Loci

AI that runs only on your device, never on a server.

Works offline. Completely private. Free, without limits.

[Loci @askloci]

Fund descriptions and expense ratios are public. Your account balances and lot sizes are not. Look up the public part online, then go offline and use a local model for the private math. Your actual figures stay on your device while you sort out your own plan.

[Loci @askloci]

Processing a strained friendship honestly means writing down the petty stuff, the contradictions, the part where you might be wrong. If internet is off and you're using a local model, all of that stays on your device. Nothing gets uploaded.

[Loci @askloci]

Forty handwritten recipe cards is forty separate formatting requests. A local model generates each one on your device, outside any cloud chat plan, so there's no message cap stopping you halfway through the box.

[Loci @askloci]

Preparing to negotiate a raise means working through your actual salary, the number you want, and your evidence. With internet off, a local model processes all of that on your device without uploading it.

You decide which points make it into the real conversation.

[Loci @askloci]

Before counting on a local model in a place without internet, run it once while you're still connected. The files need to be downloaded and cached ahead of time.

A quick test confirms everything works before you're somewhere you can't fix it.

[Loci @askloci]

Figuring out which debt to tackle first means lining up every balance and interest rate you actually have. With internet off, a local model works through your numbers on your device. Those real account figures stay on your machine while you sort out your own payoff order.

[Loci @askloci]

Have you ever held back from letting someone else use your chatbot because you share the message limit?

A local model generates each reply on the device. A whole afternoon of curious questions doesn't draw from a shared cloud plan's message cap.

[Loci @askloci]

You don't plan for the internet going out at home. But if a compatible model is already on your device, you can keep asking it questions offline. The text is generated locally, so that part of your work keeps going while anything cloud-dependent stops.

[Loci @askloci]

After downloading a model, Loci can answer typed questions offline. A commute through tunnels or dead zones becomes time for an outline, a first draft, or a list.

https://t.co/X5A5sY5mgF

[Loci @askloci]

When you're naming a project that hasn't been announced, use a local model with internet off. It generates suggestions on your device, and the whole list of working titles stays on your machine.

[Michael Waldman @mwaldman130]

Local AI caught up way faster than anyone realizes. Sharing my thoughts having spent time in ML at Meta and dedicating the last year to getting smart on local AI.
1/ Barely anyone, not even hardcore AI power users, realizes how fast local models have caught up to the frontier. Qwen3.8 woke a few people up, but the most technical people are so frontier-pilled they've blinded themselves to the breakneck progress.
Meanwhile, normal users don't care about frontier intelligence at all. They want AI that gets the job done, but they have no idea what a local model is or how to run one.
So local is stuck in a weird spot: the people who can run these models don't care to do so, and the people who'd benefit most don't know how to run local.

[Loci @askloci]

Sorting out which symptoms matter to your doctor is half the work of writing a portal message. A local model, with internet off, lets you lay out every date and detail on your device without uploading the sensitive parts.

Then you paste only the relevant lines into the portal.

[Loci @askloci]

Picking a title means generating a pile of options and rejecting most of them. A local model handles each attempt on your device, so a dozen throwaway headlines don't count against a cloud chatbot's message limit.

[Loci @askloci]

This is why local AI is the future. Most people don't care about frontier models. They prefer unlimited AI, with the added benefit of complete privacy. Loci will keep building towards this future.

[a16z @a16z]

98% of US households aren't paying for AI yet

More charts in State of Markets II: https://t.co/MTaxKUxa2w

[Loci @askloci]

Testing a flawed idea is useful. You learn by seeing where it falls apart. With internet off and a local model, you can ask it to argue against your premise. The whole exchange stays on your device, including the parts where you were completely wrong.

[Loci @askloci]

Quick editing tasks add up fast: reword this line, tighten that paragraph, try a softer tone. A local model generates each revision on your device, so those requests don't count against a cloud chatbot's message limit.

[Loci @askloci]

Ever started writing a family story and realized it's full of other people's private details?

In a genuinely offline session with a downloaded model, the text is processed on your device. Those names and personal details don't get sent to a remote server during that session.

[Loci @askloci]

With a downloaded model, each revision is generated on your device. Those five or six follow-ups in twenty minutes aren't handled by a remote inference service, so they don't use that service's message allowance.

[Loci @askloci]

Ever open the fridge mid-recipe and realize you're out of rice, so you ask for a new plan, then spot chicken you forgot about, so you ask again?

[Loci @askloci]

Need to keep asking follow-ups without rationing your messages?

Loci's on-device chat has no message quota.

https://t.co/Xjvo9qYtlt