local-first

Local-first means finishing a task—like editing a photo or a document—completely on your own computer, without ever sending it over the internet to some server. It's like cooking dinner in your own kitchen instead of ordering delivery. Delivery means placing an order, giving your address, waiting, and paying a fee, but home cooking is done on the spot as long as you have the ingredients.

The advantages are clear. There's no need to upload files, so things go faster; work continues even without internet access; and since your photos or documents never pass through someone else's server, there's less to worry about in terms of storage and privacy. The tradeoff is that your computer's graphics card has to take on all that extra work itself, so you need a certain level of hardware performance to back it up.

Recently, tasks that used to be handled almost exclusively by cloud services—like converting images into 3D models or training AI models yourself—are increasingly moving to local-first desktop apps.

An article describes the open-source app Modely, noting that "the entire process happens on the user's own computer, with no separate server or cloud API calls involved," calling this a "local-first AI tool trend." One common misunderstanding is that local-first doesn't automatically mean 'smarter AI.' It only means the processing location has shifted to your own computer—it doesn't automatically mean the output quality improves.

Try turning off your internet connection for a moment and see if the AI tool you just installed still works normally. If it does, it's likely built with a local-first approach. On the other hand, if an error pops up as soon as your connection drops, that tool has been quietly communicating with a server somewhere the whole time.