GoodWaves was a positive-only social network, open to people 16 and up. One rule held the whole thing together: only what lifts people up gets through. Not censorship — a design principle, enforced not by a thousand human moderators, but by a machine taught to recognise good.
A social space where the default wasn't outrage or noise — it was encouragement. The rule was the product.
Every contribution was read for the lift it carried, not the reaction it provoked. We measured the thing that actually mattered.
A working stack and a shipping iOS app. Real software, not a pitch deck — which is exactly why it taught us so much.
To keep a network kind at scale, no single model is enough. So we built a seven-step agentic pipeline — a chain of specialised AI agents that read, weighed, and judged every post in sequence before it ever reached another human being. Nothing published itself. Everything passed the gate first.
Building GoodWaves taught us the thing that became our entire thesis. The answer to powerful AI isn't to unleash it, and it isn't to cage it. It's to govern it — give it boundaries it cannot cross, and a gate where something other than the machine says yes. The "only what lifts people up" rule was our first gate. We just didn't know yet how far that idea would travel.
GoodWaves the consumer product didn't reach the world — we made the call to set it down. But the architecture underneath it didn't die. It became NN.