Seven of the last ten people we've brought on at Firecrawl have been on the go-to-market side and at the same time we're hiring slower on engineering. A year ago that would have seemed insane for an AI infrastructure company like ours.
The reason is pretty simple. A growing amount of our code is now AI-written. We use Devin directly in Slack and even built an internal tool called Spring that's a remote Claude Code instance connected to all of our codebases. In about six months, Devin and Spring combined have become the second highest contributor of code to our web app behind my co-founder and CTO Nick, with over 500 commits.
We're not unique here. A quarter of the startups in Y Combinator's Winter 2025 batch shipped products where 95% of the code was written by AI. This is happening everywhere.
When everyone can build, the thing that used to be rare (engineering talent) stops being the bottleneck. Something else takes its place.
The cost of building is dropping to near zero, and what you get is an explosion of apps and tools solving overlapping problems. The supply side of software is going to become completely saturated. Average SaaS customer acquisition costs were already rising roughly 14% in 2024, and that's only going to accelerate as more products hit the market.
The default assumption in tech has always been that engineering is the hard part (and it still is in some cases). You build the product, then maybe you hire some marketing people to talk about it but that's flipping. When you can clone an app in a day, the technology itself stops being the moat.
DocuSign is a good example of where this is going. The actual technology behind e-signatures is not particularly complex. Anyone could build a DocuSign competitor in a weekend now, and hundreds probably already have. But DocuSign has the enterprise contracts, the brand trust, the integrations, the muscle memory of millions of people who just associate "sign a document" with that company. I'm not saying DocuSign is going to win, but they'd improve their odds by shrinking engineering faster than sales in the next round of cuts. The moat is the relationships and brand, not the code.
You can build a v0 really quickly now, but then how do you drive people to it? That's the real skill. It's going to require really smart people working on ways to get products in front of customers through new discovery engines like AI agents, ChatGPT, Social Media, and whatever comes next.
To be clear, AI is eating go-to-market too. Vercel replaced its 10-person inbound SDR team with a single employee and an AI agent in six weeks and held their conversion rate flat. AI multiplies everything, engineering and marketing alike. The caveat on both sides is taste. The engineer behind the agent needs to know what good code and a good user experience actually look like. The marketer needs to know what's actually going to resonate with the target market, and more and more, they need to know how to build too. The marketers who can ship a landing page in Claude Code or wire up their own automations on top of doing traditional GTM are becoming absurdly leveraged.
I think what this all leads to is smaller companies with fewer people, all highly paid, all leveraging AI agents to get a ton of work out. Both teams shrink but get more efficient. The difference is that right now, significantly more is being automated on the engineering side. The high-leverage work in marketing (brand building, figuring out how to gain the algorithm on new platforms, building ecosystem partnerships, earning trust with customers) is still deeply human, still requires taste, and is harder to automate. And as building gets cheaper and more products flood the market, that work is only getting more valuable.
We're doubling down on this at Firecrawl. We're investing heavily in brand, integrations, partnerships, and building relationships that entrench us in the ecosystem and in people's applications. It's shaped who we hire too. The PMM we just brought on ships full web pages with Claude Code, does their own design work in Figma, and handles traditional product marketing. A few of our engineers help the GTM team build more complex automations, but that's partly because we sell a developer tool to developers and I wouldn't assume it generalizes. The bigger point is that marketers need to close that gap from their side, and the tools now exist to let them. The biggest problem we're solving right now isn't technical. It's making sure that humans and agents choose Firecrawl over the alternatives that could do something similar.
The big engineering salaries of the past decade aren't going away, especially for exceptional engineers solving genuinely hard problems. But the skill of selling, marketing, positioning, and getting people to actually use your product is going to be one of the most valuable skills of the next decade. I think the marketer that wins is less of a campaign manager and more of a strategist and agent orchestrator, someone who sets the goals, designs the channel experiments, and directs agents to handle the execution. It's why I feel lucky to be leading marketing at Firecrawl right now. I'm learning more about what it takes to win attention in a crowded market than I ever could have in a world where building was the bottleneck. The companies that figure this out first, the ones that invest in high-leverage GTM talent the way they used to invest in engineering, are going to have a massive advantage in a market that's about to get very, very crowded.