"AI delusion" remains loosely defined, but most of us know it when we see it: exaggerated faith in the technology’s power, applicability, or nature, coupled with a refusal to acknowledge its limitations. It’s bad in the individual, as when an elderly relative emails you a hallucination-laden ChatGPT summary. It’s worse in management, when your boss insists that AI slop is genius marketing content at a bargain rate. It is terrifying in tech CEOs. It is not, thankfully, compulsory. Consider three recent examples of technology leadership in the age of AI. One would be Pythonesque were the disconnect with engineers and the public not so profound. Another risks becoming a business-school case study in how to demolish customer trust. The third offers something less fashionable: robust pragmatism. Saving the best till last, y’all. Case 1: we present for your consideration Sam Altman of OpenAI, one of the few actually at the controls of the singularity machine. It seems churlish to pluck just one luscious fruit out of that cornucopia of choices, yet we must. Sam Altman thinks it is a sin against the natural order that so much RAM is idling in machines owned by individuals, organizations, and anywhere that isn’t Big AI . It’s oh so inefficient. How convenient that the efficient allocation of scarce resources should favor the business he runs. Free market schmarket. Altman has since said his remarks were taken out of context and that he wants consumers to have as much memory as they like. The efficiency argument still reveals whose priorities come first. This is the best kind of delusion, one with an undeniable truth at its core – memory does indeed get used far more intensively in AI model makers and inference servers than in your laptop. For a business selling the future, treating privately owned hardware as an inefficient allocation of resources has a distinctly Soviet flavor, and the American right has vowed to stamp out socialist extremism. Get on with it, chaps, you’ve only got two years left . That Altman can’t see this is the best bit of the misconception. There are two potential root causes of this Marxist-Leninist-Altmanism. One is that he’s mathed the math and worked out that the only way OpenAI can hit the numbers is if it gets every last byte in the world, hot off the fab. The other is that you can’t run on-prem on an 8 GB state-mandated Chromebook, and on-prem is otherwise going to twist his melon until it falls off. Case 2: Andy Yen, CEO of Proton, the company behind Proton Mail and Proton VPN. You, me, and the household cat know ProtonMail as the secondary email that’s worth paying for. You want the best recovery backup email for Gmail, you want it as an accessible end-to-end encrypted email service. You want it because it is all about stable spotless security, all about the end users, all about being there when you absolutely need it. It is safe to recommend, and you do…