THE NUMBER: 4.9 out of 10 — the average AI automation exposure score across all 342 U.S. occupations, according to Andrej Karpathy's weekend project. Jobs paying over $100K average 6.7. Jobs under $35K average 3.4. The people most worried about AI are the ones least likely to lose their jobs. The people who should be worried aren't paying attention.
Ethan Mollick spent the weekend writing what amounts to a coroner's report for the public internet. The comments on his posts — X and LinkedIn both — aren't worth reading anymore. Not because of trolls. Because of bots. "Meaning-shaped attention vampires," he called them. Sophisticated enough to fool everyone, purposeless enough to fool no one who's actually looking.
His prediction: humans retreat to private Discords and group chats. The public web becomes bots talking to bots about what's "doing the heavy lifting." That's not sci-fi. It's Monday morning.
This is a business problem, not a philosophy problem. Meta and TikTok sell ads based on engagement. If that engagement is increasingly synthetic, every DTC brand paying $50 CPMs is subsidizing a fiction. A guy figured out years ago he could upload ambient music to Spotify, point bots at it, and collect royalties forever. No listeners required. Now scale that logic to the $130 billion digital ad market. The advertiser thinks they're reaching a woman in Austin interested in running shoes. They're reaching a language model interested in nothing.
The opportunity hiding inside this mess: a "guaranteed human" layer. Continuous biometric verification that proves, in real time, a human is on the other end. Not a CAPTCHA. Persistent proof of life. Advertisers would pay a premium. Consumers might accept the trade. Nobody's building it yet.
Meanwhile, Karpathy scored every job in America. 42% scored 7+ for AI exposure. That's 59.9 million workers and $3.7 trillion in wages. The smart response? Whoop is hiring 600 people — not despite AI, but because of it. When execution costs drop 10x, the rational move isn't fewer people. It's more ambition.
The ad-tech stack was built for a human internet. The human internet is over. The companies that verify their audience, expand their ambition, and bet on understanding the physical world instead of autocompleting sentences will own what comes next.
On the site today: The full Signal/Noise briefing — Dead Internet, Karpathy's 342-job scorecard, and why Pokemon Go might have built the training data for the actual future → getcoai.com
Long Form: Why Elon Musk is the human orchestration router — and why the router is dropping packets → getcoai.com
From the Scroll: Mollick's "meaning-shaped attention vampires" thread, Karpathy's jobs tool, Niantic's 30 billion images → getcoai.com