Growing Marimo's YouTube channel, with Vincent D. Warmerdam
Marimo's YouTube channel hit half a million views in its first six months, mostly from Shorts announcing tiny feature releases. Vincent Warmerdam explains why a $40 drawing tablet and completely unscripted life-coding beats a heavily edited video.
Vincent D. Warmerdam works on DevRel at Marimo, a next-generation Python notebook, and has spent the last seven months building the company's YouTube channel from scratch — reaching roughly 650,000 views along the way. In this episode he walks through exactly how that happened.
Jack and Vincent get into why Marimo's small team releasing software two to three times a week is the real engine behind the channel's growth, the phased strategy that moved from pure education to Shorts to targeted community outreach, why Vincent never scripts anything beyond the intro, and his simple gut-check for knowing whether you're actually excited enough about a product to make good content about it.
The Q&A below is experimental — extracted from the episode transcript with AI. If you spot a mistake, let us know at jack@scalingdevtools.com.
What results have you actually seen from Marimo's YouTube channel?
When we first spoke, I'd worked at Marimo for six months and we had half a million views in half a year. I'm now at month seven and we just reached about six hundred fifty thousand. The thing that really caused a lot of these views is that the Marimo team is small but we make two to three software releases a week, so there's never a shortage of small things to announce. That's what drove us toward YouTube Shorts — if you make five small, demoable changes, that does quite well as Shorts. Each one gets a minimum of around two thousand views, and something like 40 to 50% of those views are from people who've never watched our channel before.
Can you give an example of what one of these Shorts actually looks like?
We support Python 3.14 now — that's a thing you can announce. You show the feature working in a Marimo notebook, tell people to download the latest version, done. That kind of video wouldn't do amazing on the new-viewer metric though — it's more for our existing five thousand subscribers who are hardcore users interested in new features. Something that did really well with new viewers, around 17,000 views, was a bit sillier: I built a 2D slider that works inside a Marimo notebook, using our custom widget mechanism. I've also made ones where you connect a gamepad through the browser's gamepad API to trigger Python functions. Individually those demos aren't enough to make someone download Marimo, but if you see ten of them, at some point you think, these people release a lot of software and have interesting features — I'm a Python person, I'd like to try this.
Do you pay much attention to subscriber counts and other metrics?
We do a gut check on the stats every week — subscribers, view counts, whether new videos are reaching new people. But the thing that's actually meaningful to me is qualitative. We have a questionnaire that pops up in a few places, like when someone joins our Discord, asking how they learned about Marimo. Some people just say one of Vincent's videos, which is a little strange since they know me on a first-name basis. It's not rigorous, but on a gut level it feels like roughly one in five people who fill that in mention discovering us via YouTube. We also hear from Airflow or Django communities that someone wants to give a talk about Marimo at their meetup after seeing a video — that pattern is something we try to repeat.
How did your approach to content change over time?
When we started, the first month or two was almost entirely education — the kind of content that's really just an extension of the docs, because people compare Marimo to Jupyter and there's a lot of Jupyter habits to unlearn. We only tried Shorts after about four months. After the frequently-asked-questions backlog was covered, we moved to more topical videos, and now we've settled into picking specific communities and offering something targeted — one main educational video a week for a specific community, plus Shorts, plus space for anything topical that comes up. An example: I'm also a Django developer, and Django's admin interface doesn't easily support a slider, but it's easy to build in Marimo. So I made a tutorial on getting Marimo working inside Django admin, and that's a way of nerd-sniping a specific community — the Django people who saw it started giving us feedback, and I try to repeat that pattern for other communities like Airflow.
How do you actually make the videos — is there a script, or an example app you build first?
Nothing is scripted except the intro. Everything else is me life-coding. I have a $40 Huion drawing tablet from Dutch Amazon — a genuinely good device — and an app on Mac called Screen Brush that lets me draw over whatever code is on screen with a keyboard shortcut. So it's mostly: write something in a notebook, get it working, hit record, look at the camera, explain what I just coded. I do my prep beforehand to make sure the code works, but I don't build an app first and then write a tutorial about it afterward the way some people do.
What's your advice for someone deciding whether they're the right person to do DevRel or content for a tool?
It really comes down to excitement — can you expose your excitement in a way that makes other people eager to listen? I do a kind of test whenever I see a new tool: if I use it and a week later I find myself building plugins for it, that's usually a sign. With Marimo, I made three fairly substantial plugins in the first month, all of which ended up in their gallery — that was my signal that something was happening here worth taking seriously. And it's not just about being passionate about the general subject; a company should look for someone excited specifically about what that company does, not just excited about the broader field.
How has making YouTube videos changed for you since you started doing this professionally, during COVID?
During COVID there was a lot of demand and not much supply — if your video had reasonable quality and explained something clearly, it was pretty easy to get views. These days it feels like the opposite: too much supply, not enough demand, and I don't think there are enough eyeballs on the planet to give every video produced a thousand views. I do make sure I've got a solid introduction and a good thumbnail, but beyond that I try to keep editing minimal for myself — I've even programmed my keyboard for video editing. If a video doesn't get ten thousand views, that's also fine. Six percent growth a week, if we can maintain it, is solid. It's a marathon, not a sprint.
What do you think is the actually hard part of building a dev tools business?
There's a meme that first-time founders worry about implementation and second-time founders worry about distribution, and distribution really is hard — you can't just build it and expect people to come. But underneath that, the real question is whether you have something that actually moves the needle for someone's professional life. You need to genuinely convince yourself of that first. Once you have, the second question becomes: why doesn't anyone recognize that yet? Promotional activity is the exchange of enthusiasm, but you need something of real value before you can be excited about it in a way that transfers to other people.
Any shoutouts or recommendations for other DevRel people?
Don't sleep on Reddit — I'm active in the ergonomic keyboard subreddit, and even small, heavily-moderated communities will accept a video if it's genuinely relevant, which is a point of pride, not a wall to complain about. And I still love a good blog — a lot of blogs are turning into slop these days, and people can tell when something is written by hand versus generated. If you're a DevRel person, you can also just come up with the craziest demos yourself and try to break your own software, then report back to the engineering team — that's a kind of experimentation engineers usually don't have time for themselves, and they appreciate it.