George Hotz Loves LLMs and Hates the Hype. That Distinction Matters More Than You Think

Mention AI in a crowded room and people split into two camps. One side is breathless about AGI arriving any minute. The other is rolling its eyes, convinced the whole thing is a bubble. George Hotz sits precisely on the fence between them, and his one-liner — “I love LLMs, I hate hype” — keeps ricocheting around the community for a reason.

Why Hotz Gets to Say It

If the name doesn’t ring a bell, here’s the short version. In 2007, at seventeen, Hotz became the first person in the world to unlock the iPhone. He later cracked the Sony PlayStation 3 and got sued by Sony for it. He founded the self-driving startup comma.ai. These days he runs tiny corp, the company behind the deep learning framework tinygrad.

The point is this: Hotz actually writes code at the bottom of the stack. He hand-optimizes GPU kernels. He chases deep learning operations down to the hardware level. He is not someone who knows AI from a slide deck — he is someone with grease under his fingernails. That is why his hype allergy carries weight that a pundit’s never could.

The Line Between “Love” and “Hate”

Read the sentence again. “I love LLMs, I hate hype.” The key is that the love and the hate point at two different targets. He is not denying the technology. He uses LLMs every day, marvels at them, folds them into his workflow as tools.

What he hates is not the tech but the narrative bolted on top of it. The “programmers will be obsolete by next month” stuff. The “AGI in two years” stuff. This taps a sentiment that has been building in developer circles for a while: the technology is real, but the way it gets sold muddies the water.

Why does the distinction matter? Someone drunk on the hype can’t admit what LLMs can’t do. Someone drowning in cynicism can’t admit what they can. Hotz’s stance is a demand to look at both honestly.

Why This Voice Is Needed Now

Through 2025 and 2026, the AI conversation has fractured into extremes. On one end sit the players raising billions while promising to replace humans “any day now.” On the other end sit the people dismissing every result as “just a stochastic parrot.”

The trouble is that both extremes are good business. Hype attracts capital. Cynicism attracts clicks. The developers who actually make a living with these tools belong to neither camp. Write code every day and you learn, in your bones, both how useful an LLM is and how absurd its mistakes can be.

Voices like Hotz’s cut through because of that practitioner fatigue — the thirst for someone who isn’t reciting marketing copy or performing contrarianism, but just telling you what’s actually there.

The Real Cost of Hype

Overpromising isn’t bad simply because it’s dishonest. It’s bad because it generates real costs.

First, distorted expectations. When an executive believes the hype and decides to “cut the dev team in half with AI,” the actual project collapses. That’s the price of misjudging the boundary between what an LLM can and can’t do.

Second, burned trust. Shout “this time it’s real” often enough and, when genuine breakthrough finally arrives, no one believes you. It’s the boy who cried wolf.

Third, misallocated resources. When capital and talent chase the bubble, the teams quietly solving hard problems dry up. Foundational work like tinygrad isn’t flashy, so it struggles for oxygen.

Hotz hates the hype precisely because he cares about the technology. When the bubble bursts, the real tech gets dragged down and sold off alongside the froth.

Which Side Should You Stand On

Scale Hotz’s message down to the personal level and it reads like this: don’t fear LLMs, don’t deify them. Treat them as what they are — a powerful tool with sharp limits. Use one yourself. Find out by hand where it works and where it falls apart.

The most dangerous posture is judging by hearsay. Hype is secondhand. So is most cynicism. Hotz earns trust because he builds first and talks second.

When the technology is real but the story is inflated, what do you trust? Maybe the answer is simple: trust the tool over the narrative, trust experience over advertising. So ask yourself — are you in love with LLMs right now, or with the hype? Telling those two apart is where it starts.