How much slop could a slop bot stop if a slop bot could stop slop? It is a child's tongue twister with a grown-up problem hiding inside it, and the answer turns out to be most of the way to a product strategy.
First, the honest part. A lot of what crosses your screen now is slop. Not malicious, not even wrong, just average. The median of everything, generated to fill a space that asked to be filled. When making a thing cost something, the average was held down by effort. Now that making costs almost nothing, the average rises to meet you everywhere, all day, politely.
Slop is an abundance problem wearing a quality costume
It is tempting to talk about slop as a quality issue, as if the fix were better models. It is really an abundance issue. The same collapse I wrote about in "Still in Figma?", where the distance between imagining a thing and shipping it went to nearly zero, also collapsed the distance between having nothing to say and saying it at length. Both are the same event seen from two sides. Cheap creation is wonderful when you have taste and a reason, and exhausting when the feed does not. The cost floor fell out of writing, of images, of code, of opinions. When supply goes infinite, no single piece can be scarce, and scarcity is where attention and meaning used to live. I made that argument about music in "Impossibly Small". It turns out to be true about everything.
Can a slop bot stop slop?
The obvious move is to fight the machine with the machine. Build a detector. Score the feed. Let a model sort the model's output into keep and discard. This works for a while and then it does not, because detection is an arms race and the generator always gets a turn after the detector. You end up with two bots in a loop and a person watching, no more sure of anything than when they started.
So the slop bot can stop some slop, briefly. That is the wrong question. The right one is what you reach for when you can no longer trust the surface of a thing, and the answer is not a better filter. It is provenance.
The scarce thing is trust, and trust has an address
When you cannot tell whether a thing is real by looking at it, you start asking where it came from. Who made this. From what. With what hand on it. This is exactly the bet under everything I work on. The moat is the authoritative record of who made what from what, which I argued in "The Graph Is the Moat". The durable asset is a real understanding of a specific person, not a pile of average output, which is the whole of "Your AI Should Know You". Slop makes both of those go from nice-to-have to load-bearing. In a world of infinite plausible content, the things that hold value are the ones with a legible origin and a memory attached.
Designing for a sloppier world
If that is right, products should stop optimizing for volume and start optimizing for trust. Some of what that looks like:
- Make origin legible. Show where a thing came from, who touched it, what it is descended from. Provenance as a first-class part of the interface, not a footnote.
- Bring back friction on purpose. Not every action should be frictionless. A small cost, a small constraint, is a signal that a human decided this was worth doing. Constraint was the feature of the iPod nano and it is a feature here.
- Privilege memory over generation. A tool that remembers you and what you have made is worth more than a tool that can produce more of anything on demand. Generation is the commodity now. Understanding is not.
- Surface the tended, not the produced. The internet keeps what nobody was paid to keep. Design for the corners that are obviously cared for, because care is the one thing slop cannot fake at scale.
The turn
So, how much slop could a slop bot stop. Some, for a while, and never enough, because the bot is fighting the wrong war. The way through is not a better machine for sorting the flood. It is building things with an address, a memory, and a hand on them, so that the question of real or not real has an answer you can trace.
The flood is not going to recede. The move is to be a place worth standing on while it rises.