Open AI as a Public Good
Published
Most of the world's AI is built inside a handful of data centers, owned by a few companies. Tendril is built on a different idea: that AI can be powered by millions of ordinary people, and the data they create can be a public good rather than a private asset. This essay makes the case for open AI, and how everyday contribution makes it real.
Most of the world’s AI is built inside a handful of data centers, owned by a few companies, in a few countries. The compute is concentrated, the training data is private, and the resulting models are products you rent. That arrangement has produced impressive systems. It has also quietly decided who AI is built to serve, and who it is not. Tendril is built on a different idea: that the data behind AI can be a public good, created by the people it is meant to help.
The difference between a product and a public good
A private asset is something one owner controls and can sell, restrict, or shut off. A public good is something many people can use without using it up, like a shared map or an open standard. Most AI training data today sits firmly in the first category. It is collected privately, kept private, and converted into a product.
Human-feedback data does not have to work that way. A record of which AI answers real people prefer is exactly the kind of thing that becomes more valuable the more openly it is shared. Open it up and many builders can improve many models with it, including models for communities that no single company finds profitable enough to bother with. Keep it locked away and that improvement only happens where it pays. The choice between those two worlds is not technical. It is a choice about who AI is for.
Why open matters in practice
When the data is open, the benefit is not confined to one company’s roadmap. A researcher working on a long-neglected language can use it. A small team building a tool for farmers, students, or clinics can use it. The communities that supplied the judgments can see what was collected and how. Open data invites scrutiny, which is also how it stays honest: provenance and licensing can be checked rather than taken on faith.
This is the heart of what Tendril is trying to make real. Not another private model, but the open human-feedback data underneath it, built in a way that anyone can inspect, use, and build on.
How everyday contribution makes it real
A public good needs a public to build it, and Tendril gives people two simple ways in. You can open a browser tab and let your device do a small amount of AI work in the background, using only spare capacity and stepping aside the moment you need your machine. Or you can pick up your phone, look at two AI answers, and tap the one that is better. Both turn ordinary spare moments into a contribution to something shared.
Those two surfaces matter most for the languages and communities the market tends to skip. The judgments people make become open data that can be used to build AI that understands them, not just the languages that were already well served. You can see which languages this focuses on, and how to help, on the languages page. Buyers who want to put this data to work, or commission more of it, can start at the datasets page.
For each contribution you earn TEND Points. They are a way to recognize and account for what you put in. TEND Points are not money, and we never promise payouts. The real return is different: a piece of the world’s AI that was built in the open, by the people it is supposed to serve, rather than handed down from a warehouse. If that is the future you want, open a tab or start tapping, and help build it.