• 

  Instant Scam Verdict

• 

  Analyze Reddit Post

• 

  ModQueue Copilot supports two live-entry paths

• 

  reddit_Account

• 

  Real Moderation Action

• 

  Moderator Case File

• 

  Policy Stress Test

• 

  Measured Moderator Impact

Inspiration

Reddit moderation is one of the most important — yet under-supported — workflows on the internet. Moderators make fast decisions under pressure, often with incomplete context.

While exploring real posts, we noticed a pattern: obvious spam or scam content sometimes slipped through, while harmless posts were removed. Decisions varied between moderators, and context was often lost between shifts.

This revealed a deeper issue: moderation is not just about detecting bad content — it’s about making consistent, explainable decisions as a team.

That insight led us to build ModQueue Copilot.

What it does

ModQueue Copilot is an AI-powered Reddit moderation assistant built with Devvit. It helps moderators triage posts faster and make more consistent decisions.

It analyzes posts directly inside a subreddit and recommends one of four actions: approve, remove, review, or reply — all without leaving Reddit.

Beyond detection, it provides:

• Clear risk verdicts with supporting reasoning
  
• Structured moderator case files
  
• Policy-aware decision comparisons
  
• Moderator handoff summaries
  
• Impact metrics such as posts triaged and time saved
  

This transforms moderation from reactive guesswork into a structured, transparent workflow.

How we built it

We built ModQueue Copilot as a full-stack Devvit application running natively inside Reddit.

• Frontend: React-based moderation console optimized for speed and clarity
  
• Backend: Hono API with a deterministic moderation engine
  
• Detection: Heuristic-based signals (earnings claims, urgency, external redirects, etc.)
  
• Data: Redis for case history, activity tracking, and metrics
  

The system works on real subreddit posts and includes seeded demo scenarios for reliable hackathon demos.

Challenges we ran into

Working within Reddit’s platform constraints was a key challenge. Moderation actions are limited to the subreddit where the app is installed, which required careful design of entry points and demo flows.

Another challenge was balancing power with clarity. Early versions were feature-rich but too dense. We refined the interface to highlight critical signals — like risk level and key reasons — at a glance.

We also focused heavily on trust. Instead of a black-box classifier, we made every decision explainable and transparent.

Accomplishments that we're proud of

• Built a fully functional Devvit app that runs inside Reddit
  
• Enabled end-to-end moderation workflows in a single interface
  
• Designed a transparent system with explainable decisions
  
• Introduced policy-aware moderation for different subreddit styles
  
• Measured real impact through risk detection and time savings
  

What we learned

Moderation is not just a technical problem — it’s a workflow and coordination challenge.

Accuracy alone isn’t enough. Moderators need clarity, trust, and consistency. Clear explanations and structured context often matter more than marginal improvements in detection accuracy.

We also learned to design within real-world constraints by building directly inside Reddit’s ecosystem.

What's next for ModQueue Copilot

Next, we aim to evolve ModQueue Copilot into a full moderation operations layer:

• Prioritizing high-risk posts in the queue
  
• Assisting with automated moderator responses
  
• Providing deeper team-level insights and consistency tracking
  
• Expanding policy customization for different communities
  

Our goal is to make moderation faster, more consistent, and scalable — while keeping decisions transparent and human-centered.

Built With

• ai
• devvit
• hono
• react
• redis
• typescript