For as long as software has existed, we’ve treated code as something carefully crafted by engineers: designed, typed out line by line, and managed in repositories and workflows built entirely for humans. That view is already outdated.
Today, code itself is becoming the foundation of generative AI. Every new capability, whether it’s a marketing dashboard, a workflow automation, or a feature triggered by a prompt, is ultimately code being written and deployed on demand. The text that fills a landing page, the chart created from a dataset, the button that appears when a user asks for a new action — all of it is generated as code.
This shift has enormous implications. When everything becomes code, and when code is generated by agents rather than humans, the bottleneck moves from writing it to running it. A world of prompt-driven software creation demands toolchains that don’t just produce snippets, but systems that can find the right context, merge changes seamlessly, execute safely, and scale instantly. Without that infrastructure, code generation is a parlor trick. With it, software becomes a living, generative fabric.
That is the world Relace is building toward.
Relace is creating the models and infrastructure that make coding agents production-ready out of the box. They’re rethinking the stack from first principles: small, specialized models trained specifically for coding tasks, coupled with infrastructure layers that surface the right context in seconds and apply edits at unprecedented speeds: their Instant Apply model merges edits at over 10,000 tokens per second, and their newly-released Repos feature allows agents to do high-frequency commits while making it simple to search, navigate and retrieve the context. Together, these capabilities offer a foundation where agents don’t just generate code — they integrate it and make it usable in the real world.
Already, Relace is powering more than forty companies working on code generation, and partners like Figma and Lovable are running their models in production today. These adopters are proving that once the backbone exists, new possibilities open up: designers spinning up interfaces, marketers generating custom workflows, professionals in every field creating software on demand without ever opening an IDE.
The timing could not be better. While massive foundation models have captured attention, the true breakthrough in code generation lies in smaller, specialized models paired with optimized deployment infrastructure. Advances in training, inference, and hardware make this possible today — but the pieces need to be assembled into a coherent platform. That is exactly what Relace is doing.
Preston and Eitan are two of the most relentless and brilliant builders we’ve met. They combine craftsmanship in training models with hard-earned expertise running infrastructure at scale, and they approach both with the kind of intuition and persistence that set the very best founders apart. It’s that rare blend of research depth and practical execution that makes Relace such a compelling team to back.
We are thrilled to lead Relace’s $23M Series A and join the board. Relace is building the foundation for a future where software is generated on demand — fast, reliable, and accessible to anyone. This is the beginning of a profound shift in how code gets created, and we are excited to build that future together. If you’re an engineer or researcher excited about building at the frontier of models and infrastructure, Relace is hiring — and it’s a great time to join.
-
Yoko Li is a partner at Andreessen Horowitz, where she focuses on enterprise and infrastructure.
-
Guido Appenzeller is an investor at Andreessen Horowitz, where he focuses on AI, infrastructure, open source technology, and silicon.
-
Martin Casado is a general partner at Andreessen Horowitz, where he leads the firm's $1.25 billion infrastructure practice.
