The Internet is 50% Fake. I Built a Detector.

2026-02-18 · Source: Siraj Raval · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Emerging Technologies & Innovation · Depth: Intermediate, medium

Summary

Suriraj, in collaboration with Juny AI, developed an open-source Chrome extension called "AI Slop Detector" or "Slop Shield" designed to identify and score "slop"—defined as low information density content characterized by high predictability and a deficit of verifiable claims. The initial version, built in approximately 4 minutes and 30 seconds using Juny AI within the JetBrains IntelliJ IDE, relied on heuristics to detect issues like low lexical variety and repeated phrases. Recognizing its limitations with complex content like research papers, the detector was improved to incorporate a local 7-billion parameter AI model, specifically a quantized Qwen model running on 4GB of VRAM via Olama. This enhanced version uses a "constitutional judge" system prompt to encode values like falsifiability and epistemic modesty, providing a more accurate and private slop score by leveraging local AI judgment.

Key takeaway

For AI Engineers and developers concerned about content quality and information overload, building your own AI-powered content filters, like the Slop Shield, offers a robust defense. You can leverage AI agents for rapid prototyping and integrate local, quantized LLMs to ensure privacy and custom judgment. This approach allows you to tailor content verification to specific needs, moving beyond generic heuristics to a more nuanced, value-driven assessment of information density.

Key insights

Low information density, or "slop," can be programmatically detected using a combination of heuristics and local AI models.

Principles

• Slop is measurable as verifiable claims over content length.
• AI models can be imbued with values via system prompts.
• Local AI offers privacy and custom judgment.

Method

Build a Chrome extension using an AI agent (Juny AI) to generate code. Start with heuristics, then integrate a local 7B parameter quantized LLM (e.g., Qwen via Olama) with a constitutional system prompt to refine slop detection.

In practice

• Use Juny AI for rapid Chrome extension development.
• Quantize LLMs for local, low-VRAM inference.
• Implement system prompts to encode AI values.

Topics

• AI Slop Detection
• Chrome Extension Development
• Local AI Inference
• AI Agent Coding
• Content Verifiability

Best for: AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• LinkedIn's war on AI slop is not just a policy update—it is an admission that the platform lost control of its feed — LinkedIn is implementing new detection and verification systems to combat low-value AI-generated content and fake profiles.
• Atlas's Last Run in the Sun — AI is rapidly advancing across enterprise, social platforms, and robotics, emphasizing agentic systems and production-grade infrastructure.
• Meta rolls out new AI content enforcement systems while reducing reliance on third-party vendors — Advanced AI systems can significantly enhance content enforcement accuracy and efficiency while reducing human exposure to graphic content.
• Building AI Agents with Local Small Language Models — Local AI agents powered by SLMs offer cost-free, private, and offline operation on standard hardware.
• Are we already getting past the point of no return with Slop software? — The discussion explores the concern that AI-assisted software development is generating an excessive amount of "slop" software, primarily categorized as "Agentic Wrappers around LLMs" or "Slop apps th…
• "That's AI Slop, You Bot!" Studying Accusations, Evidence, and Credibility in Online Discourse Towards LLM-Generated Comments — Online "AI slop" accusations function as social gatekeeping for perceived authenticity, not accurate AI detection.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Siraj Raval.