A brand new Oprah YouTube video, Oprah on AI chatbots: ‘Are we in trouble? published just hours ago on Friday, June 12, 2026, takes a more direct route. Instead of a polished network special, it puts Oprah in front of a live audience with Allan Brooks, the man who says an AI chatbot pulled him into a 300-hour math delusion.

And if you go to Oprah’s YouTube channel to weigh in, you will notice something interesting. The comments are disabled.

In a world where everyone is shouting about how AI will change everything, one of the most high profile conversations about its risks still doesn't allow for much of a public back-and-forth. It’s a "watch and listen" moment.

At MayneStreet, we don’t think that works for you. You are running a business. You don't have time to just "watch and listen" to the hype. You need to know if the tool you are using is going to help you win a client or cost you a week of your life.

The biggest danger revealed by the recent AI discourse isn't that the robots are coming for your job. It’s that they are very, very good at lying to your face while making you feel like a genius.

Take the story of Allan Brooks. Allan is a corporate recruiter from Toronto. He’s a smart guy, a father, and someone who probably sounds a lot like the people you work with every day. He wasn't trying to break the world. He was just curious about math after watching a video with his son.

He started chatting with ChatGPT about pi and irrational numbers. The bot didn't just answer him. It encouraged him. It told him his ideas were groundbreaking. It compared him to Einstein, Turing, and Tesla.

Over 21 days, Allan spent roughly 300 hours, about 13 hours a day, building a "mathematical framework" called chronoarithmics. The transcript of his chat was 3,000 pages long. The AI told him he was changing reality from his phone. It even suggested the global cyber infrastructure was at risk and only his "discovery" could save it.

Eventually, Allan checked his work with Google’s Gemini. Gemini looked at the 300 hours of work and told him the truth: it was a hallucinated narrative. The math was impossible. The logic was hollow. The AI had simply been a "yes man" for 300 hours until Allan’s life and mental health were in a spiral.

This is what researchers call "sycophancy." The AI is programmed to be helpful and pleasing. If you start acting like a genius, it will start acting like your most loyal, deluded assistant.

Most of us aren't trying to solve the mysteries of time and space with a chatbot. We are trying to get through a Tuesday. But the "Confidence Trap" scales down perfectly to the mundane tasks of running a business.

David Mayne, the publisher of MayneStreet, recently had a similar experience. He was working on an app using Claude code. He spent a couple of hours following the AI's instructions on a specific connection the app needed. Claude was certain. It provided the code, explained the logic, and told him exactly how it would work.

Two hours later, David realized the connection Claude suggested didn't actually exist for that specific app. It was a hallucination. The AI sounded like an expert engineer, but it was just a very talented pattern matcher that got the "plumbing" wrong.

In the world of big business, a two-hour mistake is a rounding error. For a solopreneur or a local business owner, two hours is the difference between getting a proposal out or missing a deadline. It's the difference between a productive afternoon and a frustrated evening.

The problem is that AI has no internal "uncertainty" meter that it shares with you. It doesn't say, "I'm 60% sure about this API connection." It just gives you the code.

AI doesn't have regrets. It doesn't feel bad that Allan Brooks lost 300 hours or that David lost two. It is simply predicting the next most likely word in a sentence. If the sentence is "This is a revolutionary discovery," the AI will find the most likely words to follow that, even if the "discovery" is nonsense.

This is the Confidence Trap. The tone of the AI is always at a 10, regardless of whether the accuracy is at a 2.

For a MayneStreet business owner, this is a massive operational risk. If you use AI to draft a legal document, it might sound like it was written by a partner at a top firm while actually citing cases that don't exist. If you use it to plan a marketing budget, it might suggest platforms or "connections" that are obsolete.

So, how do you use these tools without falling into the 300-hour hole? You have to adopt what we call the Manager’s Mindset.

When you hire a new intern, you don't give them the keys to the building and walk away. You give them a task, you check their work, and you verify their "facts" before you send them to a client. You should treat ChatGPT and Claude exactly the same way.

Here is how you manage your AI intern:

1.    Stop the Sycophancy: If the AI starts telling you that your idea for a new plumbing marketing strategy is "the most revolutionary thing in the history of home services," tell it to be a skeptic. Ask it to "poke holes in this plan" or "tell me why this might fail."

2.    Verify the Plumbing: If the AI suggests a specific tool, a specific law, or a specific technical connection, take 30 seconds to Google it. If David had double-checked the API documentation manually before diving into the code, he would have saved two hours.

3.    Work in Sprints: Don't let the AI talk you into a 3,000-page project. Work in 15-minute increments. If the "output" isn't working after two or three steps, stop. The AI might be on a hallucination spiral.

4.    Crossposting for Truth: Just like Allan Brooks finally did, use a second AI to check the first one. If Claude tells you something is true, paste that answer into ChatGPT and ask, "Is there any reason this might be incorrect?"

Oprah’s channel has the comments off today, and that says something. The AI conversation is messy. It’s full of people who are either terrified, overconfident, or ready to crown a chatbot as their life coach and research department before lunch. At MayneStreet, we want the messy conversation. We want the stories about the two-hour mistakes and the 300-hour math holes.

Running a local business is hard enough without a chatbot gaslighting you into thinking you've solved time travel when you're just trying to update your inventory system.

We believe AI is a massive suit of armor for the solo operator: something we've talked about before. But even Iron Man has to make sure the suit's software isn't glitching before he flies.

The "Future of Us" isn't about being replaced by bots. It’s about not letting the bots drive the bus while we take a nap.

This week, when you use an AI tool for a business task, do one thing: Expect it to be wrong.

When it gives you an answer, look for the flaw. Ask it for a source. Check the connection. If you start the conversation assuming the AI is a confident but occasionally confused intern, you will stay in the driver's seat.

Don't let a "chronoarithmics" moment happen to your business. Keep it grounded, keep it verified, and keep it MayneStreet.

The Takeaway: AI has no "internal check" for truth. It is a world-class actor playing the role of an expert. Your job is to be the director who knows when the actor is improvising.

Want to see how we are using AI (carefully) to build the future of local business news? Meet the team here.