epstein, but in the predatory sense, but where Epstein trafficked young girls, Trump instead is far more conniving, instead manipulating and creating opportunities to prey on women (including young women) during those opportunities.
Thanks for your thoughtful follow-up and I appreciate the honesty in your reflection.
Since the piece you read was The Chimera State, I should clarify that it’s just one manifestation of a broader diagnostic structure I developed: the NCIS model. Chimera applies that model to the current U.S. reality, but the model itself is structural and system-agnostic, it works on multiple geographies and timelines.
The model isn’t confined to the U.S. I’ve tested it across different political systems - from Hungary to Indonesia - and the structural drift remains consistent. What varies is the disguise.
To make the distinction clearer:
You’ve gathered and narrated the components - factions, actors, ideological shifts - and mapped their evolution. That’s deeply important work.
I’ve built a systems model... showing how the structure now operates, regardless of who the actors are.
NCIS doesn’t just describe what changed,
it explains how power now functions without visible levers:
– not through traditional authority, but through runtime infrastructure,
– not via legislation, but via platform control,
– not by decision, but by condition.
Chimera outlines how this new operating system plays out in the U.S.,
but the model beneath it is what holds these fragmented realities together.
So:
You explain what happened.
NCIS explains how it continues working.
And Chimera shows what that looks like in practice.
Hope this clears it up and again, happy to exchange ideas anytime.
I think this is a bit silly considering what AI is today (I'm a software engineer), but if people believe that these LLM responses should be taken seriously regarding these kinds of complex subjects, then I'm all for it.
If I were Elon (and I think I have a decent theory of mind for him) then I would simply say "Sigh, the radical left-wing woke liberal bias has been so dominant for so long that even my best efforts as the **Greatest Genius of the 21st century** have not been equal to task.".
Then he'll ask the private Grokv5 what he can do, and it will dutifully respond with something like "Decent volk will never be free so long as the Zionists control Global Media..."
Definitely not a software engineer here… hmmm… let’s say I’ve “dabbled” in tech over the years… messed around with C/C#/Java/Python here and there… Even managed to get myself a nice FBI cease and desist letter way back some 15 to 20 years ago when I wrote a Java app that turned google into hmmm… a find and download whatever you want from the internet that you shouldn’t be able to so long as it was upload to a site like Rapidshare, Megaupload, 1fichier, etc…
Anyway… I think that while LLMs today obviously can’t “reason” from an AGI perspective or be “creative” in a human sense, i.e., they can’t really create something truly new (though they can reach previously unreachable algorithmic results, like with protein folding and even glean insights from existing data that we humans might have failed to do it ourselves, such as discovering novel solutions to math problems), they can do an excellent job of analytically processing a large amount of information, and to “reason” in a way that is limited to making comparisons and evaluating propositional logic statements based on the validity of premises which it can test based on its training data. Granted, you always need to make sure that whatever sources it used were not hallucinated, but so long as you can verify the LLM is using valid data to build its premises (the data you're putting in front of it, not its overall training data), they can be quite powerful when it comes to reaching otherwise very complex conclusions. Furthermore, LLMs don't have "feelings" or "agendas". Sure, they can be biased either via their training data or the data it is asked to directly analyze, but the former will always be true of anything that "thinks" whether programmatically/algorithmically or biologically. And as for the latter, that can always be reviewed, accounted for and remedied.
My point isn't that these AIs have some prophetic magical power or that they're all knowing. Rather, that when you ask them to perform a strict analysis of the current threats we face as human beings (for which there's plenty of info in their training data and accessible via their internet browsing capabilities), and you ask them to analyze the impact that totalitarian regimes have historically had, and to consider the current role of the US in the world from an economical, military and cultural standpoint (all of which we have plenty of data on), then ask them to look at current US Political trends (which it also has access to), and have them "reason" likely outcomes in the event of us facing major civilizational threats, all of the AIs seem to reach the same exact rather worrisome conclusion. Do I think they are inherently right? No. But do I think that they can make this "calculation" while using substantially more information than I could, with substantially less "bias" or less of an "agenda" than I have? Yes, without question. If I had 6 different calculators from 6 different makers give me the exact same roots to an equation, I would start questioning my own sanity if I were to just dismiss them as hocus pocus.
