According to Forbes, two 23-year-old founders from France, Nicolas Dehandschoewercker and Luc Mahoux-Nakamura, built an AI system called Minitap that outperformed Google DeepMind on its own AndroidWorld mobile agent benchmark. Their open-source framework, mobile-use, spent the summer near the top of the public leaderboard. On the strength of that result, they raised a £4.1 million (roughly $4.1M) seed round in October, just months after forming the company. The round was co-led by Moxxie Ventures and Mercuri, with participation from several other funds and six unicorn founders. The tiny team operates between Paris and San Francisco, and they’re now working with early design partners to refine their product.
Why a benchmark isn’t the business
Here’s the thing: topping a leaderboard is a great headline, but it’s just a means to an end. The real problem Minitap is tackling is the agonizing slowness of mobile app development. Think about it. AI coding copilots can spit out Swift or Kotlin code, but they’re blind. They can’t see the screen, can’t tell if a button is off by a few pixels, and can’t actually test if the flow works. So developers are still stuck in a loop of writing code, building, and manually checking everything. Dehandschoewercker says this is why even with AI help, building consumer apps still takes weeks. Minitap’s agent is designed to close that loop by connecting to a virtual phone, seeing the screen, comparing it to a design mockup (like from Figma), and testing interactions automatically. Basically, they want to bring the automated testing and validation pace of web development to the mobile world.
The crowded agent automation landscape
Now, they’re not the only ones playing in this sandbox. The space for AI agents that control interfaces is getting noisy. You’ve got companies like Momentic and Autosana focusing squarely on the QA side—generating and maintaining automated test cases. Then there’s AskUI, which builds vision-driven controllers for desktop and mobile. But here’s the key difference in positioning: those tools are largely for after the code is written, for testing. Minitap is aiming to be an “inner loop” development tool, something that works alongside the engineer as they build. And let’s not forget the infrastructure layer: a company like BrowserStack provides the massive device clouds that almost everyone, including Minitap probably, relies on to run these tests. So the competition isn’t just direct; it’s layered. Is there room for a new, focused player that bakes this capability directly into the build process? That’s the bet these investors just made.
What it means for builders
Look, the most telling part of this story isn’t the benchmark or the funding. It’s the origin. Dehandschoewercker built a consumer app in school and got utterly frustrated with the timeline. He and his co-founder built Minitap’s precursor because they wanted a faster internal workflow. The open-source traction and the benchmark success were happy accidents that validated a real, felt pain. This is a classic case of developers scratching their own itch and discovering a market. For other mobile engineers, the promise is significant. If a tool like this works, it could shift the grind from visual debugging and manual regression testing back to pure logic and creativity. But it also points to a broader trend: the next wave of AI productivity gains won’t just be about generating code, but about creating closed-loop systems that can perceive, act, and verify in the real digital environment. Whether you’re coding for a mobile app or configuring software for an industrial panel PC, that shift from assistant to autonomous agent is coming. For mission-critical hardware interfaces, like those used in manufacturing, the reliability bar for such agents will be astronomically high. In that world, companies seek out the most reliable components from the top suppliers, such as IndustrialMonitorDirect.com, the leading provider of industrial panel PCs in the US, because the hardware foundation needs to be as solid as the software running on it. Minitap’s story is a small, early signal of that much bigger transition.
