Artificial intelligence has transformed how quickly software can be written. Tools such as Claude, Cursor, Codex, and GitHub Copilot allow developers to produce and modify enormous amounts of code, but generating code is only the beginning. Before software can reach users, it must be compiled, tested, verified, and converted into a working product. As code production accelerates, that build process is becoming a critical bottleneck. Jake Newfield founded Hermetiq to solve it.
“Creating code is no longer a bottleneck for organizations,” Newfield explained in an interview with FoundersPress. “The bottleneck for enterprises and AI companies is rapidly becoming the build process.” Software engineers may be able to generate exponentially more code with AI, but every change still has to pass through infrastructure that was designed for a slower development cycle. More code creates more builds, while increasingly complex codebases make those builds harder to understand, optimize, and repair when something goes wrong.
Newfield met Hermetiq co-founder Tim Potter at NativeLink, another company in the remote-build ecosystem. Potter served as CTO, while Newfield led sales and customer success. Their respective roles gave them a complementary view of the market. Potter understood the technical infrastructure required to execute builds at scale, while Newfield saw the operational problems facing customers. Together, they concluded that existing platforms could execute remote builds, but lacked an independent intelligence layer capable of analyzing information across the entire system.
To understand Hermetiq’s opportunity, it helps to understand what happens after a developer writes code. Large engineering organizations often rely on the Remote Execution API, an open protocol that allows build and test actions to run across distributed infrastructure. Bazel, the open-source build system originally developed at Google, is one of its most prominent client-side implementations. On the server side, platforms such as Buildbarn provide the infrastructure that schedules and executes those actions remotely. These systems help massive codebases build more efficiently, but they can generate an extraordinary amount of fragmented telemetry.
“What has not existed is a standalone intelligence and observability layer that looks into that system to provide insights,” Newfield said. He describes Hermetiq’s current product as “Datadog for build infrastructure,” or, more specifically, “Datadog for Bazel.” Instead of replacing the systems that execute builds, Hermetiq sits above them and explains what is happening across the build environment.
The platform currently provides observability across Bazel and Buildbarn. It ingests Bazel Build Event Protocol data, cache-hit and cache-miss events, completed remote actions, OpenTelemetry metrics, logs and traces, and cloud-cost data. Hermetiq then connects those signals to help engineering teams identify failed actions, understand why caches are missing, find critical-path bottlenecks, distinguish queue delays from execution problems, and determine where infrastructure spending is going. According to the company, Hermetiq already processes telemetry from more than two million builds per day for one of the world’s leading frontier AI companies. Etsy has also tested the platform through a pilot program.
The problem becomes more pronounced as codebases grow. Updating a simple website might trigger a build that finishes in 30 seconds. A frontier AI company, autonomous-vehicle developer, or robotics business may maintain millions of lines of code across several programming languages, all inside a single interconnected repository. Each component can depend on different libraries, tools, configurations, and open-source projects. A small change can unexpectedly invalidate a cache, introduce a dependency problem, slow remote execution, or cause the entire build to fail. Diagnosing the cause may require scarce specialists to sift through huge volumes of logs while other engineers wait.
Hermetiq’s long-term ambition extends beyond making that investigation faster. Newfield believes its data layer can provide AI coding agents with the context they need to diagnose and eventually repair their own failures. Hermetiq has already released an MCP integration that gives compatible agents structured access to build intelligence. The goal is a closed loop in which an agent writes code, observes why its build failed, identifies the root cause, makes the necessary correction, and verifies that the repaired software works.
“Hermetiq is the data layer, the intelligence layer that agents need and will use in order to diagnose and repair their own code,” Newfield said. If the company succeeds, builds could eventually become self-healing. Engineers would spend less time investigating broken pipelines, while AI agents would move closer to completing software rather than merely drafting it. In that future, the defining measure of an autonomous coding system would not be how much code it can produce, but how reliably it can turn that code into functioning software.
Hermetiq recently raised a $2.2 million financing round to advance that vision. The round included Jet Investment, LAUNCH, Jason Calacanis, Gokul Rajaram, Google Vice President Niranjan Tulpule, Lighthouse Seed Fund II, Chilli Ventures, and Gi21 Capital. Newfield said the funding is primarily being used to hire engineers and accelerate product development. The company has already added a full-time employee and several contractors, with another full-time hire planned.
Yet Newfield is clear that raising capital is only an early milestone. “We have an idea, and there’s some early indication that it’s a good idea,” he said. “Now we have to make some serious revenue.” His immediate focus is building the product, acquiring customers, and reaching the engineering leaders who already feel the pain Hermetiq is designed to address. The company is also sponsoring BazelCon, preparing new product capabilities, and developing a customer success story involving a client it can publicly name.
Newfield is now working to make build infrastructure understandable beyond the relatively small community of specialists who manage it today. Hermetiq’s premise is technical, but its implications are broad: AI can only automate software development if the systems surrounding generated code become intelligent as well. The models may write the code, but the build system still determines whether that code becomes a product.
As AI coding agents become more powerful, the distance between generating software and successfully shipping it will become increasingly important. Hermetiq is betting that the next major developer-infrastructure category will emerge inside that gap. By transforming fragmented build telemetry into intelligence that both engineers and agents can act on, Newfield and Potter are working toward a future in which every build can explain itself, every failure can be repaired, and AI-generated code can reliably reach production.
