From React boilerplate to ChatGPT apps
Max Stoiber has spent more than a decade shipping open source software that the rest of the React ecosystem quietly depends on. The Changelog conversation with him walks through the long arc: hundreds of small projects nobody remembers, the big ones (react-boilerplate, styled-components) that changed how teams wrote React, and the companies that came out of those communities.
The interesting part isn't any single library. It's the pattern. Each step grew out of the previous one. Stellate, the GraphQL caching layer Stoiber co-founded, emerged from years of running into the same edge-caching and persistence problems at scale. The Shopify acquisition earlier in 2024, and a follow-on deal with The Guild, validated the cache-as-product thesis. Two companies bought the same asset for different reasons: Shopify wanted the runtime, The Guild wanted the team.
Spectrum, GitHub Discussions, and what gets absorbed
Earlier in the journey, Stoiber built Spectrum, a community platform that GitHub acquired in 2018. Spectrum didn't survive as a standalone product, but its DNA ended up in GitHub Discussions, which launched in 2020 and now sits at the center of how OSS maintainers talk to users.
The story captures something real about the open source economy. Most projects disappear. A few get absorbed. A vanishingly small number become the default. The ones that win are usually not the cleverest technically; they're the ones that match what a large platform needs at the moment a platform is ready to ship.
ChatGPT apps as a new surface
Stoiber's current role sits him inside the team building ChatGPT's plugin directory and app platform. His framing of it is worth quoting: ChatGPT apps feel like a new surface for software. Not a new distribution channel, not a new app store, but a surface, meaning a new way for the same user intent to land in front of a piece of code.
If that framing is right, the winners in the next few years won't be the teams with the best prompts. They'll be the teams that figure out how a conversational surface should handle long-running state, permissions, and the boring plumbing that desktop apps spent thirty years working out. The plugin directory is the easy part. The hard part is what happens after.
The lesson underneath the projects
Stoiber keeps saying the same thing in different ways across the conversation: even if the value isn't in writing code, it's still worth it to share things with the world. The hundreds of forgotten projects were not wasted time. They were the training set for the ones that landed.
That is a useful reminder for anyone staring at a side project that no one is using. The asset isn't the repo. The asset is the developer who shipped the repo and learned what didn't work.