Hey everyone,
Some of you might remember me from the Wave Protocol days -- I was a committer on the project, then continued through the Apache Wave era after Google donated Wave-in-a-Box to the Apache Foundation. When the project was eventually retired, I moved on -- but I never moved on from the idea of Wave. Real-time collaborative documents with threaded conversations, operational transformation, federation -- nothing else has ever quite replicated what Wave was trying to be.
For years I lacked the resources, the time, and frankly the technical bandwidth to solo-maintain a codebase this size. Wave is not a small project. It's a full-stack distributed system spanning a Java server, GWT client, protocol buffers, operational transformation engine, and a WebSocket-based real-time sync layer. Modernizing it felt impossible for one person.
Then LLMs happened.
I first tried AI-assisted development about a year ago. It was... rough. The models would hallucinate completed features that weren't actually implemented, introduce security vulnerabilities, silently break the OT engine, and confidently report "all tests pass" when nothing compiled. After a few frustrating weeks, I shelved it.
But the latest generation changed everything. Claude Opus 4.6 and GPT-5.4 with 1M context windows can actually hold the entire Wave codebase in context, reason about cross-cutting changes across dozens of files, and autonomously develop features end-to-end -- what people are calling "vibe coding." These models don't just autocomplete lines; they understand the architecture, follow existing patterns, and run compilation verification before claiming success.
So I tried again. And this time it worked.
Starting from the retired Apache Wave codebase, here's what's been built:
Infrastructure ModernizationJakarta EE migration -- full javax-to-jakarta servlet migration with dual-source compatibility
SBT build system -- migrated from Gradle to SBT with native-packager for distribution
MongoDB 4.x driver -- modern async driver replacing the legacy MongoDB client
Java 17+ runtime -- updated from Java 8 with proper module system support
JWT authentication -- session persistence across deploys via signed tokens
Zero-downtime deploys -- Docker Compose rolling updates with health checks
CI/CD pipeline -- GitHub Actions with automated build, test, and deploy to production
Email authentication -- registration with email confirmation, password reset, magic link login (via Resend API)
Wave pinning -- pin important waves to the top of search results
Tag system -- add tags to waves with autocomplete, search by tag:name
Custom folders -- inbox, archive, and pin-based organization
Version history -- browse wave history with diff view and one-click restore
Saved searches -- pin frequently-used searches to the toolbar
Contacts autocomplete -- Gmail-like contact picker when adding participants
Infinite scroll -- smooth pagination with responsive page sizes
Modern top bar -- clean navigation with profile, notifications, admin access
Profile system -- user profiles with Gravatar avatars, last seen, member since
DM waves -- direct message detection with _dm tag
Inline reply navigation -- slide between reply threads with breadcrumbs
Toast notifications -- non-intrusive feedback replacing all browser popups
Skeleton loading -- smooth loading states instead of blank screens
Public wave pages -- SSR-rendered pages accessible without login
Public waves directory -- browse all public waves at /public
Share button -- one-click copy public link from wave toolbar
SEO optimization -- Open Graph, Twitter Cards, JSON-LD structured data
Auto-refresh -- public pages update content every 30 seconds
Admin panel -- user management with roles (owner/admin/user), suspend/ban
Contact page -- public contact form with admin submissions viewer
Legal pages -- Terms of Service, Privacy Policy
Server-side rendering -- full SSR pipeline for fast initial loads
Webhook monitor -- GitHub webhook integration for deployment automation
200+ pull requests merged
1,300+ commits on the modernized fork
E2E test suite -- registration, login, WebSocket wave creation
Wiab.pro integration -- ported contact tracking, saved searches, and other features from the Wiab.pro fork
For the curious:
Create an account, start a wave, invite someone. It's the same Wave you remember -- threaded conversations, real-time collaboration, inline replies -- but modernized and actually running in production.
The project is open source: github.com/vega113/incubator-wave
If you're interested in contributing, the codebase is in better shape than it's been in years. The AI-assisted workflow means features that would have taken weeks of solo work now take hours. There's a lot more to build -- federation, mobile experience, rich media, AI-powered features -- and Wave's architecture is uniquely suited for the collaborative AI era.
The wave never really ended. It just took a while to come back.
--
You received this message because you are subscribed to the Google Groups "Wave Protocol" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wave-protoco...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/wave-protocol/37d1896c-2da8-4021-b5c7-3d70e217b340n%40googlegroups.com.