Where we explore, build, and find out what works.

Staying sharp takes practice. In Labs, we run experiments, ship internal products, and test ideas on our own terms, using the same AI-native approach we bring to every project.

It's how we develop real knowledge about what works, before a client project depends on it.

[Things we've built]

These are the products and tools we've built on our own terms. Some are public, some run internally, and all of them came out of a genuine curiosity about whether something could work.

Building for ourselves is how we stay grounded in what's real.

DrillRoom

OrangeLoops product

[iOS] [AI]

An AI-powered training app for pocket billiards players. DrillRoom uses computer vision to track shots and misses in real time, giving you feedback you'd normally only get from a coach. Choose from 50+ drills across skill levels, build a consistent practice routine, and follow your progress over time.

Learn more

DrillRoom app across tablet and phone screens

[Open source]

Build to share

When something we build is useful beyond us, we open it up. These are the projects we've made available for anyone to use, fork, or contribute to.

Harvest/Forecast MCP Server

[Python] [MCP] [API]

An MCP server that connects Claude, or any MCP-compatible client, directly to Harvest and Forecast, so you can query time tracking and project planning data without leaving your AI workflow.

Explore it