AI summary by TwinMind
GitHub's Spam Problem Keeps Getting Worse....
Theo - t3․gg
18m 38s
Sponsor Segment - Ahrefs
- Speaker promoted Ahrefs SEO tools and courses
- Spent 4+ hours watching their YouTube content
- Highlighted specific video "I use ChatGPT to rank number one in Google"
- Recommended free courses before trying paid product
- Praised educational content quality and platform capabilities
GitHub as Social Network Problem
- GitHub evolved from code sharing to social platform with public repos, issues, pull requests
- Anyone can file issues/PRs on any public repository
- Platform lacks hierarchy/prioritization between users
- New accounts have same permissions as veteran contributors
- No reputation system distinguishing established vs new users
Comparison to Twitch Safety Systems
- Speaker worked at Twitch on safety tools and moderation systems
- Twitch built with presumption of abuse at scale
- Features include:
- Zero permissions by default for new accounts
- Email verification requirements
- Phone number verification
- Good CAPTCHAs during signup
- IP detection systems
- Suspicious user monitoring with hidden messages requiring approval
Current GitHub Spam Attack Analysis
- Antonio's viral tweet revealed massive spam issue on Shadcn UI repository
- Hundreds of spam issues opened by single accounts
- Spam content includes:
- Pornographic material
- Fake livestream links
- SEO manipulation attempts
- Virus/malware distribution
- Spammers use edit technique to dodge filters (post clean content, edit in links later)
GitHub's Inadequate Spam Detection
- Mostly manual rules and keywords maintained by humans
- AI content-based classifier with different models (issue, comment, etc.)
- Poorly implemented "Aros Labs" CAPTCHA system
- Easy to circumvent by posting without links, then editing
- Edit checking less aggressive than initial post checking
Fundamental Platform Failures
- New accounts can open unlimited issues on any repository
- No rate limiting or restrictions for new users
- No moderation tools for repository maintainers
- Cannot ban users from filing issues in specific repos
- Cannot bulk delete malicious content
- Issues can be closed but not deleted (by default)
Issue Deletion Limitations
- Feature disabled by default in organization repositories
- Only organization owners can enable deletion capability
- Even with feature enabled:
- Only one issue can be deleted at a time
- No bulk deletion from specific users
- Requires manual deletion of each individual issue
- Documentation contradictory and unclear
Scope of Problem
- Affects major repositories including:
- Vercel/Next.js
- Express.js (constant README spam from tutorial followers)
- Multiple high-profile open source projects
- Spam gets indexed by Google before removal
- Maintainers forced to manually clean up or build custom bots
Proposed Solutions
- Implement reputation/karma system
- New accounts should have limited permissions until proven trustworthy
- Suspicious content should be hidden pending maintainer approval
- Add aggressive CAPTCHAs for new account interactions
- Create filtering systems for common spam patterns
- Provide bulk moderation tools for maintainers
- Default to hiding potentially suspicious issues from public/search indexing
Impact on Open Source Maintainers
- Maintainers eating cost of GitHub's lack of safety tools
- Thankless job made worse by spam management burden
- Platform not prioritizing maintainer experience
- Risk of maintainer burnout from dealing with spam
- Speaker threatens to recommend alternative platforms if not fixed
