Lovable posted its February numbers last week and the figure that stuck in my head was not the $100M in monthly revenue, although that is absurd for a company most people in tech had barely heard of six months ago. It was the 146. One hundred and forty-six employees generating that kind of number, which works out to roughly $685,000 in monthly revenue per person, a figure that would make most SaaS founders quietly close their laptops and stare at the ceiling for a while.

Nobody won by having the best model

The thing that Lovable got right, and that I think a lot of AI startups are getting catastrophically wrong, is that they did not try to build a better foundation model or even a particularly novel AI capability. They took existing LLMs and dropped them directly into a browser-based workflow where people were already trying to build things. You open Lovable, describe what you want, and it builds a working web app right there in your browser tab. No CLI. No local setup. No switching between a chat window and a code editor while you manually shuttle context back and forth like some kind of human middleware.

This is the same pattern Cursor exploited in the developer tools space, the same one that keeps showing up whenever an AI product actually scales instead of just accumulating Twitter hype. The companies winning are not the ones with the most impressive demos or the highest benchmark scores. They are the ones that figured out where their users already spend time and embedded the AI directly into that environment.

The distribution gap is enormous

And yet most AI companies are still building destination products. They want you to come to their website, log into their platform, learn their interface. Every one of those steps is a leak in the funnel, a place where someone who would have gladly used the tool instead closes the tab and goes back to doing the thing manually. I saw Jason Lemkin reference a stat recently that the average worker uses something like eleven different SaaS tools per day, and the last thing any of them want is a twelfth one that requires onboarding.

Lovable understood this intuitively. A browser tab is where you already are. You do not have to install anything, configure a development environment, or convince your IT department to approve a new tool. The AI meets you in the browser because that is where the work happens for most knowledge workers, whether they are writing code or filling out procurement forms or comparing vendor pricing across six open tabs.

$685K per employee is not an anomaly

What makes the Lovable numbers genuinely interesting rather than just impressive is that they are consistent with a broader pattern. Cursor reached $100M ARR with under 50 people. Midjourney was supposedly doing $200M+ with around 40. Perplexity scaled to millions of users with a tiny team. These are not flukes. They represent what happens when distribution is so efficient that you do not need a massive sales org or customer success team to push adoption forward because the product lives where the users already work.

The old enterprise sales model assumed you needed bodies in seats to close deals, manage accounts, handle support tickets. But when your AI product is embedded in the user’s existing workflow, a huge chunk of that overhead evaporates. Onboarding friction drops. Support volume drops because users are not confused by a separate interface. Retention goes up because the tool is woven into daily habits rather than sitting as a standalone app that people forget to open after the first week.

Browser-native is the quiet winner

I keep coming back to the browser specifically because it is the one environment that almost every knowledge worker inhabits for most of their working day, and the AI tools that operate inside it have a structural advantage that is hard to replicate. A browser agent like Dassi follows this exact logic. It does not ask you to leave your browser, copy context into a chat window, and then manually apply whatever the AI suggests. It sits in the side panel, sees the page you are on, and acts on it directly.

The Lovable story is compelling because it validates a thesis that should be obvious but that the AI industry keeps ignoring in favor of capability races: people adopt tools that show up where they already are. Building a marginally better model matters less than building a product that intercepts the user at the moment they need help.

Lovable did it for app building. Cursor did it for code editing. The browser — where most of the world’s actual work gets done — is still wide open. And the companies that figure out how to embed AI there without asking users to change a damn thing about their workflow are going to look a lot like Lovable’s revenue chart.