TL;DR: A new study shows the generation most affected by AI, Gen Z, is also the one turning against it fastest. Meanwhile, companies can't get their own employees on board, and the CEOs driving the boom seem genuinely puzzled by the backlash. AI has a serious perception problem, and the people best positioned to fix it may be making it worse.

What happened: A Gallup survey released Thursday found that the share of young people who feel “excited” about AI dropped from 36% last year to 22%. Those who feel “angry” jumped from 22% to 31%, and those who feel “hopeful” dropped nine percentage points. The vibes track: Gen Z is the generation closest to the entry-level job market AI is already reshaping.

Gen Z’s AI usage is unchanged year over year, and more than half still use it daily or weekly. But the sentiment is clearly souring. Many worry AI could undermine creativity and remain unconvinced it will meaningfully improve learning or career outcomes.

Not just Gen Z thinks AI is chopped: AI skepticism runs deep across all age groups. A Quinnipiac poll from March found that over half of Americans think AI will do “more harm than good.” A 2025 Pew survey found 61% of Americans want more control over how AI is used in their lives. 80% support regulation, even if it slows development.

The call is coming from inside the house: That same tension is playing out inside businesses. Companies are investing heavily in AI, but struggling to convince their own employees to embrace it. Nearly 9 in 10 communications leaders say they’re not ready to lead an AI transformation. Meanwhile, employees are feeling anxious, angry, and left behind as their ill-equipped bosses aren’t able to reassure them.

CEOs are crashing out: Some tech CEOs are a little in their feelings that you don’t like their AI. Sam Altman recently called adoption "surprisingly slow." Jensen Huang described the public perception of AI as “extremely hurtful, frankly.” (Worth noting: These narratives are also coming from within the AI industry. An Anthropic safety researcher recently quit to write poetry, saying “the world is in peril.”) Microsoft's Satya Nadella said it plainly at Davos: Until people feel AI improving their actual lives, it risks losing “social permission” to operate at all.

Bottom line: AI is one of the only technologies in recent memory met with what one financial historian called “active hostility.” Or, as tech journalist Nilay Patel put it, “Great consumer products don’t make young people feel anger and despair the more they use them.” Which, fair. And right now, those building AI are the least likely group to assuage those fears or solve the trust gap. —AC

AI 101

This one's for those of you who are still figuring out what AI actually is. Advanced readers, scroll on. (If you're intermediate, who knows, maybe you'll pick up something new.) Over the past few weeks, a few of you wrote in asking for the true basics, so consider this your cheat sheet on AI to get you started.

1. The big three. Right now, there are three major players: OpenAI's ChatGPT, Google's Gemini, and Anthropic's Claude. They each have distinct strengths: ChatGPT is versatile and great at searching online; Gemini integrates most seamlessly with Google's ecosystem (Gmail, Docs, etc.); and Claude tends to excel at longer, more nuanced writing and analysis.
2. They’re everywhere. All three are available as apps on your phone, tablet, desktop, or through your web browser. Within each platform, your conversations sync across devices, so you can pick up a ChatGPT chat on your phone that you started on your laptop.
3. They have multiple "models" to choose from. Think of them like staff on a team: Each has different strengths, and some are better suited for certain situations than others. Unless you're doing something like a deep-dive research report, the default model is more than adequate for most situations.
4. Pay attention to your data. Once you're in, it's worth deciding whether you want "Memory" turned on—the chatbot will remember things about you across conversations and can reference them later. ChatGPT and Gemini have it on by default; Claude has it off. This can sometimes be useful if you're using it purely for personal purposes, but it can get complicated if you're also using it for work—so it's worth making a conscious call either way.
5. Each platform has usage limits, a cap on how much you can chat with it in a given time period. Hit that cap (on a free or paid plan), and you'll have to wait (or pay to upgrade).
6. There's something called a context window. That’s the maximum amount of information a single chat can hold. Once you approach that limit, the AI can start to lose track of what was said so far. This is (in part) why AI sometimes confidently says the wrong thing or makes something up with complete certainty, much like your least-favorite colleague.

Now you know the absolute basics. If you have any other AI questions, reply to this email, and we may answer them in a future edition. —SM

It’s the classic Silicon Valley origin story: One guy with a big dream harnesses the power of AI to build a $1.8 billion empire—fulfilling Sam Altman’s 2024 prophecy that AI would soon mint the first billion-dollar one-person company. Or, at least, this was the yarn a New York Times profile spun about GLP-1 telehealth startup Medvi last week, with just a few caveats scattered throughout.

Zoom in and it becomes obvious that the “$1.8 billion” figure that boomeranged around the internet isn’t a valuation, but an aspirational sales projection. And Medvi was never a one-man show. Per the profile, the plan was always to use AI to “do the branding and marketing.” Meanwhile, it let two telehealth firms handle the legwork of sourcing doctors, pharmacies, drug shipping, and compliance—basically everything except the NYT sit-down.

So what were the AI’s actual billable hours? It vibe-coded the website, mass-generated ad copy, and created deepfake images of nonexistent doctors and patients. But its true accomplishment, as Techdirt noted, was “bulls–t at scale.” Medvi has already received an FDA warning for drug misbranding and a class action lawsuit over 100,000-plus spam emails. Though it’s hard to argue with the ROI: AI manufactured just enough legitimacy to land a starry-eyed feature in the paper of record. —WK

Chaos Brewing Meter: /5