AI summary by TwinMind
A YouTuber Stole My Project (And I'm HYPED)
Theo - t3․gg
15m 12s
Upload Thing Open Source Announcement
- Upload Thing service now open source, repo link to be provided
- YouTuber Josh reverse engineered Upload Thing before open source release
- Developer filmed live reaction to Josh's reverse engineering video
- Acknowledges Josh's "fantastic job" of reverse engineering and video production
Josh's Reverse Engineering Process
- Josh analyzed Upload Thing's source code and rebuilt clone service
- 90% of apps use cloud file storage (YouTube, Twitter, Instagram)
- Upload Thing simplifies developer interaction with AWS S3
- Josh spent 24 hours building functional clone with fewer features
- Source code was accidentally published in packages from version 2.0.3 onwards
- No minification needed, source directly accessible
- Open sourcing described as "more formality than anything"
Upload Thing Service Overview
- Enables easy file upload to cloud storage for developers
- Simple implementation: install package, create page.tsx, configure API route
- Demo shows drag-drop file upload with console link output
- Multi-uploader component provided by Upload Thing
- API route definition under "upload thing" endpoint required
- Router configuration defines file types and max size limits
Technical Implementation Details
- Files stored in state until upload button clicked
- Only file metadata (not actual files) sent to server for validation
- Server-side validation enforces file limits (cannot trust client)
- Pre-signed post URLs generated with S3 policy containing:
- Signature preventing client modification
- Content length range enforcement (e.g., 3MB limit)
- File type restrictions (e.g., image/* only)
Upload Architecture Flow
- Client requests server → Server validates → Server calls Upload Thing
- Upload Thing generates S3 policy and database metadata entry
- Post URL returned to client → Client uploads directly to S3
- Direct S3 upload bypasses infrastructure, eliminates egress costs
- S3 triggers Lambda on upload completion
- Lambda calls Upload Thing server with file key
- Upload Thing translates S3 metadata and calls back client endpoint
Backend Callback Implementation
- On upload complete callback runs backend-first for reliability
- Prevents "ghost files" in S3 bucket without database logging
- Backend callback should be source of truth for file tracking
- Client callbacks useful for UI updates but not reliable for persistence
- Addresses common problem: pre-signed URLs created but upload completion not tracked
Service Infrastructure
- Uses Upstash database for edge compatibility
- Amazon SNS (Simple Notification Service) handles S3 webhooks
- API key quota tracking implemented
- File size limits enforced in webhook processing
Josh's Analysis and Validation
- Josh recognizes value despite being S3 expert
- Confirms service solves real developer pain points
- Abstracts complex S3 pre-signed URL handling
- Provides convenient success/error callbacks
- Makes file uploads "super straightforward"
Development Timeline Issues
- Team late on component library and API updates
- Should have shipped Wednesday updates, delivered following week
- Delays attributed to Reactathon conference and Vercel events
- Service maintenance took priority over feature development