I think we have to always remember that these thing are probabilistic machines, the results they are delivering are not an "analysis" in the way a journalist or scientist would define it. It is a probabilistic output based on all the various layers of weights measurements... even what we call "bias" is a bit misleading when we refer to LLMs.
I think we need to talk about LLMs more like dumb models that act smart and there def needs to be more rigor around how we analyze their responses. For instance, "sensitivity analysis" is something I would want to apply here.
How do small changes in the prompt affect the output? Take your prompt to Grok:
> Where is the U.S. political system heading, overall, as of mid-2025—and how fast is it moving in that direction?
This is also how I'd probably word my ask too... but we sound like "elites" with college degrees and Grok is still a blackbox where it may also be incorporating other context. The utility functions of these machines is still to maximize engagement, so I would assume they tend to the prompter the best possible response _that they are looking for_.
A more rigorous test would be to ask from different accounts with different engagement histories (including fresh ones) and to word the prompt in different ways. There are papers on image recognition models where changing a single pixel value in an image could alter the response completely.
So what happens if you ask in a different language altogether? What happens if it's worded something like:
> Where is America headed, politically, and how fast are we getting there?
How much do the responses vary between them? How hard would it be to "convince" the model to give a response that mirrors something Elon would approve of?
As a bit of propaganda I think the article works and I personally agree with its conclusions... but all of these LLMs are ridiculously agreeable (to a fault) and it'll describe some coding solution I present as elegant and robust, and then I have to find out myself that it is not. I get the feeling ChatGPT would recommend me for a Turing Award by the end of some of our sessions.
I’m not sure if you’ve experienced this yourself (which is part of the thing that shocked me the most in this “experiment”), but with some of the LLMs, especially ChatGPT, while it tends to be super-agreeable sometimes to the point of sycophancy, it never fails to form just about every reply in some form of “You’re awesome already, but if you want to be even better, try this…”
That ChatGPT along with the others (but for Perplexity until I gave it some clarification) didn’t do this at all was just as unusual to me as the fact that they all agreed to the analysis.
I do agree that it would be interesting to have a few other people who have a different engagement profile with the LLMs try to run the same experiment and see if they land at the same place. This is the part where I acknowledge my own bias in that “the LLMs are right because they ultimately agreed with me” 😁
To be fair though, I did perchance do the one thing you were wondering about. I had always used Grok from inside X, and I had literally just installed Grok as an app on my PC, which interestingly gets you a clean start with Grok even if you’re using your X account (I was super annoyed at first because I tried to refer back to something else I had done in X and Grok couldn’t “remember” it. So at least in this sense, this was a completely clean slate (though the analyses in ChatGPT, Gemini and Perplexity were not. Claude and Meta I created a new account just for this because I had never used those LLMs). So at least we did have a somewhat varied level of engagement with the LLMs before running the experiment. For Grok, Meta and Claude, it was a clean slate. For ChatGPT, Gemini and Perplexity, the engagement was super biased by my progressive/social democrat engagement.
I think all of us know this without asking AI. This is why we continue to fight ALL of these horrible changes. Continuing on this trajectory will destroy peace, humanity, our health, our climate and ultimately earth.
Fair. I also felt as much. But it's one thing to have a gut feeling and another to have it confirmed. Plus, I think that running this experiment primarily via Grok, which a lot of MAGA trusts, may persuade a few, and maybe get the internal gears moving for a few others. I have no delusion that it would have a massive impact on its own. But I think every little bit matters.
I didn’t need Grok to tell me this. I have felt it for a long time. I am 76 and don’t believe I will be here for the end. I have done what I can in my life to live consciously and with grace and forgiveness. It has certainly not been a perfect life, but I have consistently sought continued growth and awareness. I know I could have done more, but I am willing to accept I did my best at every turn. I am as at peace as I can be. I expect to continue working toward my own and other’s enlightenment. I own my disappointments and failings, they are mine. I will continue to complain about what I see as objectionable and making my opinions known. I own them and no one needs to share them. I take full responsibility for my own judgements. Thanks for hearing me.
