DEV Community

Johann Hagerer
Johann Hagerer

Posted on

Presentation Slides for #StopTheSlop

#slop #aislop #ai #spam

The following slides I presented on PyCon 2026 as lightning talk.

Vimeo Link - jump to 05:45:00 to watch the presentation.

Stop the Slop!

A 5-Minute Call for Action for the AI-Assisted Workplace

PyCon DE 2026 — Lightning Talk

It's Official.

"AI slop": Merriam-Webster 2025 Word of the Year.

Low-quality digital content produced, usually in quantity, by means of AI.

  • Like "spam", but now it's coming from your team.

  • Your colleagues, PRs, tickets, chat.

  • No spam filter, no mute button, no protection.

Nobody is safe.

The Numbers

  • 40% of workers received "AI slop" in the past month
  • 15% of all content received at work is AI slop
  • $9M / year in lost productivity for a 10k-person org
  • 61% hide their AI use from colleagues
  • 55% pass off AI output as their own work
  • 2/3 never check AI output before sending it

-> AI should save time. Now, we're drowning.

curl Rage-Quit

  • Daniel Stenberg (curl maintainer) shut down a bug bounty after too much AI slop:

"We now ban every reporter INSTANTLY who submits AI slop. We are effectively being DDoSed."

  • curl's security.txt now reads:

"We will ban you and ridicule you in public if you waste our time on crap reports."

  • 📊 20 submissions in 16 hours. None were real vulnerabilities.

Code Reviews Gone Wrong

  • A study of 1,1k developer posts:

Individual devs ship faster. Reviewers, maintainers & the community pay the price.

  • Example: An AI agent hallucinated external services, then mocked out these services, i.e. an internally consistent but verbose hallucination.

"They're literally just using you to do their job — critically evaluate their AI slop and give it the next prompt."

What Can You Actually Do?

As a colleague:

  1. Be transparent — "I used AI for this" lets collaborators calibrate effort.
    • BTW: I actually used AI for this presentation
  2. Keep generations small — Every slide in this presentation is short and thus easily quality assured.
  3. Manually refine — Read all AI output. Make sure you understand it and remove unnecessary and wrong information.

What Can You Actually Do?

As a developer:

  1. Two-prompt rule — If the 2nd attempt doesn't improve, stop prompting. Write it yourself.
  2. Keep generations small — If you can't explain the output, reject it.
  3. Be transparent — "I used AI for this" lets reviewers calibrate effort.
  4. Don't dump AI plans on teammates — Pasting a raw Claude plan into a ticket is asking others to do your work.

What Can You Actually Do?

As a team:

  1. Review = understanding — If no one understands the code, it doesn't ship.
  2. Name it — If you see slop, say "this looks like slop." Normalize the feedback.
  3. Rule of thumb: If you can't defend it in a review, don't submit it.
  4. Establish best practices: DSPI (look it up!)

Remember:

AI slop is not a technology problem.

It's a laziness problem.

The tool is fine. The copy-paste is the crime.

Resources:

  • 🌐 stoptheslop.dev — Internal dev team policy template
  • 📄 arxiv.org/html/2603.27249 — "An Endless Stream of AI Slop" (academic study)
  • 📝 daniel.haxx.se/blog — curl maintainer's AI slop chronicles
  • 🔧 github.com/hardikpandya/stop-slop — Claude skill to de-slop your prose

Sources:

  • BetterUp Labs / Stanford Social Media Lab (2025) — hbr.org/2026/01/why-people-create-ai-workslop-and-how-to-stop-it
  • Global survey of 32,352 workers across 47 countries — theconversation.com/ai-workslop-is-creating-unnecessary-extra-work-heres-how-we-can-stop-it-267110
  • Heidelberg / Melbourne / Singapore, 2026

Top comments (0)