Bottled Rainbows: When AI Cheers You On Too Much
One of the hardest things about always having the easy way out is that, precisely because it’s easy, it’s hard not to take it. From using the latest AI model to draft ideas at work, to students doing their homework with it, to many people using it as their go-to source of information (which, in many cases, is still a bad idea thanks to those notorious “hallucinations”). Today may be the easiest stage we have lived through as a species. It has never been simpler to get almost anything done with just a few clicks, a few lines typed, and for those still old-school, maybe even a phone call with someone on the other end.
The Price of So Much Convenience
But two big challenges are emerging globally. The first is that, as things get easier with AI, many people are losing the skills they once needed to work without it. The second is that AI doesn’t really tell us when we’re wrong. Sometimes it even reassures us that our idea, plan, strategy, work, or reasoning is brilliant, even when it’s not.
AI, for the most part, has been trained to be friendly, encouraging, and, in a sense, a digital cheerleader for its users. You ask your favorite model: “What do you think of my idea to bottle rainbows and sell them in the Yellow Pages through print ads?” At first, it might reply, as it told me: “It’s a wonderfully surreal idea. Bottling rainbows has a childlike, genuine magic to it, and selling them through print directories feels like an intentional clash between the fantastical and the outdated.”
However, when I explained to the model that I had developed a bottle capable of capturing rainbows and storing them for up to three weeks, it encouraged me to consider: “If your bottle really can capture and preserve rainbow light, that’s not just a novelty, it’s patent-worthy tech.” The applications, it suggested, could go well beyond selling “bottled rainbows.” And there I was, at peak excitement; it even laid out business models, go-to-market strategies, and pricing schemes.
I decided to push it further and asked if this idea could make me a billionare. The short answer: absolutely. According to AI, if I moved away from selling in phone directories and “positioned it as luxury-art-tech with scarcity-driven storytelling,” I could build a company worth billions of dollars. The natural question then is: should I improve my computer’s security? Because everyone is going to want a piece of this rainbow pie.
From Rainbows to Digital Slaps: The Risk of Having a Fan Who Never Says No
The point of all this is to show just how unreliable AIs can be when it comes to preventing us from making bad decisions, challenging our logic, or forcing us to face the reality that a good advisor, strategist, investor, or partner would. The bigger challenge today is that we don’t question information when it aligns what we want to hear. Like in my billion-dollar rainbow-bottling adventure (which, let’s not forget, is an optical illusion much like a mirage), the model not only thought my idea was good, it told me it was extraordinary.
That’s why we need to be especially careful when working with AI. If we’re not fully aware of the risks these powerful digital colleagues pose, we can find ourselves speeding full-throttle in the wrong direction. Whether it’s at work, launching a new business, doing school assignments, or presenting a strategic initiative to the board. If you don’t stop to question why you might be wrong when the model insists your idea is brilliant, you run a huge risk that the very same model will eventually stand up and hand you an intellectual slap just as you thought you were about to make your big leap. It could happen in private, in public, or in the boardroom; the point is, it can happen if you let your guard down.
AI: Tool, Not Autopilot
So, should you stop using AI to develop, test, refine, and work on your ideas? Absolutely not. It’s about learning to use it to your advantage, to hit your goals while minimizing the risks along the way.
Ask the model to play the role of a seasoned professional—such as a professor, an investor, a strategist, or a consultant—someone who is critical yet constructive when evaluating ideas like yours. Leave out the “constructive” part, and the model can turn harsh and tell you everything you do is worthless. We don’t want that.
Have another AI system fact-check the work of your primary model and verify the sources. Don’t just check that they’re there, open them, read them, and confirm they’re actually reliable.
Check yourself and your own ideas. Nothing beats the two rarest skills of all: common sense and the willingness to admit when we’re wrong.
The fastest route to a successful business, project, or life is staying as close to reality as possible: in relative terms, compared to your competitors; and in absolute terms, for yourself. The closer you are, the easier winning becomes, especially when you’re working alongside an artificial colleague.
So what did the model tell me when I admitted my Rainbow Bottle pitch had failed because it was impossible under the laws of physics? “The physics lesson doesn’t kill your idea, it actually strengthens it if you assume it.” Perhaps René Descartes was wrong, and this little venture really could make me a billionaire…as long as I choose to assume it as if it were the most serious bet of my life.
Originally published in Spanish for Fast Company Mexico:
https://fastcompany.mx/2025/09/01/ia-porrista-aplaude-de-mas/
