Late last year at Firecrawl, we were releasing a new open source example app every single week. Almost every one went viral on X. Hundreds of thousands of eyeballs on products built with Firecrawl, week after week. From the outside it looked like we had figured out developer marketing, but the numbers told a different story.
The idea was to open source full applications for popular use cases so anyone who wanted a similar workflow could get started quickly. Following big product launches with an open source version people could actually use became our signature move. I put one engineer on it full time, shipping a complete app every seven days. It worked in the ways you could see. Our open source website builder, Open Lovable, crossed 20,000 stars on GitHub. Open Agent Builder and Open Scouts each pulled in thousands more. Posts about these projects were racking up hundreds of thousands of impressions, and across all our repos we crossed 100k stars. People were talking about Firecrawl constantly.
But our main KPI on the marketing team was signups. When the numbers got an honest look, a super viral example would boost signups for a week and then they'd fall right back down the next. It wasn't compounding the way it should have. It all felt more like a treadmill than a flywheel.
The more subtle problem was brand dilution. Because we were launching so many standalone apps, people started getting confused about what Firecrawl actually was. "Are you competing with Lovable?" "Is this like Scouts?" Even our investors were confused, which led to some awkward conversations. We tried posting from founder accounts to separate things, but we were all too associated with Firecrawl for that to work.
At one point the team was shipping more open source examples than actual product, and something about that felt wrong. The engineer behind all those examples was a beast, and when it came down to it, a great product has way more leverage than even the most viral open source example. Moving him to product was one of the clearest resource allocation decisions we've made.
Stopping something that feels like it's working is hard. All that attention and all those stars were real. But the best resources on the team were going toward something that gave bursts instead of compounding growth. At a startup you have to kill good things to make room for great things. Focus is everything when your resources are limited, and every person on a strategy that isn't compounding is a person not on one that could be.
So the question became what would actually move signups and compound over time, and we landed on two bets alongside product marketing. One was SEO and GEO, because more people are using AI to write code and that means more people becoming software creators. If Firecrawl is the number one recommended web data API across AI tools and coding agents, signups grow as those platforms grow. More content builds more authority which drives more recommendations. Beyond content, we started pursuing integration partnerships with platforms whose users already need web data. Our recent integration with Lovable is a good example, where Firecrawl comes free by default and users convert once they hit the limit. This ties our growth to the growth of other platforms. Product marketing ties it all together, because constant shipping and sharing with quality builds brand trust that compounds through word of mouth and mindshare in ways that viral moments just don't.
Since making these changes about three months ago, our weekly signups have doubled, driven by strategies that build on themselves week over week. Looking back, the open source examples were the right play for the stage we were in, when we were still establishing ourselves and needed raw momentum. But staying on that path would have meant optimizing for attention instead of growth. I think a lot of startups get stuck here, doing the thing that feels like it's working because the attention is real, when the harder and better move is reallocating those resources to what compounds.