There are slop cannons and turbo brains and people who don't use AI. And then there are people who have good judgment.

"A weak human player plus machine plus a better process is superior, not only to a very powerful machine, but most remarkably, to a strong human player plus machine plus an inferior process."

There is a story about Kasparov that when he lost to Deep Blue in 1997, he started chess tournaments where human players augmented their game with computer chess engines...and it made some people a lot better but not necessarily the chess masters.

A weak player with a better process beat a strong player with a worse process. Once computers became competent, performance depended less on intelligence and more on orchestration. It was knowing when to trust the machine, when to override it, and how to structure the collaboration so that each side compensated for the other's weaknesses.

Shapiro thinks this comes down to knowing when something is wrong:

catching errors is the entire point of the human half of the equation. That's the job. The AI handles speed, breadth, and pattern recognition. The human handles quality control, contextual judgment, and the final call on what ships and what doesn't.

AI still has trouble really knowing anything in the way humans know things. At least humans who think about what is in front of them.

Humans make mistakes because they are tired, or distracted, or tricked by biases, and for many other reasons...but LLMs make mistakes because they don't know any better. Sure humans do the same, but I think this is the primary failure mode for LLMs.

And so the person who doesn't bother thinking about their LLM's output or who can't evaluate whether it's wrong or right produces slop.

people who discovered ChatGPT, decided it was magic, and now produce astonishing volumes of confident, polished, wrong work. They ship hallucinated legal citations, auto-generated emails that miss the entire point of the conversation, code that compiles but breaks in production, and financial models built on assumptions they never examined because the output looked clean.

Shapiro calls these people Slop Cannons.

The Slop Cannon is the person who uses AI without the judgment to evaluate what it gives them. And the thing that makes this quadrant so dangerous is that the output looks good. AI writes fluently. It structures arguments coherently. It formats documents professionally. The surface-level quality is often excellent. Which means the errors it produces are harder to catch, not easier.

There's more to Shapiro's post and I found the whole thing an interesting read.