Frequently asked questions

PHNTM One is a private AI assistant appliance: a small desk device with a 10.1″ touchscreen that runs its AI model on its own hardware instead of in the cloud. It chats, remembers what you tell it, sets reminders, answers questions about your own documents, and talks — with the internet unplugged. It costs $799 once, with no account and no subscription. What is a private AI appliance? →

Yes, for a specific kind of person. It does the everyday assistant jobs — memory, reminders, drafting, questions about your documents, voice — in Private mode without sending anything to a data center, which neither ChatGPT nor Alexa can do. It is not as smart or as fast as a frontier cloud model. Home Assistant control on your own network rather than a vendor's cloud is planned for 1.0 and is not yet release-verified. The compare page has both tables: the rows we win and the rows we lose.

If you're technical and already own strong hardware, you should — it's free, and a good GPU will outrun a Raspberry Pi. PHNTM One is for people who want the finished thing rather than the project: memory, reminders, document search, voice, a touchscreen and phone chat already integrated, always on, on a dedicated box that isn't your work computer.

PHNTM One is designed, assembled and tested by Jacob DeCamp, a one-person hardware operation in the United States. It is sold only at phntmcore.com. It is not affiliated with other companies or products that use the name “PHNTM”. Meet the builder →

Local chat, memory, documents, tap-to-talk and Notes are described in the current manual. Current device & 1.0 release scopeThe October 7 accepted build documents local chat, memory, document questions, tap-to-talk, Notes and optional local wake detection. Current work strengthens reminder delivery, approval checks, phone credential handling, document processing and recovery. Made-to-order purchasing is open at $799. Planned additions stay separately labelled; units do not update automatically.Check the status beside each capability → See each capability and status.

Planned PHNTM One 1.0 capability, not yet release-verified. The intended integration uses the Home Assistant you already run to control the lights, plugs, scenes and thermostats you allow over your local network, with a token that stays on the device; locks and garage doors wait for a tap on the screen. No vendor cloud and no new account. It is the voice and the screen, not the hub, so you need Home Assistant. How it works.

The local model can answer without internet, but can be wrong and has no live news. Planned · not release-verified Planned 1.0 capability, not yet release-verified: local Wikipedia and reference packs with source-linked answers. The current local model can answer general questions, but that is separate from a reference library.

Planned · not release-verified Planned 1.0 capability, not yet release-verified: several named kitchen timers with on-screen countdowns and completion sounds. Restart timing has not been established. Documented · local Save one-off or recurring reminders on the device. Today’s accepted fixes preserve pause state, recurring schedules across timezone changes and pending delivery records. Alerts still depend on a verified clock. No RTC coin cell is installed, so an offline cold power loss can leave wall-clock time unknown. Timely offline alerts after that power loss are not established.

Yes. Send a photo of a receipt, a letter or a whiteboard, or a scanned PDF, and it reads the text on the device and files it with your documents; answers cite the photo they came from. It reads the words in a picture; it doesn't describe what's in a photo.

Documented · local Text replies support ten language choices. Speech in all ten languages remains planned for 1.0, not yet release-verified. Do not assume a text-language choice provides the same spoken support.

Owner-initiated signed updates use a documented local procedure with verification and self-test rollback. The Skills screen, on-screen Update Center, account setup screens and paired-phone list remain planned 1.0 scope, not yet release-verified. Existing CLI/API and signed-update paths are separate from these proposed screens.

The manual documents bounded service and index repairs; available evidence determines what it can try. These are not a promise that every fault recovers. Planned · not release-verified Planned 1.0 capability, not yet release-verified: a weekly summary of incidents, repairs and items waiting for approval. Read the self-care scope.

Yes. The assistant — the model, your memories, the chat — runs entirely on the device. Unplug your router and it keeps working. You only need a connection for extras you choose: Boosted mode with your own key, or the optional Telegram and email connections. Updates come as a file you put on the device yourself. The clock sets itself over the internet, so after a power cut with no connection the time can be unknown until it's back.

Documented · local Local chat, memory, document retrieval and speech run on the device in Private mode. Connected Telegram, email and calendar accounts have their own network paths. Optional cloud-model answers send context to your provider. See the paths and controls.

Gemma 3 4B — Google's open-weights model, quantized (Q4_K_M) to fit the Raspberry Pi 5 — plus a local embedding model for memory and document search, both running on the device's own hardware. The device always shows you which brain answered. An honest framing: a small local model is genuinely useful, not a frontier model — that's exactly the trade the product makes for total privacy, and if you want more power for a hard task, Boosted mode uses your own API key.

No. You buy the device once and there is no required subscription for local use — there's nothing for us to bill because nothing runs on our servers. If you enable Boosted mode, you pay Anthropic directly, on your own key, for what you use; none of that goes through us.

Your device keeps working, full stop. It doesn't depend on our servers for anything — not for answers, not for login, not for staying alive. It's a Linux computer you own, SSH access is documented, and your data is on your hardware in your house. This question is exactly why the product is built this way.

Yes. It runs Debian Linux, SSH is available and documented, and we will never remove it. You never need a terminal — setup happens on the touchscreen — but the machine is yours, and we mean that literally.

Scan the code on the device's screen and a private chat opens on your phone, served directly from the device over your network — no app store, no account, no cloud relay in between. Your phone is talking to the box on your desk, not to a server somewhere.

Documented · local Tap the mic or press Alt+M and talk it through. Local recording stops on silence, another tap, or after 15 seconds. Tab moves keyboard focus. Optional wake detection has its own setting and starts off. Verified on dev · local Catch an idea before it wanders off. Type a note or record up to 60 seconds, review the transcript, then choose Clean up, Develop idea, or Project checkpoint. Originals stay separate from AI drafts. Processing stays local in every mode; long notes can take several minutes. Voice, Notes and optional wake detection.

Fast enough for a desk assistant, and honest about the trade. The local model writes about two words a second, so a short answer usually takes tens of seconds and a long one can run past a minute; the first word usually arrives within a few seconds. Questions about the device itself come back almost instantly. The one slow moment is the first answer after the box has been idle for a while, which is slower, once, while the model reloads into memory (an idle box releases the RAM on purpose). After that it's back to its usual pace. This is the cost of running a real model on your own hardware instead of a data center — and it is why Boosted mode exists for the rare hard task.

Raspberry Pi 5 with 8 GB RAM, a 10.1″ 1920×1200 touchscreen (5-point touch), speaker, active cooling, and 256 GB NVMe M.2 SSD storage included — in an enclosed case. Full details on the specs page.

Yes. International orders are available by quote, subject to delivery availability. Request a shipping quote with your country, city, postal code, and number of units. Orders shipping outside the US are subject to shipping charges. We confirm your destination, shipping cost, and power supply before payment. Import duties, taxes, and customs fees may also apply; your quote explains which costs you pay. US shipping remains free.

PHNTM One is built to order in small batches. Right now ordering is open. Your unit ships within 7–10 business days after payment, with a tracking number to your email the day it leaves. Every unit is assembled, burned in for at least 24 hours, and tested before it ships. Free shipping in the US; shipping & returns has the rest.

$799: one price, once, no required subscription for local use, and the voice bundle (USB microphone + mini wireless keyboard) is in the box. You can order one now through secure Stripe checkout. Applicable tax is shown before you pay. Questions first? Email me directly.