For a while, that question had one practical answer: the cloud. Open models were a novelty, the good ones lived behind someone else's API, and running anything capable on your own hardware was a hobbyist's compromise.
That changed. Open-weight models are now genuinely capable — good enough that the bottleneck moved. The models aren't what's keeping AI in the cloud anymore. The software around them is.
Privacy should be a property, not a promise
When your data never leaves the building, there is nothing to leak, nothing to subpoena from a third party, and no terms-of-service update that quietly changes the deal. That's worth something to a researcher with unpublished work, a school with student records, a company with internal procedures — or anyone who simply believes their documents are theirs.
Ownership compounds
Hardware you own gets better as better open models arrive. It doesn't get worse when a provider changes its pricing, deprecates a model, or exits the market. Cost stops being a meter running in the background and becomes a choice you made once.
Applications are the last mile
A capable model on your machine is a starting point, not a product. Most people don't want a chat window; they want to learn the thing, finish the task, get the answer right. That takes applications designed from the first line of code around local models — their strengths, their limits, and the hardware people actually own.
That's the work of JMNI Labs. Ward brings a private AI assistant into medical and legal offices; Recitation turns any textbook into a private course. And our research lab exists so that everything we build is grounded in measured reality.
Local-first isn't nostalgia for a pre-cloud world. The hardware caught up, and the models caught up. The applications are next.