Releasing an AI product as open source isn't an obvious choice. You're giving competitors a head start. You're accepting that someone will fork your codebase and compete with you directly. You're choosing community over control.
We did it anyway. Here's why.
The decision
When we were deciding on a license, we considered three options: proprietary (standard SaaS), permissive open source (MIT/Apache), and the Unlicense (public domain). We chose the Unlicense because it has zero ambiguity: anyone can do anything with the code, forever, with no attribution required.
The reasoning was partly philosophical and partly strategic. Philosophically, we believe that AI tools should be inspectable and forkable. If Tripplet's entire business model depended on keeping our UI hidden, that would be a sign that we don't have a compelling enough product. Our actual moat is the models, the infrastructure, and the team — not the Next.js frontend.
How it's worked out
Six weeks after publishing the repo, we have 2,400 GitHub stars and 47 external contributors. Three of those contributors have submitted PRs that we merged directly into production. One — a developer who wanted better keyboard shortcuts in the Code workspace — now has their code running for every Tripplet user.
We've also received detailed bug reports and feature requests from developers who read the source code to understand how something worked. This level of feedback quality is impossible to get with a black-box product.
The trade-offs
Has anyone forked Tripplet and tried to compete? Yes — we've seen at least three forks that are building commercial products on top of our UI. We're fine with this. If they succeed, it validates the space. If they build something better, we learn from them.
What we keep proprietary: the model weights, the inference infrastructure, and the training pipelines. The application code is open. The intelligence is ours.