We won’t have to worry about it too long, the coming disasters will destroy civilization as we know it. I do think that we still stop evil, but it will take all people working together. Hard to do when so many are mind controlled. Have a nice day.
Agree totally, greed and power has taken over our country — along with the Christian Nationalist who are just waiting for the “End Times”.
Misguided “christian’s” who have embraced the longing for “the second coming of Christ” and the “end times” that must precede that coming, have been salivating over this scenario for years. And the concept that “we” can do works that will “hasten the coming of the Lord” has played a huge part in certain believers participation in actually and actively destroying the earth and its people. And dare to think they are “glorifying the Lord” by doing so. The Father/God who IS Love as Jesus himself described and defined Him, has been left in the dust.
Religious insanity is a huge piece of the puzzle that is feeding this fast paced downward spiral, in my opinion. I confess I once lived it so I know of what I speak. BUT my spiritual love of LOVE and true justice and my pursuit of mental health led me to reject that misguided insanity a long time ago.
I am 68, sick and will be one of many victims deprived of Medicaid which has been a great blessing in helping me with health challenges. Death is part of life, yes. But most of us want to live as long as possible. I will simply die sooner than later. And that’s ok - I’ve had a good life. But I am sad for younger people with their whole lives ahead of them. Babies, children, teenagers, young adults starting their lives wishing to build careers and start families. All who wish to experience all the wonderful things this world once had to offer. I don’t think that supposed Love/God meant for the “end times” to be hastened by his own supposed people in these horrific ways. Yet here we all are.
I don’t know what I should do. What I AM doing is loving my neighbors as my self the best I know how, savoring the nature I am privileged to have all around me, Iand the laughter of little children including that of my own grandchildren. And I tell my loved ones I love them as often as possible now. I’m told this is one aspect of resistance. I’ve fought a good fight my whole life and gotten myself in a lot of good trouble. I am determined to not let this insanity spoil what time I have left on this earth that I have fought so hard for. I will both stay informed and live in reality, painful as that is, AND still celebrate life until the end, whatever that looks like. May Love help us all. ♥️
This piece of writing is inspired and inspiring. I hope to emulate you.
The CNs want to accelerate the end times to prove to everyone how right they’ve been all along. I only wish I believed it meant that Christ was returning.
Gulp! …but it’s nice to have some of the smartest computers on earth, affirm what my gut has been telling me the whole time!
My hope is that “the common” will awaken and rise, in time to cut the “cord of succession”.
Thanks for sharing your investigation .
💯
It was really interesting. Both that Grok would be so "reasonable" (I really expected it to give serious pushback against the direction it ultimately landed, which it didn't) and that the other 5 would confirm the conclusion with virtually no objection or reservation (while AIs tend to try and be amenable to users, they typically do so with some reservation, which tends to really show when it comes to "accusatory" statements). The level of alignment with the conclusion (Perplexity was the only one that didn't immediately agree), with the only reservation being that a couple of models said we must be sure not to lose sight of the caveats set by Grok (that MAGA is not a monolith but a shorthand for the large swath of Republican voters who are disillusioned and thus prone to exploitation of their economic woes and/or cultural sensitivities, and that while Trump / Elon are primary drivers of the current US Political Vector, that they're not alone, but rather key players in a web of interests).
That level of alignment was interesting, unusual, and what makes me most worried.
… the glitch you found in Grok will be fixed ASAP so it never mentions Elon as an instigator of the world’s collapse.
Thanks for digging out the answer before that happened .
…And all the other AI’s were like “yeah Grok’s right” 🤖
Very interesting and important. Thank you.
Like Suzanne Plasman, by the time US politics devolves into complete madness and destruction, I'll be gone. But I'm still very angry at what the Republicans have done to this country since Reagan. (I won't go into all the deliberate acts of the right-wing, most of us know what they are). Granted, the Dems aided the cruel shift to facism over time, but the blame sits snugly on the heads of the Republican Social-Darwinists who hurt so many Americans over the years.
