A Token for the Oracle

AI has a Soothsayer Problem

Jul 13, 2026

Recently I listened to an episode of The Tech Report, where Natasha Bernal talked about the rise of AI layoffs and AI psychosis in company leadership. It was fascinating to listen to, and was full of quotable sections like this one:

“…you have scenarios where you’ve got people who are asking ChatGPT or any of the other chatbots for their opinion to reinforce what they’ve done is right, almost as like a pseudo-HR function.”

However, one of the questions that appeared a few times is why? What causes C-suite executives, VPs and investors to go so insane when it comes to AI and AI agents. Throughout the episode, Bernal focused a lot on the mental factors such as undiagnosed mental health issues. I’ve heard other critics like Brendan Dell mention the fact that executives are so far removed from the actual work being done that they don’t know just how many mistakes AI makes. I don’t necessarily have the answer to why, but I wanted to raise another one that caught my eye recently.

AI causes Psychosis because it’s trained too never challenge you or tell you no.

As it turns out, AI has the same problem that soothsayers, mediums, and astrologers had in the past: AI has a vested interest in keeping you happy, agreeing with you, and keeping you using the chatbot.

The Soothsayer Problem

Throughout history, even into modern times, world leaders have turned to…less than qualitative data collection to help them make decisions:

In Ancient Greece, leaders traveled to Delphi to consult The Pythia, the high priestess in the Temple of Apollo, who inhaled earths gasses (or chewed oleander…or both?) to prophesy the future.
Roman senators and leaders consulted Augurs, who used the flights of birds and other “magics” to divine the future.
Hitler and the Nazi Party created a group called The Ahnenerbe, and financed real missions to send the group all over the world to find evidence the ancient, mystical Aryan race that the Nazis believed existed.

Finally, “modern” leaders from Abraham Lincoln to Ronald Reagan consulted mysticism and astrology to make strategic planning decisions. Now, as a Christian, I am not going to bash someone’s personal beliefs on what matters or works, however, I noticed a few problems from problems for using that kind of data source.

Problem 1: Unreliable Narrator

If you look at how these soothsayers and advisors divined their answers, there is a huge problem: using hallucinogenic substances, birds, or other supernatural means are not the most reliable methods for determining solutions to problems. There is no way to double check the validity of the claims and it leads to unintended consequences:

Ceaser consulted soothsayers before he crossed the Rubicon river, which ultimately convinced him to march on Rome.
Napolean often believed a personal “Lucky Star” guided his future, leading him to ignore strategic decisions during the 1812 Russia campaign
Hitler often turned to Karl Ernst Krafft, an astrologer who convinced the Nazi party that astrology could predict victory and strategy decisions, and so he was taken in as a close advisor. Once the Allies discovered this fact, they used false astrological signs and news to throw off Germany’s campaign.

I noticed a pattern when I was looking at this, which leads to the second problem: these soothsayers are just telling people what they want to hear.

Problem 2: Excessive Agreeableness

I will use the Hitler point as a case study. Krafft told the Nazi regime that they would need to end the war by 1943 to win. However, when SS officer Rudolph Hess, a strong proponent of the use of Astrology, fled to Scotland in 1941 to sue for peace, Hitler had the entire Nazi apparatus purged of astrologers, mystics, and the occult, which caused Krafft to flee. The lesson here is simple:

For soothsayers, mediums, and other supernatural oracles, getting the information wrong, or saying the wrong thing could be dangerous.

In that world, one of the safest things to do is give relatively vague answers to protect yourself, and, if you can flatter or reinforce the consultees ego, all the better. This kind of response, like the yes-man culture, creates an echo chamber, which mean everyone in that room fall for the same pitfalls, and no one learns anything.

Now I am sure your wondering, what does this have to do with AI and AI psychosis? Well, it look like AI has the same problems as soothsayers, and we’re running into the same problems.

AI, the Modern-Day Oracles

AI, at first glance, looks like it will change the world. You can ask it anything and it will respond in plain English with an intelligent sounding answer. AI can now write your code for you and with agents, it seems like AI is an all-powerful technology that can do anything. Like the soothsayers of the past, it seems like they possess superhuman abilities, and because they’re trained to be pleasant and agreeable they seem like wisely sage you can trust.

You can’t. Because of how the original labs designed the models, plus the rising legal challenges connected to AI-induced mental health and AI providing instructions for young users who’ve expressed suicidal thoughts, plus a staggering 1/3 of organizations using AI reporting issues from AI accuracy, at least from my perspective, you have a sycophantic tech tool that’s been built to keep you using it.

The AI models excessively agree with you, largely supporting what you have to say, regardless of where the conversation goes. When you ask it a question it can’t find answer, it hallucinates an answer that sounds right but isn’t right. The “safeguards” that are often touted by AI maxis also are not ubiquitous, leaving with an uneven security landscape. The point is, AI has been designed as a soothsayer, not as an encyclopedia, and although that doesn’t make the tech worthless, it again raises the question: are these models overhyped and overvalued.

Now For Something Constructive…

I didn’t want to end with an AI BAD sentiment, especially since this recent trend of AI skepticism opens up some important and constructive topics for AI’s future. Now that less people are caught up in the AI hype, we can start to ask questions like:

Should AI strategies be top-down structures, with a small number of individuals making decisions?
Should we start standardizing AI safety audits (industry or nonprofit driven, not government driven)?
Are AI models good enough to deserve the valuations and hype we given them?

Also, and one of the most important questions that’s shown up in my mind, should governments be so heavily involved in these new technologies, and can we really trust public officials when so much is on the line?

Overall, I think this skepticism is good for AIs continued growth. As more people ask questions of the AI labs around security, safeguards, and hallucinations, it will force them to produce better products or lose market share. Despite the crazy we’ve seen over the past few years, hopefully the increase debate will lead to a healthier AI future, and I think that’s something to look forward to.

Thank you for reading the article!

If you found this article valuable, then subscribe to the newsletter on LinkedIn or Substack. I am on a mission to make economics, finance, and business easier to understand for the average person, and each subscription helps!

WealthNWisdom

Discussion about this post

Ready for more?