Five years ago, Jon Sakoda, founder of Decibel Ventures, said in a post on X that one should never underestimate the power of small groups. He added that a group of five people with a shared purpose can out-execute 500 people with competing priorities.
“Small teams generally move faster and more easily navigate a dynamically changing world,” he said.
The statement has never been more relevant than it is today, in 2025.
AIM recently reported how the AI-enabled coding platform Cursor, not only crossed $100 million in annual recurring revenue (ARR) but also achieved it in record time with 20 employees. Cursor’s case isn’t an isolated one, several other startups have reaped the benefits of existing in an AI-first landscape.
Last year, MidJourney, an AI image generator, reached $200 million in ARR with just 11 employees. Similarly, Bolt.new, another AI coding platform, has crossed $30 million in ARR with just 20 employees, in just a little over 4 months and has registered 3 million users.
Lovable, a Cursor competitor, has reached $10 million in ARR with 15 employees in just 60 days. Similarly, ElevenLabs, an AI text-to-voice startup, has reportedly reached ARR figures of nearly $90 million with 50 employees.
How did we get here?
High Risk, High Rewards
In a conversation with Dan Shipper, co-founder of Every, Mike Maples, co-founder of Floodgate Ventures, attributed it to the difference in the risk profiles with which small startups and big companies operate.
He cited an example of how Elon Musk’s SpaceX can launch something into space, and if it blows up, the team can quickly get to making a better one. “NASA is not going to do that,” said Maples. “If NASA launches a rocket, they don’t sit there and say ‘easy come, easy go’ it blew up,” he added.
He then said that the difference in the risk profile associated with Musk, and SpaceX, for example, can change the speed with which he can move. The situation is similar for many small teams building AI products that may seem to compete with the bigger names in the competition.
“Not having to be burdened by what could go wrong is a big factor in trying things that could go right,” said Maples.
Both Shipper and Maples indicated that startups will, and should focus on a “niche that incumbents won’t get into” and ideas they don’t want to do themselves, and can’t conceive of.
Moreover, Maples indicated that specialised domain expertise can create defensible positions where large companies cannot easily compete with their generic AI offerings– especially in complex, multi-disciplinary fields.
Lovable’s Playbook
Take Lovable, for example. The startup was launched only 30 months ago, and it has 300,000 monthly active users, with 10% paying users.
In a podcast episode, Anton Osika, founder and CEO of Lovable, spoke about the origin story of Lovable, which may have a few insights on how to scale a small team startup fast.
The startup began by releasing an open-source tool called GPT Engineer on GitHub, which rapidly gained popularity, amassing over fifty thousand stars. This initial success helped Lovable build an early user base that valued their product, ultimately playing a crucial role in driving the company’s growth.
Lovable was then launched under a waitlist to iterate based on user feedback. When the product got ‘really good’, it was launched for more users. Given that Lovable is an AI system building code for products, it tends to get stuck in the process. “What we did was to painstakingly identify places where it got stuck,” said Osika.
The team had to “tune the entire system quantitatively and have a fast feedback loop to improve it in the areas where it got stuck”. Processes like these are crucial to ship a product that is both usable, and functional.
Osika attributes the success of Lovable mainly to the existence of foundational models, which he says is akin to how the discovery of oil gave birth to many products and industries. He said that Lovable’s job was to “obsess” over presenting the foundational model in the right way to the user.
Second, he also emphasised the idea of building in public. He said that to gain insights, awareness and feedback on the product, the company regularly posts on social media about everything that is in the works.
Osika credits its team that can ship ‘really fast’ and have good taste for the right abstractions along with an obsession to make the product better and better. He also said that Lovable strives to hire people with the “absolute superpower” in some dimension to have a generalist brain.
And the growth story doesn’t seem to stop. Osika recently revealed that Lovable grew 50% in a single week, and is now adding 1,500 new customers per day.
“Word of mouth growth is picking up. I’ve had a lot of people tell me how many people they have shown Lovable to recently – this is an exponential driver,” he said.
Using AI as a Force Multiplier
Interestingly, Osika shared that the startup has set up a version of Lovable to improve its own capabilities. Building AI with AI has been another crucial factor in the success of these startups.
“Everyone uses AI all the time in writing code,” he said, adding that it is great for experimentation.
Osika also said that almost everyone in his team uses Cursor to write code. Several industry leaders also believe in using AI as a force multiplier for growth.
Micheal Mignano, a partner at Lightspeed Ventures, echoed a similar sentiment in a post on X. “They [high growth, small team startups] all use the best AI tools to get leverage in every part of their workflow,” said Mignano.
“And they work together in real life, most sitting directly next to each other (to make communication easier) while executing against very strict goals for what they need to ship and by when. Just hardcore, hyperfast building, all day,” added Mignano.
This might just be the beginning of the journey. As long as frontier labs don’t specialise in every use case possible, there will always be room for more such startups.