Litmus
Research into filtering out low-quality, AI-generated contributions before a person has to review them. The plan is to build a tool around it.
Most experiments and projects stall because they lack structure, and the good parts never get a real chance. Here, each one earns its place. Before it grows beyond one place, it passes three tests:
There are three people. One is a generalist. One has a specific skill. The third can manage both the skill and the people.
If you only want things done and you pick the first, it feels like it's moving, but something unpredictable always pops up. If you give the second a set task, you get quality, but it feels slow. The third shares the opportunity fairly, so nobody takes from the others and everyone grows over the long term. With them, you move without fear.
Humane
A lot of AI tools swap the roles. You end up fixing their mistakes, and the job becomes supervising the tool.
We design it the other way. The tool takes the volume, like the tenth attempt at the same fix. A person keeps every decision that carries real risk. And it has to cost less than doing the work yourself. Otherwise it isn't worth shipping.
We judge every tool on how much it does, the risk it leaves with a person and what it costs to run.
Governance first
Open source is filling up with volume: generated pull requests and bots replying to bots. When everything gets through, it buries the people doing careful work.
So governance, for us, is mostly about protecting their attention, and about telling contributors plainly what they're signing up for.
Local first
The usual approach is to build for everyone at once and hope it fits somewhere. We go the other way. Every layer, from the hardware and software to the compute and the people around it, should work well for one place first.
Scale comes later, as lots of local setups linked together rather than one big centre everyone depends on.
Built this way, each piece keeps running when the network drops or a vendor shuts down.
Three projects at different stages. Each one started as an experiment and made it this far.
Research into filtering out low-quality, AI-generated contributions before a person has to review them. The plan is to build a tool around it.
A planner that lets you start writing before it asks you for anything. No account and no cloud sync. The AI runs on your own machine.
Helps people help each other with technology they already own, even in a blackout. The first 25 nodes are starting now.
Incubate
If your open source project or research needs governance and a fast start, tell us about it. We help write its rules and find its first contributors, and it leaves with its name.
We incubate opportunities that grow out of a real problem, not ones engineered to chase an outcome.
Or write to us directly.
hello[at]huyoworld[dot]com