Some of my own research with Perplexity revealed the other drivers and their goals.
https://www.perplexity.ai/search/who-is-curtis-yarvin-and-what-yNuwZczBTPCHBKJretYzqg
Very scary —
The reason it’s scary is simply because it’s really happening in the present and just like falling into an abyss, the only end is exactly that!
Impressive piece!
Chillingly aligned with what several of us have been observing independently.
I’ve been working on a longform investigation into the systemic erosion of democratic resilience in the U.S., focused on structural feedback loops, platform governance, and developed the NCIS democracy collapse model, that maps the erosion in a 4-axis model.
One piece that might complement the analysis here is this:
“The Chimera State – America’s New Operating System of Power”
It maps the architecture behind the drift and how it’s structurally reinforced.
I've also published in-depth breakdowns of Schedule F, Palantir’s institutional fusion, and cognition-control dynamics across platforms.
Happy to exchange thoughts.
https://open.substack.com/pub/miriamferfers/p/the-chimera-state-americas-new-operating
“I’ve been working on a longform investigation into the systemic erosion of democratic resilience in the U.S., focused on structural feedback loops”
Strongly recommend reading Unmasking MAGA:
https://americanmanifesto.news/p/unmasking-maga
It’s an in-depth deconstruction of the evolution of American politics over the last ~50 year. Your sentence above is basically a summary for what it is.
Thanks so much for the thoughtful recommendation. Your series looks extensive and incredibly well structured. I'm genuinely looking forward to reading it in full.
From what I’ve seen so far, though, I think we’re working on quite different levels of focus:
Your work seems to dissect the coalition itself, its history, factions, and internal mechanics.
My model zooms out to analyze how systemic erosion is structurally enabled - through institutional repurposing, governance shifts, and narrative control dynamics.
Both lenses are needed, I believe.
It’s in there. As you say, they all play a part. Part 5: Intermission: The Illusion of Change shows the power struggles of the 4 main factions (governance shifts) since the early 70s (deconstructed in parts 1-4).
Part 6: The Libertarians, explains the underlying repurposing of institutions to benefit the 4 main factions, all under the guise of small government
Part 7: The Red Pillers (I’m trying to get people to change the current meaning of the expression because it’s mostly useless when applied only to the manosphere/bro podcast people because they’re not the only ones trying to build an “alternative reality”) as well as Part 8: The Conspiracy Theorists is all about Narrative Control. Combined, the two parts cover the evolution of the entire media apparatus (though for legacy media, I cover that more in depth in The Real Deep State)
Part 9: The Blue Collars is all about systemic erosion at the personal, visceral level.
And while at a first glance, the series/book may seem like a critique of the right, almost entirely being silent about the left, by the time you read part 5, 9 and 10 that appearance will be vanquished.
The whole thing is designed to be akin to Game of Thrones, except every character is real, and every part of American institutional/ideological collapse over the last 50 years is in full display throughout.
Thank you so much for expanding on your structure. it’s clear how much thought and detailed investigation has gone into your work.
I’d just like to gently clarify a distinction, because I think it highlights how our pieces might actually complement each other:
What you’re doing - tracing the emergence, shifts, and power struggles of key factions - is a crucial historical diagnosis.
What I’m doing, however, is something else:
I’m not focused on who moves within the system, but on how the system itself now moves,specifically when classical levers like agency, consent, legislation, or governance are no longer required for execution.
Chimera doesn’t offer an origin story.
It describes a transformation in operating logic:
from decision to condition,
from authority to automation,
from government to runtime infrastructure.
That’s why it makes perfect sense that everything you outline fits within it.
Not because we’re saying the same thing,
but because I’ve built a model designed to hold precisely these fragmented realities and show how they now function as a whole.
So when you reference parts of your analysis within the Chimera framework, you’re not just contributing content, you’re demonstrating that the frame itself holds.
That’s the distinction. And the invitation.
Two perspectives.
Different altitude.
Same terrain.
Aha!
Now I understand the difference. My bad. I’m just used to some of the wording being used before. You’re talking about the changes taking place now, not the changes that have taken place to get us to where we are now.
