Marie and Chris hop on CodePen Radio for a little chat on how we're still happily using WordPress for a variety of things here on CodePen. We've got it hosted on a subdomain and on a different (WordPress-specific) hosting plan than our main monorepo, so it has its own deployment process and such. We use WordPress for our blog, where this very podcast is published, because WordPress handles RSS well, and a podcast is an RSS feed with an audio file. We also, quite importantly, use it for our docs. For 2.0, we re-wrote the docs and used a Custom Post Type in WordPress to organize them, which feels much more organized.

When we set out to build the 2.0 editor, another editor in the CodePen cannon, the goal was actually to reduce the number of editors we have and support. Ideally, reduce it to one. The trick is that anything an existing Pen or Project can do, this new editor needs to be able to do as well. All while making sure it's forever future extensible. That was the wild engineering task in front of us, and we're happy to report it's done.

To be clear, at the time of publishing: all Projects have been converted to 2.0 Pens, so Projects doesn't even exist anymore. Someday we'll convert Pens too, but not particularly soon. Instead, you can choose to convert Classic Pens (with a button click under the Save menu) if you want to.

With the recent launch of Custom Agents, Notion became the collaborative AI workspace where teams and agents work side by side. And now, their new Developer Platform is turning that workspace into infrastructure developers can build on.

Marie has her eye on what seems to be super popular with CodePen users so far this year.

With the recent launch of Custom Agents, Notion became the collaborative AI workspace where teams and agents work side by side. And now, their new Developer Platform is turning that workspace into infrastructure developers can build on.

Rachel and Chris on the podcast this week discussion the (ahem, rather large news) that every CodePen 2.0 Pen is versioned. That's right, you can jump back in time to previous versions at any time in case mistakes were made. You can take a look at the preview and code of past versions in case you just need to reference something. It all works extremely fast and efficiently. We get into the why and how it all works.

Shaw and Chris talk about how the keyboard shortcut situation is challenging, but in the best shape it's ever been in for our 2.0 editor. Between the operating system, browser, CodeMirror, and Emmet, the space is fairly crowded, but we've got enough room to offer lots of useful stuff. The commands are more findable than ever with our new Command system and the Omnibar.

We've had realtime features on CodePen for ages. Back when it was pretty damn hard. Our Collab Mode is an obvious one, where users can code together. (This "just works" in the 2.0 editor, it's not called anything special.) That and Professor Mode used to have realtime chat (until we pulled it because very few people used it). We went through iterations of this including our own implementations and using services like PubNub.

Code collaboration in realtime is particularly difficult. We used FirePad/FireBase for a while for this, and honestly that was a nice abstraction. Increasing the difficulty though, we need to layer on additional needs like syncing settings, presence, and just arbitrary message delivery (e.g. "a new build is ready!"). We used Ably for much of this for quite a while.

These days, our whole realtime system integrates with our caching layer, goes over the network using standard web sockets to our beefy Go servers, and is integrated with CRDTs/Y.js to keep everybody looking at the same thing. Plenty of open source software at work, but no more third-party services. This means lower operating costs and a simpler architecture. The cost is that problems are, well, our problems.

Rachel and Chris on the show this week to discuss a series of big changes over the last, say, six months or so with our billing system. We've essentially re-written this thing several times, and obviously this is the best time. Having three plans, two payment providers, teams, and fifteen years of history is a lot to manage.

An important aspect of the journey was getting the billing information into a single table in our database, and relying more on dynamic calls out to the payment providers when needed rather than trying to keep too much data in sync.

Of course we wanted to clean up the codebase and get payment APIs ported over to our latest system, but the biggest need this was all satisfying was UX. We wanted a proper pricing page, better pages for people to manage their billing, and really easy upgrade modals inside our 2.0 editor.

The good news is, it all worked.

Having that first response from the server to users browsers be, ya know, full of good, useful, renderable HTML is certainly better than an empty

Stephen and Chris jump on the podcast to discuss this journey for us as it relates to Next.js. We got it "for free" for our Ruby on Rails pages, but as we're slowly changing frameworks over the years, we've got new challenges, and some gray-hair inducing bugs.

We started on Next.js' "page router" and made changes to our Apollo Cache setup to support SSR. Then, over time, moved to the "app router" and did lots more work getting ready getting the most complex page on our site, the 2.0 editor, completely SSRd. That meant getting rid of using things like localStorage for anything view related and fighting bugs related to React somehow mounting itself to the wrong root.

Stephen and Chris chat about how the 2.0 editor (well, the compiler really) understands and supports Browserslist, a config format that lives in your package.json file that multiple blocks can understand and make browser support decisions against.

With the recent launch of Custom Agents, Notion became the collaborative AI workspace where teams and agents work side by side. And now, their new Developer Platform is turning that workspace into infrastructure developers can build on.

As we're getting close to rounding out the Beta period of the 2.0 editor, we're trying to close out any bugs we find or users report quickly. They could be browser-support related, network conditions related, account capability related, or just bugs in how the 2.0 editor and technology behind it works. It's complicated enough that the best way to debug things is to see exactly what the user sees when they have trouble.

A very cool side effect to having built the 2.0 editor with Apollo is that we have a nearly complete look at what is happening in the editor by virtue of the Apollo Cache (we talked about what that is here). We built a tool that can export that as JSON data, and we can load it locally to see exactly what the user sees. It's a bit fancier than that, doing things like saving browser console error logs and stuff, but that's the gist of it.