I have been noticing something in the conversations lately. Between friends, and inside some companies, the decisions almost always come wrapped in a few articles. Someone read a text, or heard the story of another company, and that is it. The path is already chosen. And most of the time it is only one side of the story.
I am talking here specifically about AI, because it is the topic that fills my feed the most right now. But the point is not whether AI is good or bad. It is not whether it is everything they say, or just smoke. It is not whether it is already close to AGI, or still far away. It is not whether the token price is subsidized today, and one day the bill arrives, and maybe your company limits you to Model X instead of Y just for cost. It is not whether the cybercriminals are already using this better than your defense, and your company is exposed without knowing. It is not whether your data already ended up in some training run without anyone asking permission, and now it is too late to control. There are countless facts, and each one of them has people shouting with conviction on both sides. That is not the subject of this thought.
As I always say, this Notes & Thoughts does not exist to teach anyone anything. It is my space to think out loud. So what comes here is a reflection for myself, before it is anything for you.
When the decision costs money
What bothers me is not the enthusiasm. It is the way it drives big decisions.
We are talking about choices that steer a company. They involve cost, a lot of investment money, people, a team, months of work. And even so the basis is usually a handful of articles. Many of them are advertising in disguise, or carry an interest that is not written in the first line. The same people who sell the tool also write the study that proves the tool works.
No technology that moves this much money deserves to be judged by one side only.
And there is a detail almost nobody mentions. When this happens in technology, the decision is rarely made by an engineer. In law, rarely by a lawyer. In a leak, rarely by a plumber. Whoever signs off is almost never the one who knows the problem best. And that is why I prefer the more boring path. Listen to the specialist, run a small POC, and prove value by solving a real problem, instead of following an article that already came with the conclusion ready. I am not saying the specialist should have the final say on everything, do not get me wrong. I am only saying it is worth hearing them before the money is committed.
What rarely happens is to stop and listen to the person who says the opposite. Not to change your mind on the spot. Just to test your own. Because if the thesis survives the pushback, it gets stronger. And if it does not, better to find out before the invoice arrives, not after.
This same enthusiasm that opens the door to a good bet is what hides the bill. There is no shortage of recent examples of companies that ran to the agents and then looked at the token cost with a startled face (see Uber). The exciting side of the article was in my feed. The side with the bill was not.
It is not only AI
And here is the part that intrigues me the most. This does not happen only with AI.
I see the same pattern in other topics. In politics, most of all. We fall in love with the subject, pick a side, and let the whole feed turn into a mirror. The algorithm learns fast what pleases us. From there on it delivers more of the same, always confirming what we already felt. It looks like information, but it is only echo.
Nothing is so bad that it cannot get better. And nothing is perfect seen from one angle only.
The problem is that listening only to the side that satisfies us is comfortable. Falling in love with the thesis is the easy part. The expensive part is to keep listening to the one who disagrees with it. It takes work. Sometimes it annoys me. But that is where the part of the decision I had not yet seen lives.
Saint Thomas Aquinas already did this as a method. In the Summa Theologiae, before answering any question, he first wrote the objections against his own position, and in the strongest version he could. Only after that did he give his answer. Seven hundred years before my feed, he already knew one thing. Your thesis only stands after it faces the best argument against it, and not the worst.
What I try to do
I have no formula, and I am not here saying I do this well. I have more gray hair than answers.
What I try, in practice, is simple. When a topic excites me too much, I get a little suspicious of myself. I look on purpose for the people who write from another point of view, based on other arguments. I follow people I disagree with, not to fight, just to not lose the other angle. And before I bang the gavel on something that involves other people’s money, I stop and ask one silly thing. Who wins if I believe this.
Just today this became clear on my screen. I read that June 2026 marked a more skeptical turn in the generative AI hype, with OpenAI leaning toward delaying its IPO to 2027. Lack of confidence, doubt about the value, who knows. On the same day, I read that OpenAI released GPT-5.6 Sol, their strongest model yet, in a preview for a few companies. Both sides, on the same screen, on the same day. If I followed only one feed, I would have seen only one of them, and I would have walked away thinking I understood the whole story.
It is not the skepticism of someone who thinks nothing is any good. It is the opposite. It is liking the topic enough to want to see all of it, and not just the piece that pleases me.
In the end, AI will keep being what it is, with or without my hype. What is in my hands is how I decide. And a good decision rarely comes from a feed that only agrees with me.
Pax et bonum.