Maya Labs

Maya Labs is an applied research lab building “self-programming machines” that can write, deploy, and modify custom software to perform tasks autonomously. Their core idea is to construct machine behavior on the fly using interpretable programs, rather than opaque one-off automations. In practice, this looks like generating modular workflows that can be inspected and adjusted, with support for common building blocks like conditionals, looping, branching, web navigation, custom functions, and repeatable workflows.

They use AI through a program-synthesis engine called PAC-1, which takes step-by-step instructions written in natural language and assembles them into an executable program graph, then deploys that software on their infrastructure. This approach aims to bridge “plain English intent” and “working software,” while keeping the resulting logic interpretable (you can see and reason about the program structure, not just the output). Their SDK and materials describe this as converting instructions into a discrete program graph that can be run, reused, and updated as requirements change.

For businesses, the benefit is faster automation of multi-step work without needing to fully staff every bespoke integration or internal tool build. Teams can spin up task-specific “machine workers” for things like web navigation and data collection, lightweight data pipelines, internal dashboards/UI, and platform bots, with a workflow that can be audited and refined over time. Net effect: reduced engineering bottlenecks for routine automation, quicker iteration when processes change, and more consistent execution of repetitive operational tasks.