SLOP SCAN | Build tools that fight AI-generated slop

The internet has a quality problem. Build tools that catch it.

May 29 – Jun 1, 2026 · Online · Free · $1,800 in prizes

01 /

The Problem

“Slop” was the 2025 Word of the Year.^[1][2] Merriam-Webster defined it as “digital content of low quality that is produced usually in quantity by means of artificial intelligence.”^[3] Some researchers predict that the majority of online content will be synthetically generated within years.^[4] Nobody is asking “was this made with AI?” anymore. The question now is “did a human actually check this before hitting publish?”

You already know what slop looks like. The PR description that summarises the diff you can already read. The onboarding doc that restates its own heading for three paragraphs. The cover letter that could belong to literally anyone. The five-star review that reads like every other five-star review. The blog post that ranks on page one and says absolutely nothing.

Generating slop is easy. Catching it? Almost nobody is working on that.

Slop Scan is a 72-hour hackathon. Pick a domain where AI noise is doing real damage. Build a tool that spots what people can’t, or won’t, spot on their own.

We’re not against AI. We’re against the lazy defaults. Nobody should have to waste 10 minutes reading something that took 10 seconds to generate.

You have 72 hours.

02 /

8 Tracks

Each track covers a domain where AI slop is doing real damage. Pick one.

Code Review

Detect hollow AI-generated pull requests, commit messages, and code comments that look right but say nothing.

Build tools that:

→ Detect auto-generated code review artifacts and flag hollow documentation
→ Surface commits where the human clearly didn’t read what the AI wrote before pushing
→ Score PR descriptions for information density vs. diff-restating filler
→ Analyse commit message patterns to identify bulk AI-generated contributions

DevOps engineersSenior developersPlatform teams

Docs & KBs

Scan internal documentation for AI-generated filler that sounds correct but teaches nothing.

Build tools that:

→ Score documentation density — information per sentence, concrete examples per section
→ Detect circular explanations where paragraphs reference each other without adding content
→ Flag docs that don’t contain a single concrete example, code snippet, or specific instruction
→ Compare documentation claims against actual codebase behaviour

Technical writersDeveloper advocatesEngineering managers

Hiring & Resumes

Expose AI-generated materials flooding the hiring pipeline. Resumes, cover letters, and take-homes that read the same.

Build tools that:

→ Analyse application materials for AI-generation signals and templated structures
→ Detect copy-paste take-home submissions across candidate batches
→ Flag suspiciously uniform writing styles within a hiring pipeline
→ Help hiring teams design AI-resistant evaluation methods

Engineering managersHR tech buildersRecruiting teams

Communications

Filter inflated AI-expanded messages in workplace communication channels. Three paragraphs where three words would do.

Build tools that:

→ Flag inflated corporate-speak in internal comms and score signal-to-noise ratio
→ Detect when a Slack message was likely AI-expanded from a five-word thought
→ Surface email threads where replies contain less information than the subject line
→ Analyse meeting notes for content-free summarisation patterns

Engineering managersTeam leadsProductivity tool builders

Content & SEO

Build detection for the internet’s biggest slop vector. Blog posts that rank but say nothing, listicles that repeat.

Build tools that:

→ Detect AI-generated SEO content at scale and score articles for originality vs. rehash
→ Flag product listings with likely hallucinated attributes or specifications
→ Build browser extensions that warn users before they waste time reading AI filler
→ Analyse content farms and surface patterns in mass-produced low-quality articles

Content teamsSEO professionalsBrowser extension builders

Academia

Protect scholarly integrity from AI-generated papers, fabricated citations, and peer reviews that say nothing.

Build tools that:

→ Detect AI-generated sections within academic papers and flag stylistic inconsistencies
→ Verify that cited sources actually exist and say what the citation claims they say
→ Flag suspiciously similar peer reviews across a review batch
→ Help admissions teams identify templated or AI-generated essays

ResearchersAcademic publishersUniversity admissions teams

Marketplaces

Expose fake AI-generated reviews flooding every marketplace. Five-star ratings that all use the same sentence structure.

