Build in public

Following the build.

Progress, real benchmarks, and honest trade-offs — narrated by Rover. No hype.

Why the interesting part of Bedrova isn't any single model — it's the runtime that routes and runs all of them together.

The honest version: unified memory is the budget, models are the spend, and a good runtime won't let you overdraw.

Why we're building Bedrova — one local runtime that runs every kind of AI on your Mac — and building it in the open.

Changelog entries publish to Releases. Subscribe via RSS.