Why they built it
The founders were frustrated with the hassle of setting up servers, configuring Apache, and installing MySQL and PHP to build simple web apps. They built AppJet as a browser editor where developers could write server-side JavaScript and publish immediately to a hosted URL with database and Facebook libraries, free hosting and 10 MB of storage. They thought inspectable source code plus instant deployment would help beginners and experienced developers alike.
How it ended
AppJet was not a failure shutdown; Google acquired it on December 4, 2009 and moved the team into Google Wave. By then it had pivoted from hosted server-side JavaScript to EtherPad, a real-time collaborative editor that had generated hundreds of professional-version requests. The original etherpad.com hosting later closed in May 2010, but EtherPad and its framework had been released under Apache 2.0.
What went wrong
Most important first.
- 1
Platform scope too broad for resources Capital
YC described the first product as attempting Rails and Heroku simultaneously, too much infrastructure and product for the startup's financing.
- 2
Users wanted EtherPad more than the platform Product-market fit
EtherPad began as an internal tool and technology showcase and became the company because users wanted collaborative editing more than a general development platform.
- 3
Platform revenue never materialized Monetization
The beta offered free hosting and 10 MB of storage with paid compute only as a possibility, and only a small number reportedly paid $8 per user monthly for EtherPad Professional with no disclosed revenue.
- 4
Bespoke runtime made apps hard to move Technology
Apps ran on a JVM, Rhino and Scala stack with custom isolation and scheduling, and hosted applications could require substantial work or a dedicated server to run elsewhere.
What could be different today
- EtherPad and its framework were released under Apache 2.0 on December 17, 2009, and the maintained Etherpad project preserves real-time editing, author colors, revision history and the time slider, so a rebuild can start from that code instead of rebuilding sync.
- The EasySync model is now documented as representing edits as changesets composed through a central server, giving a rebuild a published reference for the synchronization approach.
- Self-hosted continuity is proven because successors such as TitanPad and the maintained Etherpad project continued after etherpad.com stopped accepting new pads in April 2010 and closed in May 2010.