A criticism about AI in general is its ability to generate "slop"—but I feel the label masks the nuance behind it.

By design, AI slop exists because people trade speed for quality. Convenient creation. Quick and dirty.

When you expose such work to the public, the perception of "unfinished" can taint your creations.

However…

If your value stream favors speed over quality, then slop isn't so bad. In fact, it's desirable.

So why would you ever want slop?

• Prototype to learn: You ship an ugly functional alpha so users can give you feedback on what needs fixing.

• Functional benefit: Working software to complete a workflow now is more valuable than waiting for something more polished. Function over finish. (Especially if you commit to fixing the rough edges later.)

I lean on these two a lot when building prototypes with AI.

• Acceptable lower standards: The people consuming your work don't care if it's sloppy; in some cases, they may prefer to consume your work raw and unfinished.

• Speed to communication/first to market: Where getting information fast is more valuable than tightening the prose.

While I personally don't think slop is justified in these two bullets, I can understand why others do.

As a designer, I hate consuming AI slop. But, if that slop generates important value for you or its consumers, it shouldn't be dismissed.

As a designer, here's how you can embrace speed and minimize slop:

1. Commit 100% to fixing it. This is the downstream solution, where you publish slop but also tell your consumers you vow to fix and improve along the way. This is exactly how Anthropic launched Claude Design: With a very broke alpha that it committed to improving in the months post launch (prototype to learn + functional benefit).

2. Build quality upfront. The classic Lean upstream solution. Prevent slop from generating in the first place. This means creating solid AI inputs, guardrails and linters to fix as much slop as possible before release. This is where I try to center my work.

So the next time you hear someone complain about AI slop, pause and ask: Are they citing a specific example, or is it a vague blanket statement? And if specific, is the slop valuable to either the creator, the consumer or both?

Rather than settle for an instant visceral reaction to AI slop, slow down and unpack the reasoning behind the label. That analysis may surprise you.