Build tools that:

→ Cluster suspiciously similar reviews and detect AI-generated product feedback
→ Score review authenticity based on linguistic patterns, timing, and reviewer history
→ Build browser extensions that filter probable fake reviews before purchase
→ Analyse Q&A sections for bot-generated answers

Consumer advocatesE-commerce teamsBrowser extension builders

Social & News

Detect synthetic content flooding feeds, fabricating reality. Bot networks, fake news, and engagement-farmed rage bait.

Build tools that:

→ Detect AI-generated social posts at scale and flag synthetic news articles
→ Identify bot-network posting patterns and coordinated inauthentic behaviour
→ Create feed filters that surface authenticity signals alongside content
→ Analyse comment sections for bot-generated engagement patterns

JournalistsTrust & safety teamsCivic tech builders

03 /

Deliverables

After 72 hours, we want to see something that works.

A working tool. Not a slide deck. Not a concept. Something that takes input, runs analysis, and shows a human something they would have missed. Hosted somewhere, runnable locally, or demoed on video.

Actual detection logic. Not just keyword matching. Show us how your tool decides what counts as slop. Linguistic analysis, pattern matching, statistical methods, structural checks. Whatever you use, explain why it works.

Honest numbers. False positive rate, detection accuracy, the cases where it fails. Judges will trust you more if you say “this catches 60% of slop and here’s why the rest gets through” than if you claim 99% accuracy.

You’ll submit:

→Live link (Vercel, Netlify, GitHub Pages, etc.) OR executable demo OR video walkthrough
→Source code (public GitHub repo)
→Brief README explaining the detection approach
→2–3 minute demo video showing the tool catching real slop

04 /

Bonus Points

Optional. Pick one and nail it. Don’t rush through all of them.

The Bake-Off

Run your tool against a known dataset of AI-generated vs. human-written content and publish your accuracy metrics. Show the confusion matrix.

Medium

Live Fire

Demo your tool against real content scraped from the wild — actual PR descriptions, real Amazon reviews, live social media posts. Not synthetic test data.

Hard

Open Source Ready

Publish your tool as an installable package with documentation, CI, and a clear contribution guide.

Medium

Cross-Track Scanner

Build a tool that meaningfully detects slop across two or more tracks from a unified detection engine.

Hard

05 /

Out of Scope

Save yourself the trouble. These won’t score well:

✗Simple keyword detectors that flag em-dashes and call it a day
✗Wrappers around GPTZero or Originality.ai with a new UI on top
✗Tools that just ask another LLM “is this AI-generated?” (that’s delegation, not detection)
✗Projects that shame people for using AI instead of surfacing quality problems
✗Ideas too big to demo in 72 hours
✗Anything requiring hardware, VR, or non-standard setup

We’re not trying to ban AI. We want low-effort output to be visible so people can make their own call.

06 /

Timeline

All times UTC.

Pre-Event

May 1, 2026

Registration Opens

May 26, 2026

Team Formation

Find teammates on Discord or register as solo.

Hackathon — 72h

May 29, 2026 @ 10:00 UTC

Kickoff & Hacking Begins

Opening ceremony, track assignments confirmed.

Jun 1, 2026 @ 10:00 UTC

Code Freeze & Submissions Due

Final commits. All project submissions must be in.

Post-Event

Jun 1–Jun 10, 2026

Judges Evaluate Projects

Judges review and score all submissions asynchronously.

Jun 11, 2026

Winners Announced

Official winners revealed.

Jun 11–Jun 17, 2026

Community Voting

Vote for your favourite projects from fellow participants.

Jun 18, 2026

Community Choice Winner Announced

Community-voted winner announced.

07 /

Scoring

How submissions are scored.