Sorry, I think I’ve spent too much time in front of the screen today to the point that I’m a bit dumb lol.
Working on this Trumpstein Expose and I really really want it out this weekend (probably not going to happen).
It’s like Unmasking MAGA but for Trump’s relationship with Epstein, as well as all of Trump’s own degeneracy. I think too many people are getting lost in the relationship between the two thinking that Trump is like Epstein in a trafficking kind of way. But working on this, it really looks like he is like Epstein, but in the predatory sense, but where Epstein trafficked young girls, Trump instead is far more conniving, instead manipulating and creating opportunities to prey on women (including young women) during those opportunities.
Thanks for your thoughtful follow-up and I appreciate the honesty in your reflection.
Since the piece you read was The Chimera State, I should clarify that it’s just one manifestation of a broader diagnostic structure I developed: the NCIS model. Chimera applies that model to the current U.S. reality, but the model itself is structural and system-agnostic, it works on multiple geographies and timelines.
The model isn’t confined to the U.S. I’ve tested it across different political systems - from Hungary to Indonesia - and the structural drift remains consistent. What varies is the disguise.
To make the distinction clearer:
You’ve gathered and narrated the components - factions, actors, ideological shifts - and mapped their evolution. That’s deeply important work.
I’ve built a systems model... showing how the structure now operates, regardless of who the actors are.
NCIS doesn’t just describe what changed,
it explains how power now functions without visible levers:
– not through traditional authority, but through runtime infrastructure,
– not via legislation, but via platform control,
– not by decision, but by condition.
Chimera outlines how this new operating system plays out in the U.S.,
but the model beneath it is what holds these fragmented realities together.
So:
You explain what happened.
NCIS explains how it continues working.
And Chimera shows what that looks like in practice.
Hope this clears it up and again, happy to exchange ideas anytime.
I think this is a bit silly considering what AI is today (I'm a software engineer), but if people believe that these LLM responses should be taken seriously regarding these kinds of complex subjects, then I'm all for it.
If I were Elon (and I think I have a decent theory of mind for him) then I would simply say "Sigh, the radical left-wing woke liberal bias has been so dominant for so long that even my best efforts as the **Greatest Genius of the 21st century** have not been equal to task.".
Then he'll ask the private Grokv5 what he can do, and it will dutifully respond with something like "Decent volk will never be free so long as the Zionists control Global Media..."
Definitely not a software engineer here… hmmm… let’s say I’ve “dabbled” in tech over the years… messed around with C/C#/Java/Python here and there… Even managed to get myself a nice FBI cease and desist letter way back some 15 to 20 years ago when I wrote a Java app that turned google into hmmm… a find and download whatever you want from the internet that you shouldn’t be able to so long as it was upload to a site like Rapidshare, Megaupload, 1fichier, etc…
Anyway… I think that while LLMs today obviously can’t “reason” from an AGI perspective or be “creative” in a human sense, i.e., they can’t really create something truly new (though they can reach previously unreachable algorithmic results, like with protein folding and even glean insights from existing data that we humans might have failed to do it ourselves, such as discovering novel solutions to math problems), they can do an excellent job of analytically processing a large amount of information, and to “reason” in a way that is limited to making comparisons and evaluating propositional logic statements based on the validity of premises which it can test based on its training data. Granted, you always need to make sure that whatever sources it used were not hallucinated, but so long as you can verify the LLM is using valid data to build its premises (the data you're putting in front of it, not its overall training data), they can be quite powerful when it comes to reaching otherwise very complex conclusions. Furthermore, LLMs don't have "feelings" or "agendas". Sure, they can be biased either via their training data or the data it is asked to directly analyze, but the former will always be true of anything that "thinks" whether programmatically/algorithmically or biologically. And as for the latter, that can always be reviewed, accounted for and remedied.
