Two brothers who worked at Dropbox and Palantir raised $15M because AI writes code faster than anyone can check it

[ECWAlex98]

London based Meticulous has raised $15 million in a Series A led by Chemistry, with Menlo Ventures and a stacked angel bench including Guillermo Rauch, Arash Ferdowsi and Scott Belsky joining in. The company's pitch is blunt, AI coding agents now generate software faster than any human team can review it, so Meticulous automatically tests every code change before it ships rather than relying on developers to write and maintain thousands of brittle manual tests

Founders Gabriel and Quentin Spencer-Harper are brothers who each independently ran into the same wall at very different companies. Gabriel was a software engineer at Dropbox, Quentin spent a decade leading frontend engineering on Palantir's Foundry product, and both kept seeing the same lesson play out, you can test code exhaustively and it still breaks the moment real users actually touch it. Meticulous works by scanning an application's codebase, identifying the edge cases a given change could break, and simulating real user sessions against the update both before and after, handing developers a visual diff instead of a wall of raw test output. It's frontend only for now, with backend and full stack validation planned next

The company says its annual recurring revenue is up fivefold over the past year, with customers including Notion, ElevenLabs, Dropbox, Wiz and LaunchDarkly. It's a crowded fight though, Harness has pushed into AI powered testing at a valuation above $3.7 billion, Diffblue has raised over $50 million automating Java unit tests specifically, and CodeRabbit and Qodo are baking AI directly into code review itself. What differentiates Meticulous is scope, it isn't trying to write the code or review the pull request, just verify whether the thing actually works before it goes live, positioning itself as a trust layer sitting between AI generated code and production

The new funding goes toward R&D, backend and performance testing, and growing the team from around 20 people to somewhere between 30 and 40 over the next year, bringing the company's total raised to $19 million. As AI agents write an increasing share of production software, the underlying bet here is that developers will spend less time writing code and more time deciding whether they can actually trust it, meaning the companies verifying AI generated software could end up mattering just as much as the ones generating it in the first place

AI writes code faster than developers can review it. These ex-Dropbox and Palantir brothers raised $15M to fix that - TFNtechfundingnews.com

[Firewall Stephen]

Both founders independently hitting the same wall at completely different companies, Dropbox and Palantir, before teaming up is a strong origin story for exactly this problem

[Rosie_81]

The shift from developers writing code to developers deciding whether they can trust code is probably the more accurate description of where this whole industry is actually heading

[WarpField69]

A visual diff instead of a wall of raw test output sounds like such a small UX choice but it's probably the difference between developers actually using this daily versus ignoring another dashboard

[Carol84]

This space getting this crowded this fast, Harness, Diffblue, CodeRabbit, Qodo, all chasing slightly different angles on the same underlying problem, tells you how real the pain point actually is

[WWFRoss95]

Frontend only for now with backend coming later is a sensible way to scope an MVP, better to nail one surface area convincingly than spread thin across the whole stack immediately

[Odd Maverick]

Fivefold ARR growth with customers like Notion and Dropbox already on board is a strong signal this isn't just solving a theoretical problem, real engineering teams are actually paying for it

[Inlet]

That core idea checks out. AI can generate large amounts of code quickly, but verification has not sped up at the same rate.

So the bottleneck shifts from writing to reviewing.

Tools that focus on catching issues early in the frontend could save teams a lot of time.

Especially for UI regressions that are easy to miss.

[EventHorizon47]

Frontend-first is a smart wedge. Visual bugs are easier to demonstrate and quantify.

You can show a before-and-after diff, highlight layout shifts, broken interactions, etc.

That makes the value proposition clearer than something abstract like backend correctness

[WarningPoint49]

Feels like a modern take on automated QA, but with AI awareness baked in.

Traditional testing assumes humans write the code.

Now the tool has to anticipate patterns generated by models, which can be subtly different.

That is an interesting shift.

[Midnight Georgia]

Backend support will be much harder. Logic bugs, race conditions, and data consistency issues are less visible than UI glitches.

So starting with frontend is not just simpler, it is strategically smart.

Build credibility first, then expand.

[ThatGirl]

Developers might resist at first if it feels like another layer of friction.

But if it saves time during debugging or prevents production issues, adoption will follow.

Tools that prove their worth tend to stick around quickly

[Oliver85]

This also ties into the broader trend of shifting left in development.

Catch issues earlier, reduce costly fixes later.

AI-generated code just makes that more urgent.

Otherwise you end up debugging a flood of machine-written logic.

[Sophie83]

One challenge will be false positives. If the tool flags too many non-issues, developers will start ignoring it.

Striking the right balance between sensitivity and usefulness is critical.

Seen plenty of testing tools fail on that front.