Skip to Content

Video: Does AI Think for You? (Slop Wars)

What happens when you hit enter, in under three minutes
November 13, 2026 by
Hugo Acurio
Slop Wars video: does AI think for you?

Under three minutes inside the black box. Your words become tokens, the model predicts the next piece again and again, and a vague prompt lands on the average of everything it read. That average is the slop. Bring what only you have (experience, systems thinking, judgment) and the answer has to lean on it. It reasons, inside the frame you hand it.

The follow-up to the Slop Wars article: what happens when you type a prompt.

Read the Slop Wars article

Transcript

There's a debate running on LinkedIn. Is AI slop proof that AI does the thinking for you? Let's settle it. Large language models do reason, one probability at a time. What they don't bring is human thinking, wisdom, judgment and taste.

Instead of writing about the black box, I'm opening it. I build AI in regulated healthcare, where a wrong answer costs more than likes. So, how does it work? Your words get cut into small pieces called tokens. The model predicts the next piece. Then the next. Then the next. Every time, it rereads everything before it, your prompt included. A short answer is a few hundred of those guesses.

Picture everything it ever read as a sky full of stars. Where the stars bunch up is what most people wrote, most of the time. That bright middle is the average. Ask it for a post about AI in healthcare and you land dead center. 'In today's rapidly evolving landscape.' Synergy. Alignment. There's your slop. And the model did its job. It gave the most likely answer to a question anyone could have asked.

Now bring what only you have. For me, that's 15 years in healthcare. Regulatory and pharmacovigilance. Systems thinking. Judgment. The prompt changes: 'The new PV rule lands November 4. Ecuador first, Mexico next, one ERP. Design the intake.' Same model, different part of the sky. Out here the stars thin out, so the answer has to lean on what you brought. The human thinking.

And yes, it reasons, and it's fast. In my case, the problems came to it already solved. It turned what was in my head into writing, code and video, faster than I'd ever manage alone. But it reasons inside the frame you hand it, and that frame is your judgment. It also cannot reliably tell when it's wrong. When I put an AI sales agent on our WhatsApp channel, it never typed a single price. It writes the sentence. Our ERP supplies the facts.

So, does it think for you? It reasons with whatever you bring. Bring nothing and you get everybody's average, on demand. Bad writers didn't show up in 2026. They just got faster. Which makes your prompt the most honest writing sample you own. What did you bring to your last prompt?

Share this post
Tags
Archive