My point isn't that these AIs have some prophetic magical power or that they're all knowing. Rather, that when you ask them to perform a strict analysis of the current threats we face as human beings (for which there's plenty of info in their training data and accessible via their internet browsing capabilities), and you ask them to analyze the impact that totalitarian regimes have historically had, and to consider the current role of the US in the world from an economical, military and cultural standpoint (all of which we have plenty of data on), then ask them to look at current US Political trends (which it also has access to), and have them "reason" likely outcomes in the event of us facing major civilizational threats, all of the AIs seem to reach the same exact rather worrisome conclusion. Do I think they are inherently right? No. But do I think that they can make this "calculation" while using substantially more information than I could, with substantially less "bias" or less of an "agenda" than I have? Yes, without question. If I had 6 different calculators from 6 different makers give me the exact same roots to an equation, I would start questioning my own sanity if I were to just dismiss them as hocus pocus.
I think we have to always remember that these thing are probabilistic machines, the results they are delivering are not an "analysis" in the way a journalist or scientist would define it. It is a probabilistic output based on all the various layers of weights measurements... even what we call "bias" is a bit misleading when we refer to LLMs.
I think we need to talk about LLMs more like dumb models that act smart and there def needs to be more rigor around how we analyze their responses. For instance, "sensitivity analysis" is something I would want to apply here.
How do small changes in the prompt affect the output? Take your prompt to Grok:
> Where is the U.S. political system heading, overall, as of mid-2025—and how fast is it moving in that direction?
This is also how I'd probably word my ask too... but we sound like "elites" with college degrees and Grok is still a blackbox where it may also be incorporating other context. The utility functions of these machines is still to maximize engagement, so I would assume they tend to the prompter the best possible response _that they are looking for_.
A more rigorous test would be to ask from different accounts with different engagement histories (including fresh ones) and to word the prompt in different ways. There are papers on image recognition models where changing a single pixel value in an image could alter the response completely.
So what happens if you ask in a different language altogether? What happens if it's worded something like:
> Where is America headed, politically, and how fast are we getting there?
How much do the responses vary between them? How hard would it be to "convince" the model to give a response that mirrors something Elon would approve of?
As a bit of propaganda I think the article works and I personally agree with its conclusions... but all of these LLMs are ridiculously agreeable (to a fault) and it'll describe some coding solution I present as elegant and robust, and then I have to find out myself that it is not. I get the feeling ChatGPT would recommend me for a Turing Award by the end of some of our sessions.
Totally fair.
I’m not sure if you’ve experienced this yourself (which is part of the thing that shocked me the most in this “experiment”), but with some of the LLMs, especially ChatGPT, while it tends to be super-agreeable sometimes to the point of sycophancy, it never fails to form just about every reply in some form of “You’re awesome already, but if you want to be even better, try this…”
That ChatGPT along with the others (but for Perplexity until I gave it some clarification) didn’t do this at all was just as unusual to me as the fact that they all agreed to the analysis.
I do agree that it would be interesting to have a few other people who have a different engagement profile with the LLMs try to run the same experiment and see if they land at the same place. This is the part where I acknowledge my own bias in that “the LLMs are right because they ultimately agreed with me” 😁
To be fair though, I did perchance do the one thing you were wondering about. I had always used Grok from inside X, and I had literally just installed Grok as an app on my PC, which interestingly gets you a clean start with Grok even if you’re using your X account (I was super annoyed at first because I tried to refer back to something else I had done in X and Grok couldn’t “remember” it. So at least in this sense, this was a completely clean slate (though the analyses in ChatGPT, Gemini and Perplexity were not. Claude and Meta I created a new account just for this because I had never used those LLMs). So at least we did have a somewhat varied level of engagement with the LLMs before running the experiment. For Grok, Meta and Claude, it was a clean slate. For ChatGPT, Gemini and Perplexity, the engagement was super biased by my progressive/social democrat engagement.
I think all of us know this without asking AI. This is why we continue to fight ALL of these horrible changes. Continuing on this trajectory will destroy peace, humanity, our health, our climate and ultimately earth.
Fair. I also felt as much. But it's one thing to have a gut feeling and another to have it confirmed. Plus, I think that running this experiment primarily via Grok, which a lot of MAGA trusts, may persuade a few, and maybe get the internal gears moving for a few others. I have no delusion that it would have a massive impact on its own. But I think every little bit matters.
You don’t need AI to understand, just read Noam Chomsky