4.4
Announcing General Machines
We are announcing General Machines. The infrastructure problem for frontier AI is not well understood from the outside, but it is well understood by the teams building it. We exist to solve the infrastructure side of that problem: the data collection pipelines, the evaluation frameworks, and the measurement systems that make AI development faster and more principled.
Why We Exist
The models that will define the next decade of AI are being trained on datasets that still have to be built largely by hand, evaluated against benchmarks that weren’t designed for the tasks they’re being deployed on, and measured with tools that were built for a different generation of capabilities.
Our first two focus areas reflect where we believe infrastructure is most underbuilt relative to the pace of deployment.
The Agentic Web
AI systems are increasingly being deployed to navigate the commercial internet on behalf of users - to find products, compare prices, and complete purchases. There is almost no principled evaluation infrastructure for whether these systems actually work in the wild. We are building the trace datasets and evaluation benchmarks that let labs and product teams measure commerce agent behavior at scale, on real retail environments, not sandboxed replicas.
Physical AI
Robots and spatial AI systems require data that cannot be scraped from the internet or generated cheaply with a language model. Ego-centric video - first-person recordings of tasks performed in real environments - is the closest proxy to how a robot or spatial agent experiences the world. Collecting it at the quality and scale that frontier teams need requires a purpose-built pipeline. That is what we are building.
The Team
We are a small team. We are not trying to build the next foundation model. We are trying to make sure the teams that are building them have the data and measurement infrastructure to do it well.
If that problem is familiar to you from your own work, we would like to hear from you. Reach us at founders@generalmachines.ai.