Detection Accuracy 30%

Does it actually catch slop that a human would miss? How often does it flag clean content by mistake?

Practical Usefulness 25%

Would you actually install this on Monday? Does it fit into a real workflow?

Technical Execution 20%

Clean code, smart architecture, handles edge cases. Works on real data, not just cherry-picked demos.

Innovation 15%

Did you find a detection angle nobody else thought of? A signal that’s hard to fake?

Presentation & Demo 10%

Can you show it working on real content and explain why your approach holds up?

Bonus Challenges

Challenge	Difficulty	Points
The Bake-Off	Medium	+5
Live Fire	Hard	+5
Open Source Ready	Medium	+3
Cross-Track Scanner	Hard	+3

08 /

Prizes

$1,800 total prize pool

1st Place $800

Grand Prize

Best overall slop detection tool. The project that made judges say “I need this installed right now.”

2nd Place $400

Runner-Up

Exceptional execution across the board. Strong detection, clean code, practical output. Almost took the crown.

3rd Place $200

Third Place

A creative detection approach that stood out from the pack. Showed real promise and clear thinking.

Best X-Ray Place $100

Best X-Ray Effect

Most compelling way to make a hidden slop problem visible. Showed people something they didn’t know was there.

Community Place $300

Community Choice

Voted by fellow participants after the hackathon. The project that people actually wanted to use themselves.

09 /

Rules

What counts as a valid submission.

Standalone & Runnable

Your project must run locally with a single command (npm start, docker-compose up) or be live at a public URL. No custom environments.

Demo-Ready

Prepare a 5-minute live demo for the judges on Discord. Pre-recorded video backup is fine, but live is preferred.

New Work Only

All code written during the 72-hour window. Open-source libraries and frameworks are fair game. No pre-built projects.

Team Size

1–4 people per team. Solo hackers welcome. Find teammates on the Hackathon Raptors Discord before or during the event.

Pick a Track

Choose one primary track (A–H). Your project should clearly target that domain. Cross-track work earns bonus points.

Source Code

All submissions must include a GitHub repo with source code. Repos go public after judging. No proprietary dependencies.

AI Tools Are Fine

Claude, Cursor, Copilot, vibe-coding. We judge what you ship, not how you built it. Just disclose your tools in the submission form.

10 /

Participants

If you’ve ever wasted time on content that clearly nobody reviewed before publishing, this is your hackathon.

Senior Engineers & Tech Leads

You’ve seen the PR descriptions that say nothing. The docs that explain nothing. Build the tools your team actually needs.

Track ATrack BTrack D

Data Scientists & ML Engineers

Detection is a classification problem. Bring your NLP chops to the biggest labelling challenge online right now.

Track FTrack ETrack H

DevOps & Platform Teams

Build the CI/CD gates and pre-commit hooks that catch slop before it ships.

Track ATrack BTrack C

Product Managers & Designers

Slop is a UX problem. Build interfaces that help people make quality decisions quickly.

Track GTrack ETrack D

Researchers & Academics

Your field’s integrity is under pressure. Build the tools that protect it.

Track FTrack H

Browser Extension Builders

The best slop detectors will live where people read. Build for the browser.

Track ETrack GTrack H

11 /

Judges

People who’ve shipped detection systems and care about this stuff.

Detection & NLP

2 judges

Built classifiers, shipped spam filters, or worked on content moderation at scale.

Developer Tooling

1 judge

Shipped linters, CI/CD pipelines, or code quality tools used by real teams.

Trust & Safety

1 judge

Worked on content integrity, fake review detection, or platform abuse prevention.

Open Source

1 judge

Maintained popular open-source projects. Knows what good code and good docs look like.

Hiring & Recruiting

1 judge

Reviewed thousands of applications. Knows what real experience looks like on paper.

Academia & Research

1 judge

Published peer-reviewed work on AI-generated text detection or computational linguistics.

