Minitap raises $4.1M to fix slow AI mobile development
Mobile teams move slower than they should. Even with strong AI coding assistants, shipping on iOS and Android drags while web teams push features daily. Minitap says it can change that pace, and it just raised $4.1M to prove it.
The San Francisco startup's seed round was co-led by Moxxie Ventures and Mercuri. The 23-year-old founders, Nico Dehandschoewercker and Luc Mahoux-Nakamura, recently claimed the #1 spot on the AndroidWorld benchmark for AI control of mobile devices-topping research groups from Google DeepMind, Microsoft, and ByteDance within 40 days of starting the project.
Why mobile still lags despite smarter AI
LLMs like Claude and tools like Cursor sped up web work, but mobile is a different beast. Testing across thousands of device configurations, hardware quirks, and OS versions is hard to automate end to end. When bugs show up on real devices, most AI agents stall.
That's why companies such as Duolingo and Hinge reportedly run five times more experiments on the web than on mobile. As CEO Nicolas Dehandschoewercker put it: "Mobile is 60% of internet usage but moves at 10% of web speed." Teams that run 10x more experiments learn faster and win more often.
What Minitap built
Minitap is pitching two core pieces that bring agentic workflows to phones:
- mobile-use: an open-source framework that lets AI agents control mobile devices like real users-taps, swipes, inputs, and flows.
- minitap cloud: on-demand infrastructure to spin up any iOS or Android configuration across thousands of devices in parallel.
Together, they form a closed loop: an AI coding environment writes app code, tests on real devices, detects bugs, fixes itself, and ships working features. The goal is to cut a typical six-week feature cycle down to days.
Proof points and momentum
The funding round drew attention from the AI infra crowd, including Thomas Wolf (Hugging Face), Petter Made (SumUp), and operators from OpenAI and Google DeepMind. That interest tracks with the team's benchmark performance and the clear pain mobile orgs feel today.
Near term, Minitap wants product managers to describe a feature, attach a Figma design, and have the system generate, test, and ship an A/B test in a single afternoon. Longer term, the vision is mobile apps that optimize themselves automatically by running continuous experiments and iterating on live signals-without human intervention.
Why this matters for engineering and product leaders
- Throughput: If autonomous testing on real devices holds up, your team can attempt far more experiments with the same headcount.
- Feedback speed: Faster test cycles mean you validate UX hypotheses in days, not sprints.
- Coverage: Parallel device matrices reduce the "works on my phone" problem that stalls releases.
- Cost profile: Infra spend may rise, but savings in manual QA, flaky test maintenance, and rollbacks can offset it.
What to ask before you try it
- Guardrails: How are production credentials, destructive actions, and PII locked down during autonomous runs?
- Determinism: How does the system handle flakiness in UI flows, network jitter, and animations across OEM skins?
- Observability: What traces, session replays, and logs can engineers inspect when the agent "fixes itself"?
- Rollback and approvals: Can you enforce human-in-the-loop gates for shipping, especially for payments and auth?
- Integration: Does it slot into your CI/CD (Fastlane, Gradle, Xcode Cloud), feature flags, and A/B platform?
How to pilot with minimal risk
- Start with a sandbox: Give the agent a non-critical flow (e.g., onboarding variant) and fake backends.
- Constrain surface area: Limit permissions and scope to a single feature flag and a small device matrix.
- Benchmark honestly: Compare cycle time, test coverage, bug escape rate, and engineer hours pre/post.
- Track costs: Monitor device-hour burn vs. QA savings and rollback reduction.
If it works, your workflow changes
PMs write crisp specs and attach a Figma file. The agent scaffolds code, runs it across device pools, patches failures, and proposes a ship-ready branch with experiment configs. Engineers shift from writing glue code and brittle UI tests to setting guardrails, reviewing diffs, and curating experiment design.
You get more shots on goal with tighter feedback loops. That's the real advantage.
The bottom line
Minitap is making a direct claim: autonomous agents can make mobile shipping 10x faster by testing and iterating on real devices at scale. The team's early benchmark win and new funding give them room to test that claim in production settings. If they deliver, mobile orgs won't need to mirror web velocity-they'll meet it.
If you're upskilling your team on AI-assisted coding and agent workflows, explore practical training paths here: AI Certification for Coding.
Your membership also unlocks:
