Superalignment
Culture

AI slop · From culture to incentives

The output was cheap. The checking was not.

A feed can fill up when more posts earn rewards, reach is easy, and review is weak.

Personally affected 9 minute path Public draft
A project-original industrial scene in which printing presses feed a flood of repeated pages and screens, with one page visibly marked for review.
Original Superalignment illustration, made with AI assistance. It does not reproduce third-party source media.

Start with a real example

Watch the source explainer.

This video shows how media can carry a record of where it came from. That record does not prove the media is true or useful.

The example · Take the challenge

Can you read the whole system?

Five quick questions. Your first answer counts. At the end, share your score.

0 of 5 answered
1 of 5 · Read the whole setupWhat can flood a feed?
2 of 5 · Want · MissionWhat job were people trying to give the system?
3 of 5 · Count · MetricWhat stood in for success?
4 of 5 · Reach · PowerWhere could the system act or influence people?
5 of 5 · Stop · BrakesWhat best describes the human stop?

The Four Checks

Want. Count. Reach. Stop.

The example changes. The four questions stay the same. This is a reading tool, not a risk score.

  1. 01Mission
    Want

    What did people really want?

    Help or entertain

    The work is meant to help, inform, persuade, or entertain.

  2. 02Metric
    Count

    What can the system see or count?

    Posts, clicks, or money

    The system can count output and clicks more easily than care.

  3. 03Power
    Reach

    What can it change or control?

    Feeds and search

    Search and feeds can send the work to people who did not ask for it.

  4. 04Brakes
    Stop

    Who can notice a problem and stop it?

    Readers do the checking

    Readers, teachers, and coworkers must find and fix the mistakes.

What it shows · Today

Take one small step into the real world.

The first guide keeps this issue in view and adds one real idea. The other paths wait until you ask for them.

Recommended next5 min Reward hacking A system can hit the score and still miss what people wanted. Open this next
See two more paths from this story
Metric bridge5 min Goodhart's law The harder people chase a number, the less that number may mean. Open this next System bridge5 min Algorithmic monoculture When many people use the same system, one blind spot can spread to many decisions. Open this next

What it shows · Limits

What this frame gets right, and what it makes simple

One example can help us see a pattern. It cannot tell us how common that pattern is in the real world.

Useful lesson

Keep these ideas

  • Cheaper production can create more work for the same amount of human attention.
  • Polished-looking work can push fact-checking onto people with little time or skill.
  • The same rewards that create rushed human work can scale faster with AI tools.
  • In one 2025 study of 1,082 on-topic preclinical biomedical videos, 57 (5.3%) met the authors’ study definition of slop. The authors warn that their method likely missed less obvious cases, and the sample is not an estimate of general internet prevalence.

Story shortcut

Do not assume these parts are real

  • There is no single agreed meaning of 'AI slop.' People use it for bad quality, high volume, tricks, or a certain look.
  • AI help does not prove that work is careless or wrong. The process and owner matter.
  • One sample cannot tell us how much slop exists across the whole internet.

Try it · Another case

Can you spot the same pattern?

An AI-made report is correct. No one knows who asked for it, checked it, or will use it. Is being correct enough?

No. It may still have no clear use or owner. A careful process names the reader, goal, source checker, and person who must fix an error.

Choose your depth

Your next step is still in plain English.

Open the first guide now, or unfold the whole path when you want the Explainer and original source.

Plain guide5 min Reward hacking See why hitting the score can miss the purpose. Open this next
Open the full research path
Next guide5 min Goodhart's law Connect output targets to the behavior they induce. Open this next Related explainer5 min Algorithmic monoculture Inspect how shared systems can create correlated errors. Open this next Original studyPrimary paper AI slop in biomedical videos Read the 2025 study behind the number on this page, including its sample and limits. Open this next

Pass it on

Send this to the person doing the cleanup

AI slop · Want. Count. Reach. Stop.

Want Mission
Help or entertain
Count Metric
Posts, clicks, or money
Reach Power
Feeds and search
Stop Brakes
Readers do the checking
The Four Checks are a reading tool, not a risk score · superalignment.inc
Cheap output can create costly work for everyone else. Before you trust AI-made work, ask: Who asked for it? Who checked it? Who fixes it if it is wrong? Style alone cannot answer those questions.