01PRIVATE AI APPLIANCE COMPANY

AI is becoming electricity. Some will rent it. Some will own the generator.

We build premium private AI appliances for specific jobs, then scale them through owned brands and creator partnerships.

Pay once. Own it. Nobody meters it, changes it, or reads it.

CONCEPT DECK · JULY 2026

PRIVATE AI. OWNED.01 / 15
02THE PROBLEM

Cloud AI is rented.

Accounts. Meters. Changing rules. Your questions stored on someone else's servers. For most uses that trade is fine. For your finances, your kids, your contracts and your restricted work, it is not.

Cloud AIRented, metered, logged
UsOwned, unlimited, private
PRIVATE AI. OWNED.02 / 15
03THE PRODUCT

A complete AI for one job. On a drive you own.

Plug it into a Mac or Windows laptop. One app opens. The AI runs on your machine, answers from vetted source material, and keeps everything local. No account. No subscription. No data leaves the room.

FPO · NOT ACCURATE PRODUCT DESIGN
  • 01
    Plug in
    works in minutes, no admin rights where your device policy allows
  • 02
    Adapts to the machine
    profiles the computer and picks the right model tier automatically: light, balanced or max. No configuration.
  • 03
    One job
    each product does one thing completely
  • 04
    Answers with sources
    retrieval from vetted material, cited
  • 05
    Local memory
    progress and history stay on the device
  • 06
    Annual editions
    fresh models, updated knowledge, once a year
PRIVATE AI. OWNED.03 / 15
04WHY NOW

Local AI is ready. The product layer is empty.

Open models now run well on everyday computers. A solo founder shipped 4,000+ offline AI drives in one year and proved non-technical buyers pay for plug-and-play. The remaining work is trust, design, packaging and distribution. That is product work, and it is the work we are best at.

  • 01
    Open models run on normal laptops
  • 02
    4,000+ units sold by one competitor proves the format
  • 03
    Buyers pay for packaging, not model weights
  • 04
    Nobody owns the vertical layer yet
PRIVATE AI. OWNED.04 / 15
05COMPETITION

few sells storage.

The field segments by user type: preppers, privacy professionals, IT power users, builders. Nobody segments by job. We sell finished products for jobs people already pay to solve.

OffGrid AI, PortableMind
StrengthProved buyers pay for offline AI hardwareGapSurvival niche, one generalist stackUsEveryday jobs, vertical depth, annual editions
Docket Mini and Pro
StrengthEncrypted multi-role workstationsGapBuilt for professionals, complexUsOne job per product, zero configuration
The Ultimate USB
StrengthBootable OS with local AI toolsGapIT and security users onlyUsPlug in, one app opens
DIY, Ollama and AnythingLLM
StrengthFree and infinitely flexibleGapSetup, tuning and maintenance become the productUsNever think about it
Cloud AI
StrengthThe smartest modelsGapRented, metered, loggedUsOwned, unlimited, private
PRIVATE AI. OWNED.05 / 15
06THE MODEL

One engine. Many verticals.

The engine is built once: launcher, hardware detection, local retrieval, signed updates, safety layer, support automation. Each new product is a knowledge pack, a brand and a channel. A new vertical costs weeks, not quarters.

BUILT ONCE

  • launcher
  • hardware auto-tiering
  • local RAG runtime
  • signed update system
  • safety layer
  • AI-first support

PER VERTICAL

  • curated knowledge pack
  • brand and site
  • channel
  • annual edition pipeline
PRIVATE AI. OWNED.06 / 15
07FIVE VERTICALS

Five products. Ledger leads, the market confirms.

  • 01
    LEDGER · Private tax research and small business document assistant
    "Your financial questions should not become someone else's training data."
  • 02
    TUTOR · Private offline tutor for families
    "Your child's learning profile should stay in your house."
  • 03
    VAULTCODE · Offline coding assistant for restricted work
    "Your restricted code should not leave your machine."
  • 04
    HEARTH · The family vault
    "Your family archive should not live inside another company's account."
  • 05
    BRIEF · Legal document explainer
    "Your contract questions should not require an upload."
PRIVATE AI. OWNED.07 / 15
08LEAD VERTICAL

Ledger leads.

Ledger answers from IRS, SBA and state publications and cites what it retrieves. Tax law changes every year by statute, so the annual edition renews itself. The buyer already pays for tax software, and financials are the data people most refuse to put in a cloud. Extension deadlines in September give it a live selling window now, and January is the landslide.

