Yesterday on Episode 374 of the EdTech Situation Room podcast, I was excited to FINALLY announce the first official release of PodCommons — an open-source community podcast discovery engine built with React, Firebase, and Gemini AI. No advertisements, no data tracking, no algorithmic manipulation by “Big Tech” algorithms or companies. Just my own algorithms, co-created with Claude Sonnet 4.6!
I want to acknowledge and shout out my former middle school computer programming student, Aiden, who taught me about Firebase in Spring 2025 in the context of his own vibe coding project.
Check out PodCommons and join my online PodCommons community (with a Google login) on podcasts.wesfryer.com. For more geeky details, check out the READ ME file and code files on the project GitHub repository. I’m continuing to document my journey of teaching, learning and coding with AI on ai.wesfryer.com. This is my first coding project to use an API to an artificial intelligence / AI platform, and is the most complex coding project I’ve ever created in my life. I’m quite proud of this and have been using the site daily for the past month to discover, listen to, and share a variety of podcasts. I’ve also shared a large number of past and current podcasts I’ve created / create via the website’s top tabs.
What is PodCommons?
PodCommons is a self-hostable podcast discovery platform that combines personal curation with community features and AI-powered episode analysis. The core idea: Your listening history and taste preferences should drive what you discover next, not a black-box algorithm you don’t control.
Built as a forkable, open-source project so educators, communities, and podcast enthusiasts can deploy their own instance and curate podcasts for their specific audiences.
v1.0.0 Highlights
AI-Powered Episode Analysis
Every episode is analyzed by Gemini 2.5 Flash at import time, generating topic tags from a controlled vocabulary, a taste score (0-100) reflecting relevance to the curator’s interests, and a one-sentence summary. Over 3,700 episodes were tagged in the initial database at a cost of less than $2 USD.
Discover Algorithm
The Discover tab surfaces episodes using a weighted ranking formula combining AI taste score, recency, community engagement (likes, comments), and curator-defined feed weights. Users can adjust sliders to tune how much each factor influences their feed.
Topic Filter
Episodes are filterable by topic across 15+ categories including AI & Technology, Democracy & Civic, Education & Learning, Faith & Spirituality, Media Literacy, and Podcasting & Audio. Topic counts are displayed per category.
Community Features
User accounts via Google Auth with trust tiers (visitor, member, trusted, admin)
Episode likes and comments with auto-hyperlink detection
Discussions tab for community conversation
Content flagging system for admin review
Invite code system for controlled community onboarding
Podcast Suggestion System
Any visitor can suggest a podcast via the Suggest page. Admins review suggestions in the Admin dashboard; approving a suggestion automatically adds the RSS feed and triggers an immediate AI-analyzed poll.
Admin Dashboard
Full admin interface with tabs for Feeds, Episodes, Users, Suggestions, Flags, and System settings. Includes manual RSS poll trigger, episode pruning tools, and stats overview.
RSS Polling Infrastructure
Two Firebase Cloud Functions handle feed polling: a scheduled function that runs every 4 hours and a manually-triggerable HTTP endpoint. New episodes are analyzed by Gemini at import time. A stats cache is written after each poll so the About page loads instantly.
Episode Pruning
Configurable pruning script (pruneEpisodes.mjs) keeps the database lean by removing episodes older than N days that have no community engagement, while always preserving first-party podcast episodes and any episode with likes or comments.
About Page
Live stats (podcast count, episode count, member count, likes) served from a Firestore cache updated with each poll. Includes an introductory video and open-source fork invitation.
Tech Stack
LayerTechnologyFrontendReact 18, Vite, Tailwind CSSDatabaseFirebase FirestoreAuthFirebase Auth (Google)HostingFirebase HostingBackendFirebase Cloud Functions (Node.js)AI AnalysisGoogle Gemini 2.5 Flash
First-Party Podcasts (Curator’s Own Shows)
This instance curates four first-party podcasts by Wes Fryer:
Speed of Creativity
Database at Launch
408 podcasts indexed
3,301 episodes with AI topic tags and taste scores
Gemini 2.5 Flash analysis on all episodes
Forking and Self-Hosting
PodCommons is designed to be forked. See SETUP.md for full deployment instructions covering Firebase project creation, environment variables, Firestore security rules, Cloud Function deployment, and first-party podcast configuration.
The codebase is structured so you can replace the curator’s taste profile and first-party podcasts with your own in a single configuration file.
MIT License — fork it, modify it, deploy it for your community.
Known Issues / Roadmap (v1.1)
Discussions tab bug: shows all podcasts rather than filtering to only those with actual comments -- needs a Firestore query fix to only return podcasts/episodes where comment count > 0
Discussions tab UX redesign: flip the layout so the comment text is the primary element (headline), with the episode it belongs to shown below it indented -- makes community contributions the focus rather than podcast metadata; similar to a Mastodon or Reddit reply thread pattern
Admin “Trust User” button not functional (Firestore rules fix needed)
Manual “Add Feed” button missing from Admin Feeds tab (workaround:
addFeed.mjsscript)Podcast title special character handling on import (apostrophes can mangle)
Maintenance mode (admin-toggled landing page for open-Firestore-rules windows)
Mobile UX audit pass
“Suggested by @username” badge on podcast pages
User-defined custom episode tags
Feedback link (Google Form) in footer
Acknowledgments
PodCommons was built across many late-night vibe coding sessions in April-May 2026 using Claude Sonnet as an AI pair programmer by me, Wes Fryer. Updates and reflections on the development process are shared at ai.wesfryer.com.
Please join my live instance of PodCommons and share feedback!
