Local AI,
straight answers.
The questions we get asked most about running AI on your own hardware — answered from actually building and running it. No hype, no sales pitch.
Is it actually private, or does it phone home?
Your prompts and documents stay on your machine — that part's real. But "local" doesn't automatically mean airtight: the app running the model (Ollama, LM Studio) has normal network access, and a couple phone home metadata by default — LM Studio collects anonymous usage analytics, and it'll sync your device list if you let it. Not your prompts, but still. Genuine privacy is one step of config: turn off the tool's telemetry and keep it off the internet. Then pull the cable and it still answers — that's the test. It's also exactly how we build our boxes: nothing to turn off, because nothing's phoning home in the first place.
Why not just use ChatGPT, then?
Ownership and privacy. No subscription, no rate limits, no policy change breaking your setup, and your sensitive data never leaves the building. If none of that matters to you, use the cloud. If it does — and for real business work it usually does — that's the case for local.
What does owning it cost against my subscriptions?
Do your own math — the Boone page has a calculator with deliberately conservative floor prices. The typical answer: a small business bleeding $1,500–3,000 a year in subscriptions buys the Software Edition once for $139.99, and updates are an optional $34.99 a year — the box stays fully usable without them. It usually pays for itself in weeks.
What hardware do I need?
Less than you'd think to start. A modest gaming GPU (8–16GB) runs small models fine for learning. A 24GB card runs a 30B-class model comfortably — our daily driver is a 35B mixture-of-experts. The big unified-memory boxes are for loading huge models slowly. Don't overbuy for a 70B you won't actually use day-to-day.
Do I need internet?
Only once, to download the model. After that it runs fully offline — air-gap it and it keeps working. A dropped connection doesn't take your AI down with it.
Which model should I run? Is it as good as ChatGPT?
Different tool, not a worse one. The "what model" answer is always two or three names that rotate every few months — right now (mid-2026) most people land on the Qwen 3.6 pair (27B dense for quality, 35B-A3B MoE for speed — on our own 24GB Arc B60 the MoE measured ~31 tok/s at 8K context, over 4× the old 32B dense), or Mistral Small if you want lean. Don't chase the biggest number: a fast 30B-class model does real work, and for anything that needs frontier-level reasoning the cloud still wins. Local wins on ownership, privacy, and knowing your own documents.
Which app — Ollama or LM Studio?
Both run the same models; pick on how you like to work. LM Studio is the friendly GUI — install, click, chat — best for your first week. Ollama is what most people settle on once they want it scriptable: a REST API on by default, runs headless on a server, loads and unloads models on its own. Start with LM Studio to learn the feel, move to Ollama when you want to wire it into things.
Is it hard to set up?
The easy 90% is genuinely easy: install Ollama or LM Studio, download a model, start chatting — no terminal required. The hard 10% is squeezing specific hardware to its limit, and most people never need to touch that part.
Why does it forget what I told it a minute ago?
Almost never the model — it's the context window, and it's the most common "my local model is dumb" complaint there is. Out of the box, Ollama defaults context to 2048 tokens (num_ctx) — maybe 1,500 words of conversation plus the reply. Run past that and it quietly drops the oldest tokens, so the model can't see the start of what you were doing. It's a setting, not amnesia. Fix it in one line: in Ollama, /set parameter num_ctx 8192 (or PARAMETER num_ctx 8192 in a Modelfile); in LM Studio, the context-length slider. Set it to fit your work — bigger context costs VRAM, so match it to the job.
How do I start?
Install LM Studio or Ollama, pull a small model, and point it at a few of your own documents. An hour with it on your own files will tell you more than any benchmark.
Can it use my own documents?
Yes — and that's the real unlock. You point it at your files (manuals, contracts, notes) and it answers from them, cited to the page, instead of guessing from its training. A grounded mid-size model reading your actual documents beats a giant one on the internet that's never seen them. (It's called RAG.)
What's it actually good for?
Drafting and rewriting, summarizing long documents, answering from your own files, coding help, cleaning up messy data, and running quiet little automations. The bread-and-butter, all day, for the cost of electricity.
Can local AI read paper — photos of invoices, nameplates, receipts?
Yes, and this is where the honesty matters: a vision model reads most of a document perfectly and then silently drops or misreads a field with total confidence — we measured a model rating every field "high confidence" while some were wrong. Our answer is two readers that fail differently (a vision model plus plain deterministic OCR), disagreement flagged instead of guessed, and every dollar amount or rating confirmed by you against the photo crop. Across our 41-image test set — glare-shot motor plates, 1936 typewriter invoices, crumpled receipts — zero wrong high-risk numbers were ever committed silently. Never trust a raw single-model transcription with numbers that matter.
Can I talk to it?
Yes — push-to-talk, or an optional "Hey Boone" wake word. Everything voice happens inside the box: your words never leave the building, a light shows whenever the mic is live, and one tap mutes it. It's off by default — turning it on runs a one-minute mic check that tunes it to your actual mic and room. And his voice is built from openly licensed parts, no cloned humans — pick it out of a lineup.
Do I have to learn all that replacement software myself?
No — that's the part nobody else does. Every app on the box arrives with its maker's own manual read in beside your documents, so you ask in your own words ("how do I cut the boring part out of my video?") and get the answer from the app's own pages, cited, offline. If you don't know an app's vocabulary yet, it offers plain starter questions in shop language, not software language.
What's the camp on the screen actually for?
It's the status display — no dashboards, no notification badges. The fire burns while the AI works; the cairn stands while your backups are current; paper lanterns light along the dock, one per outside connection, so a dark dock means nothing is reaching beyond your walls. You read the state of your box the way you'd read a room. (And yes, a few of the camp objects are playable. We'll let you find the rest.)