Founding batch
$199
First 100, numbered, ships in weeks
Annual edition
$59 / year
New models, new law, new knowledge
Positioning
Research, not advice
Cites sources. Never files, never advises.
PRIVATE AI. OWNED.08 / 15
09THE PROOF PLAN

Five tests. One founding batch. Six weeks.

Every vertical gets an identical page. Ledger sells a real numbered founding batch at full price. The other four take $29 refundable reservations. Identical pages, identical budgets, so the numbers are comparable. Phase 0 identifies obvious pull. It does not prove absence of demand. Obvious losers die, marginal results retry in season.

MetricPassKill
Cost per paid order (lead)under $75over $110
Cost per reservationunder $50over $75
Visitor to orderover 1.5%under 0.5%
Refunds during testunder 10%over 25%

TOTAL EXPOSURE ~$12.5K · MOST OF IT RESELLABLE HARDWARE

PRIVATE AI. OWNED.09 / 15
10SEQUENCE

Ledger now. Tutor on a proven engine.

  • 01
    July
    five pages live, Ledger founding batch on sale
  • 02
    August
    engine hardens on Ledger's forgiving professional buyers
  • 03
    September
    Ledger ships into extension season
  • 04
    October
    Tutor launches into report card season, guardrails first, typed questions only in v1
  • 05
    November
    Hearth gifting drop if its numbers pass
  • 06
    January
    Ledger 2027 Edition, the first annual renewal
PRIVATE AI. OWNED.10 / 15
11ECONOMICS

Hardware margin now. Software margin next.

Drive
$149–249
about $100 gross profit per unit
Download
$79–119
follows once a vertical is proven, 80%+ margin
Annual edition
$49–79
target 30% attach, required by reality in tax

100 drives cost about $8k and resell near cost if the test fails. The founding batch collects about $20k before pack production finishes.

VOLUME ECONOMICS · DIRECTIONAL
UnitsBlended gross profit
100~$11k
1,000~$109k
10,000~$1.1M

Assumes about $100 gross profit per drive, $85 per download, 30% annual edition attach. Before CAC, team costs, returns and overhead.

PRIVATE AI. OWNED.11 / 15
12HOW IT SCALES

Rent audiences. Don't buy customers.

Customer acquisition is the dominant cost in this category. Creator drops delete it. A creator's branded appliance, their knowledge pack, a numbered limited run, 20 to 30 percent revenue share. The audience already trusts the brand, so acquisition cost approaches zero. Every drop is also a live demand test for a future owned vertical.

  • 01
    Creator provides brand, channel, trust
  • 02
    We provide engine, pack production, fulfillment, support
  • 03
    Numbered runs create urgency and collectibility
  • 04
    Annual editions renew the relationship every year
PRIVATE AI. OWNED.12 / 15
13WHAT KILLS THIS

Three ways this dies. Named now.

  • 01
    OS vendors absorb local AI
    Apple and Microsoft ship good-enough local models. Answer: occupy ground they cannot ship. Compliance-scoped, creator-branded, vertical depth.
  • 02
    Support load
    Random hardware generates tickets on its own schedule. Answer: compatibility quiz before purchase, auto-tiering, AI-first support, refund-friendly margin. Above 0.2 hours per unit, we stop and fix.
  • 03
    Small model quality
    Local models bound every promise. Answer: scope each product to jobs small models do well today. Retrieval, citation, explanation. Never open-ended expertise.
PRIVATE AI. OWNED.13 / 15
14GATES BEFORE SCALE

The numbers decide. Not the deck.

Setup works first timeabove 90%
Refund ratebelow 10%
Support loadbelow 0.2 hours per unit
Annual edition interestabove 25%

If these hold, we scale. If they fall short, we learn first. No outside money conversation until the founding batch clears every gate.

PRIVATE AI. OWNED.14 / 15
15THE ASK

Six weeks to a market answer.

Core team funds Phase 0. $12.5k exposure, most of it resellable hardware. Five pages, one founding batch, and the market chooses the first product.

  • 01
    Fund Phase 0
    $12.5k
  • 02
    Build the five pages
    2 weeks
  • 03
    Ledger founding batch on sale
    late July
  • 04
    Read the numbers
    week 6

PRIVATE AI. OWNED. · END OF DECK

PRIVATE AI. OWNED.15 / 15