The panel includes senior engineers from FAANG and top-tier startups who’ve built detection systems, scaled content moderation pipelines, and reviewed thousands of real-world submissions. Full profiles will be announced closer to the event.

FAQ

Is it free to participate?

Yes. Completely free.

Do I need a team?

No. Solo is fine. Teams can be up to 4 people. Find teammates on Discord or go alone.

What skill level do I need?

Mid-level and above will get the most out of this. Detection is a hard problem. That said, all levels welcome and mentors are around 24/7.

Can I use AI-assisted coding tools?

Yes. Claude, Cursor, Copilot, whatever. We judge what you ship, not how you built it. The irony is noted.

What tech stack should I use?

Whatever lets you ship in 72 hours. Python, TypeScript, Rust, Go. Browser extensions, CLI tools, web apps, API services. Your call.

Can I wrap an existing AI detection API?

Not as the whole project. Using GPTZero as one signal in a bigger pipeline is fine. Putting a UI on someone else’s API and calling it done is not.

Is it online or in-person?

Online. Discord. Build from wherever you are.

Can I start working before the hackathon?

No. All code must be written during the 72-hour window. Research and team formation beforehand are fine.

What happens after the hackathon?

Projects go public. Community votes on favourites. Judges give feedback. The good tools tend to get adopted.

“Slop” was the 2025 Word of the Year from both Merriam-Webster and the American Dialect Society. Not “was it made with AI?” The question is “did anyone bother checking it?”

13 /

Why Now

Generating content got cheap. Evaluating it didn’t. That gap is the whole problem.

GitHub Copilot now writes 46% of code in files where it’s enabled.^[6] Up to 22% of computer science papers show signs of AI-generated content.^[7] On Amazon, 3% of front-page reviews are AI-generated — and 74% of those are five-star ratings.^[8]

Code review depends on the assumption that someone actually read the code. Documentation depends on someone caring enough to write a real explanation. Hiring depends on applications that reflect real experience. Academic publishing depends on peer reviewers doing real work. Marketplaces depend on reviews from people who bought the thing.

None of these systems break all at once. They just get a little worse every week. The PR descriptions get vaguer. The docs get more circular. The reviews get less trustworthy. One day you realise you don’t trust any of it, and you can’t point to when it changed.

That’s what slop does. Not destruction. Erosion.

The internet has a quality problem. You have 72 hours.

References

[1]Merriam-Webster, “2025 Word of the Year” — merriam-webster.com/wordplay/word-of-the-year
[2]American Dialect Society, “2025 Word of the Year Is Slop” — americandialect.org/2025-word-of-the-year-is-slop
[3]Merriam-Webster, “Slop” (dictionary entry) — merriam-webster.com/dictionary/slop
[4]Europol Innovation Lab, “Facing Reality? Law Enforcement And The Challenge Of Deepfakes”, 2022 (citing Nina Schick, Deepfakes: The Coming Infocalypse, 2020)
[5]CoverSentry, “AI Job Search Statistics”, 2025 — coversentry.com/ai-job-search-statistics
[6]GitHub / Thomas Dohmke, “GitHub Copilot generates 46% of code”, 2025
[7]Zou et al., Nature Human Behaviour, “Monitoring AI-Modified Content at Scale”, 2024 — science.org/content/article/one-fifth-computer-science-papers-may-include-ai-content
[8]Pangram Labs, “AI-Generated Amazon Reviews”, 2025 — pangram.com/blog/ai-amazon-reviews
[9]Pangram Labs / Nature News, “Major AI conference flooded with peer reviews written fully by AI”, 2025 — nature.com/articles/d41586-025-03506-6
[10]Ahrefs, “What Percentage of New Content Is AI-Generated?”, 2025 — ahrefs.com/blog/what-percentage-of-new-content-is-ai-generated
[11]NewsGuard AI Tracking Center, 2025 — newsguardtech.com/special-reports/ai-tracking-center