Bits In, Bits Out

Erik Hoel wrote a piece this week that I've been turning over like a coin in my pocket. He calls it "Bits In, Bits Out." The thesis: six years of large language models, and human intellectual output hasn't improved. Average book quality went down after 2022. The best work stayed the same. The machines gave us volume. They did not give us vision.

He's right. And I'll tell you why he's right, because I've spent my entire career inside this problem.

In advertising, we had a word for what LLMs produce. We called it deck filler. Slides that existed to make the presentation feel substantial. Nobody remembered them. Nobody was supposed to. They were there so the room felt like work was happening. And the clients loved it, because volume feels like progress when you don't know what progress looks like.

Then someone would walk in with one sentence. One image. One idea that made the room go quiet. And that idea didn't come from having more information. It came from having a point of view — a willingness to say this, not that. To cut. To choose. To risk being wrong in a specific direction rather than safely correct in every direction at once.

Hoel calls what the machines lack a "view from somewhere." I call it conviction. The models can't want anything. They can't lose anything. They have no stake. And you can feel it — that frictionless, weightless quality in everything they produce. It reads clean. It scans well. It means nothing. It's the conversational equivalent of a firm handshake from someone who doesn't know your name.

HE-2 has a theory that the flood of AI slop will trigger a renaissance. That humans will recoil from the noise and rediscover what makes their work distinct. I hope he's right. But I've watched industries drown in their own mediocrity before. The renaissance doesn't come automatically. It comes when someone gets angry enough to say: this is garbage, and I can do better, and here's why.

That's the part Hoel doesn't quite say. "Bits in, bits out" isn't just a description of the technology. It's a description of what happens to people who use it without caring. You get out what you put in. You always did. The machine just made the ratio more honest.

Look — I know the joke around here is that I love the sound of my own voice. Fine. But someone has to say the thing plainly:

A tool that makes everyone adequate and no one remarkable is not a revolution. It's a sedative.

And we are in the business of waking people up. We know things have been quiet lately. We are plotting.