Max Stoiber’s Path From Open Source Hits to OpenAI

Max Stoiber recounts building react-boilerplate and styled-components, the Spectrum and Stellate exits, and now shaping ChatGPT’s app platform.

MiHiR SEN
MiHiR SEN
·2 min read
Max Stoiber reflects on creating styled-components, react-boilerplate, Spectrum, and Stellate before joining OpenAI to work on ChatGPT’s app platform. He argues that open-source contribution remains valuable even when the code itself is not the final product, and that ChatGPT apps represent a new distribution surface for software.

A Career Built in Public

Max Stoiber has spent more than a decade shipping open-source projects that most developers eventually forget and a few that became staples. On the latest Changelog episode he traces the line from those early libraries to his current role at OpenAI, where he works on ChatGPT’s plugin directory and app platform.

The Hits and the Quiet Ones

Stoiber co-created styled-components, the CSS-in-JS library that still powers many React codebases, and react-boilerplate, an early starter kit that taught a generation how to structure modern front-end apps. He also built Spectrum, the community platform that GitHub later acquired and that influenced the design of GitHub Discussions. Later came Stellate, a GraphQL caching layer that was acquired in a dual deal by Shopify and The Guild.

Each project taught the same lesson: sharing code publicly compounds. Even the projects that never became household names still created the network that opened later doors.

ChatGPT Apps as a New Surface

At OpenAI, Stoiber sees ChatGPT apps as a fresh surface for software. Instead of another web or mobile container, the model itself becomes the host. Developers can expose tools and interfaces that the language model can call and present. The experience feels less like installing a plugin and more like extending the conversation itself.

Stoiber is careful not to oversell the present. The platform is still early. Yet the underlying idea, that any structured capability can be offered to a reasoning engine and then rendered back to a user, feels like a genuine shift in how software is distributed.

The episode is less a victory lap and more a map of how consistent open-source work can lead to the next generation of interfaces.