We shipped our first local-first prototype three years ago and it was ugly. Slow, small, and buggy. We asked ourselves why we were building it — and the answer became the whole point of PeaBox Lab.
Every time we handed a tool to a friend, the same question came up: Where does my data go? With cloud AI you get a confident answer, but you never quite know. Projects you pipeline through an external service can't be unshared.
So we made a different call on every product: run it locally, keep the output local, send nothing anywhere without an explicit click. Speech Studio transcribes your recording on your machine. PhotoLab edits your photos on your machine. The Workbench talks to a model running on your CPU or GPU — no API key required.
It's not always the fastest option. Local inference costs compute and patience. But what you get back is the thing most software promised and never delivered: your work stays yours.