Blog

Notes on the multi-model future.

Essays on conversation design, model collaboration, and building Poyrock.

Pricing that follows model costs, not the other way around

Model API costs vary by orders of magnitude. Our pricing has to reflect that honestly — here's how we're thinking about it.

How Relay keeps full context intact while fanning a single thread out to models with wildly different context windows.

Early experiments in cross-model collaboration: Claude plans, GPT challenges, and the thread gets smarter than any single answer.

A look at the @mention pattern — how two characters replace model pickers, settings screens, and copy-paste.

Why the future of AI interaction is a single conversation where models meet you — not thirty open tabs you manage yourself.