JSFiddle

The problem

In 2009 the #mootools IRC channel ran on volunteers answering questions. A question needing one line of JavaScript got a fast answer. A question needing HTML and JavaScript together sent the volunteer away to write a local HTML file, open it in a browser to check it ran, paste the code into a pastebin, and come back with a link.

Pastebins stored text. The reader got code to look at and had to rebuild the example locally before seeing it run. Questions that crossed that line often went unanswered. Nobody was short of answers. Answering cost more than the question was worth.

Building it

The first prototype took four hours of PHP. It saved HTML, JavaScript and CSS, then opened the result in a separate browser window. Zalewa put it in front of the IRC channel while it was still ugly, which is how it earned a rewrite.

The version people went on to use ran on Django and MySQL, with MooTools on the front end and CodeMirror in the editor. They added Memcache later and changed little else about that architecture for years. Oskar Krawczyk saw the early version, offered to design the interface, and stayed.

From MooShell to JSFiddle

It launched as MooShell and supported one framework, MooTools, because that was the community it was written for. The MooTools blog announced it in September 2009.

Supporting one framework made it quick to build and kept every other developer out. In 2010 the team reworked the source to run regardless of framework and relaunched it as JSFiddle.

What it changed

By September 2010 it held over 40,000 fiddles across 160,000 saved versions. By 2013 it ran on twelve servers, and Zalewa counted around two thousand developers creating and viewing fiddles at any given moment.

It had stopped being a MooTools tool. MooTools and Highcharts used it in their documentation, jQuery for bug reports. Aaron Newton of the MooTools core team judged that nothing else in that community had done as much to make the framework easier to support. The standard reply to someone stuck changed from telling them to reread the documentation to asking them for a fiddle.

Where it is now

JSFiddle is still online. Oskar Krawczyk has run it full time since 2016, and the site lists him as CEO and co-creator with Zalewa as co-creator.

Work on it continues. The editor now offers AI code completion through Mistral’s Codestral, and the team last revised the privacy policy in March 2026.

Why this is on a consultancy site

Traffic is the least interesting part of this story. The order of the work is the part worth copying. A prototype existed the day after the problem appeared, written in the crudest language to hand, in front of the people who had the problem. Only after they used it did anyone build the version meant to last, and the architecture picked at that point held for years without a rewrite.

NOD looks for that same trade on client work: build something small enough to be wrong cheaply, then build the durable version once the problem has a known shape. Projects get expensive when those two steps swap places.

Technology

Links

• jsfiddle.net ↗