An initial market of a few thousand engineers. And $2.2M to go after it.

Hermetiq raised $2.2M in seed funding on August 24 to fix the builds that AI-generated code keeps breaking. The round is well signposted. What is more interesting, and far less documented, is the distribution question sitting underneath it: how do you build an acquisition engine when your addressable market is a few tens of thousands of engineers?

I had the chance to speak with Jake Newfield, Hermetiq's CEO, about exactly that.

The problem underneath

Compiling code on a small project takes ten seconds. On a monorepo running into tens of millions of lines, the kind Google, Airbnb or SpaceX operate, a naive build would take hours. Bazel, the open source build system Google released, is what prevents that. It rebuilds only what changed, shares results across the team through a common cache, and distributes work across a fleet of machines.

When it works, a forty minute build drops to ninety seconds.

When it does not, nobody knows why. That gap is what Hermetiq sells into. And the failed build is not even the worst case. As Jake put it, engineers burn enormous amounts of time working out not just why a build failed, but why it was slow, and why it was not optimised. Cache misses, unexplained slowdowns, builds that break for no visible reason: three symptoms of the same underlying condition, which is that build systems at scale produce almost no usable evidence about their own behaviour.

Hermetiq turns build telemetry into that evidence. Which action failed, which cache key missed, whether you are queue bound or execution bound, where wall time actually goes along the critical path. The product also feeds that context to AI coding agents through an MCP plugin.

Why the timing works

The thesis fits in a sentence. Agentic coding tools produce far more code, far faster, with far less human review. So they produce far more broken builds.

Jake frames the market accordingly: enterprise first, but really anyone with a lot of code, because anyone with a lot of code hits a build bottleneck eventually. And more companies have more code every month, precisely because of agentic coding. The problem is compounding on its own.

It is an infrastructure bet rather than a model bet. While most capital chases the tools that write the code, Hermetiq is betting on what stops that code from reaching production.

Year one: Slack and LinkedIn DMs, in founder mode

This is the part I wanted to understand, and Jake was unusually direct about it.

The go to market for the entire first year ran on two channels. The public Bazel Slack channel, where he spent time talking to engineers about the problems they were hitting. And LinkedIn, not as a content play but as direct messages, sent one at a time, by him.

Founder mode in the literal sense: the CEO in the community, in the DMs, having the conversations himself.

It worked well enough to land a handful of reference customers, including one of the leading global AI labs now running the platform in production at more than two million builds a day, and well enough to raise. Jake describes the transition plainly: the first phase was about finding a few important customers who could prove out what they were building. With the round closed, the company is moving into an acceleration phase, and building the go to market machine that comes with it.

What he wants out of that phase is volume of conversation. He wants to talk to as many engineers using Bazel as he possibly can.

Users before buyers

This is the strategic choice I found most interesting, and the most counterintuitive.

Jake is completely unsentimental about how narrow his market is:

"If a company isn't using Bazel, I don't think there's any point in me reaching out to them - until we expand to other build systems."

How many people is that? By his own estimate, tens of thousands of engineers. Not millions. And he wants to reach every one of them.

Those engineers are the users, not the buyers, and that ordering is deliberate. He was heading to an event full of technology executives the week after we spoke, and still described users as the higher priority.

"We don't have a product that's going viral yet. We need to figure out how our product is working and how to market better. And the way we're going to do that is with users, not with the buyers."

There is a formula we use for this at uclic: users are the roadmap, executives are the buyers. Inverting that order at seed stage is one of the most expensive mistakes in developer tooling. You sell at the top, you sign a pilot, and six months later the engineers have not adopted the product because nobody ever asked them what they needed.

Choosing the harder, slower path here is the right call, and most founders under pressure to show a growth curve do not make it.

The next room to be in

The most immediate expression of that strategy is BazelCon, running October 13 to 15 in Amsterdam, where Hermetiq is a sponsor.

It is the same instinct that took Jake into the Bazel Slack channel in the first place: stay close to the community, in the room, talking to the people who actually run these builds every day. In a market whose total audience fits into tens of thousands of people, being present where they gather is not a marketing line item. It is continuity with how the company has operated from day one, at a moment when it finally has the means to do it at scale.

What the cap table says

The round was led by Jet Investment through its Jet Venture 1 fund, with $750,000, making it the largest single backer. But the angels are the more revealing part.

Hermetiq's very first investor was Niranjan Tulpule, VP of Developer AI at Google, which is about as targeted a validation as a build tooling company could hope for. Jason Calacanis followed through his LAUNCH fund, then Gokul Rajaram, who sits on the boards of Pinterest, Coinbase and The Trade Desk. Etsy is currently running a pilot.

Still, the cap table is not what stayed with me after the call. What stayed with me was a company that knows the precise size of its market, refuses to inflate it, and has decided to talk to its users before its buyers.

That choice produces no spectacular chart next quarter. On a narrow technical market, it is also close to the only thing that compounds.