Explainer · 01

Why local?

Local AI means the model runs on a computer in your home or office instead of in someone else's datacenter. That one change — where the thinking happens — decides who can see your work, who sets the rules, and what happens when the internet goes down.

Where your words go today

When you type into a cloud AI, your message travels to the provider's servers, gets processed there, and the answer travels back. Every document you paste in makes the same trip.

What happens to it afterwards — how long it's kept, whether people at the company can review it, whether it helps train future models — is set by that provider's policy. You agreed to it when you signed up, and it can change.

For most everyday use, that's a perfectly reasonable trade. The real question is whether it's reasonable for your work.

What running it locally changes

  1. It stays in the building

    Your questions, documents and answers are computed and stored on hardware you own. Nothing is sent anywhere to answer you, so there's no second copy somewhere else to be retained, reviewed or leaked.

  2. The rules are yours

    No account, no terms of service that update next quarter, no model quietly swapped out for a different one. The machine behaves the way it did the day it was installed until you decide to change it.

  3. It works offline

    Unplug the router and it keeps answering. No outage on someone else's status page, and no usage limit hitting at the moment you need it most.

  4. The cost is known up front

    You pay for the hardware once instead of per person, per month. Whether that ends up cheaper depends on how many people use it and how heavily — we'll run those numbers with you honestly.

The honest trade-offs

  • The biggest cloud models are still ahead. For drafting, summarising, searching your documents and coding help, the gap is small. For the hardest frontier reasoning, it isn't.
  • It costs more to start. Builds start at $2,500. A cloud subscription is a small monthly fee.
  • It's a physical machine. That's the point, but it also means it can need a repair. Every part carries its manufacturer warranty, and repairs are house calls.

Who it's actually for

If your work comes with an obligation to keep things private — attorneys, therapists, accountants, doctors, founders with an unreleased plan — cloud AI is a policy question before it's a technical one. Running the model locally answers it.

If your work doesn't, cloud tools are probably the right answer, and we'll tell you so.