Offline Base
Building AI for people who are wary of AI
Why Offline Base builds for AI skeptics, what Pew and Gallup found in 2026, and how owned, local AI and hands-on teaching answer them.
In 2026, for the first time, a majority of Americans under 30 told Pew they are more concerned than excited about AI. Many of our friends are in that group. Belief 4 on our About page (opens in a new tab), "No one gets left behind in an AI-native world," starts there: "We are young, and a lot of our friends want nothing to do with AI."
The belief goes on: "We want our peers to access and benefit from this technology without feeling they are making a moral concession." We think AI can be a force for good, and we want the people who distrust it most to have a version that fits their principles.
The survey figures below come from Pew Research Center and Gallup. Our own reading of them is in the section "Rented, remote AI is a design choice."
What Pew found among young adults
Pew Research Center published a short read on August 18, 2026, from a survey of 3,488 US adults fielded June 22 to 28, 2026. Its lead finding: "For the first time, a majority of adults under 30 (55%) now say they're more concerned than excited about AI."
Share of US adults who say they are more concerned than excited about AI:
Group | 2021 | 2025 | 2026 |
|---|---|---|---|
Adults under 30 | 31% | 47% | 55% |
All US adults | 37% | 50% | 52% |
The other side of the question moved as well. Among adults under 30, the share more excited than concerned fell from 25% in 2021 to 11% in 2026. Young adults started out less worried than adults overall, and in 2026 they are more worried.
The same report asked about work. Of all US adults, 71% expect AI to lead to fewer jobs in the United States over the next 20 years, up from 64% in 2024. Among adults under 30, that figure is 73%, up from 61%.
What Gallup found about data centers
Gallup published "Americans Oppose AI Data Centers in Their Area" on May 13, 2026. In a phone poll of 1,000 US adults taken March 2 to 18, 2026, 71% opposed building an AI data center in their local area, and 48% strongly opposed it.
Gallup asked opponents why in a separate Gallup Panel web survey of 2,054 adults, run April 1 to 15, 2026, so the 71% and the reasons come from different groups of people. About half of opponents pointed to excessive use of resources, including water (18%) and energy (18%). Another 16% mentioned pollution.
Rented, remote AI is a design choice
This part is our opinion. Much of the wariness in those numbers attaches to how AI reaches people today: rented by the month from a few large companies and run in data centers that someone has to build near somebody's home. Those are design choices.
A product can change where the model runs, who owns it and what it costs to try.
Owned, local and checkable
Base Drive is the first product we plan to launch. It is a USB drive that sets up private, offline AI on the Apple Silicon Mac or Windows 11 PC you already own. As of October 2026 it is marked coming soon and cannot be ordered yet.
- Owned. The planned price is $19.99, paid once, with no tiers. The Base Drive page (opens in a new tab) calls it "about what one month of a chatbot subscription costs, except you pay it once and keep the thing." There is no account, no telemetry and no usage meter.
- Local. The installer copies the AI onto your laptop, and after setup it keeps working with Wi-Fi off.
- Checkable. The installer is open source, every file on the drive is listed, and setup verifies published SHA-256 checksums before it copies anything. How Base Drive earns the trust a USB drive needs explains each step.
Your own computer runs the model; the drive carries the installer and the models. That is our answer to belief 2, "Running models locally answers the data-center problem." The same belief admits the research on water, power and noise "is not fully established, and we say so." The energy case for local AI, with our caveats goes further.
Starting costs nothing. At the try page (opens in a new tab) you can send three messages to Liquid LFM2.5 1.2B, running on a Jetson Orin Nano Super we operate, without signing up. It is a rate-limited demo that can go offline, and the page says "These chats aren't stored." Three things you can try on offlinebase.com without an account covers the other free demos.
Plain language first
The home page headline is "Making Local AI Easy for All." Our first newsletter issue, "Intro to Offline Base," was frank about where things stand: "We think that Local AI is still too niche and nerdy."
Plain words are part of the fix. Most pages on offlinebase.com have an "Ask a small model to…" box. You can ask for the page in Spanish, or tap "ELI5" to have it explained like you're five, and a small model on our own Jetson boards rewrites the text in place. We left that box off the app privacy policy on purpose, because its wording has to stay exact.
We also try to state limits early. The "How we work" section of the About page opens with "We'd rather under-promise." Our terms (opens in a new tab) warn that any model on the site can be "confidently, fluently wrong," and that small local models are more prone to it. A skeptic should hear that from us before buying anything.
Teaching people to run it themselves
Two of the four recordings on our demos page (opens in a new tab) teach local AI by showing it.
- Local AI in the NYC subway. A model answering underground, "with no signal to fall back on." With the network gone, the answer has to come from the device itself.
- Learn to run LLMs locally, a CUNY workshop. A hands-on session that walks a room through running language models on their own machines. The page calls it "the same thing Base Drive does for you in one click," which is what Base Drive is built to do once it launches.
Teaching matters to us because a person who has run a model on their own laptop can judge it from experience. They can see what stays on the machine and what the model gets wrong, then decide for themselves whether it belongs in their life.
Limits as of October 2026
- You can't order Base Drive yet; it is marked coming soon.
- A small local model trails the largest cloud systems on the hardest questions. What a million real questions say about small local models looks at where that line falls.
- Local AI speaks to the data-center worry in Gallup's poll and to keeping what you type private. The jobs worry in Pew's survey is a bigger question than any product we make.
If you are one of the friends who want nothing to do with AI, the try page (opens in a new tab) takes a minute.

