01
Seed Round — March 2026

AI that runs on your device.
Not someone else's server.

Digital Disconnections is building the on-device AI platform — voice assistants, health apps, and productivity tools that run entirely on the hardware you already own. Private by architecture. No cloud. No subscriptions. No data collection.

Stage: Seed
Market: Edge AI & Privacy Tech
Products: 3 (2 in late-stage testing)
HQ: Founding the Commons, SF
02
The Problem

Your device is powerful enough.
The software isn't.

Modern phones ship with dedicated neural processors capable of running AI locally. But every major AI product still sends your data to the cloud. The hardware is ready. The software hasn't caught up.

01

The Cloud Bottleneck

Siri, Google Assistant, and Alexa all route your voice to remote servers. Every question, every command, every overheard conversation — processed on someone else's hardware. Lose your signal — camping, flying, rural areas — and they go silent. No connection, no assistant.

02

The Data Extraction Model

“Free” apps monetize your behavior. Health data, voice recordings, keystrokes — collected, profiled, and sold. And every byte they store is breach bait — T-Mobile, 23andMe, and health apps have made that clear.

03

The Subscription Lock-In

AI features gated behind monthly fees for cloud compute you don't need. Your phone's neural engine sits idle while you pay rent for inference that could happen locally.

03
Why This Matters

This isn't abstract. It's personal. It's about dignity.

We didn't start Digital Disconnections because we saw a market opportunity. We started because the people we love were being hurt by the systems we helped build.

01

The Woman Who Was Asked to Train Her Replacement

Early this year, our CEO Jamal's mother-in-law came to us shaken. Her employer — after thirty years of her expertise — asked her to train an AI system to do her job. Three decades of institutional knowledge, distilled into a dataset she'd never own or benefit from.

We proposed something different: help her build a tool from her expertise, one she would own. Her knowledge, her IP, her future — not her employer's training data.

02

The Period Tracker That Knew Too Much

Our CTO Cassidy's wife used a popular menstrual cycle tracker — until she started receiving eerily timed ads right before her period. The app was selling her most intimate health data to advertisers. Her cycle, her moods, her fertility — monetized without her knowledge or consent.

That's why we built Cara. A period tracker that runs entirely on your device. No accounts, no servers, nothing to sell — and in a post-Dobbs world, nothing to subpoena.

Every product we build starts with a person, not a spreadsheet.

These aren't edge cases. They're the norm. Millions of people are having their expertise extracted, their health data sold, and their attention auctioned — by the very tools they depend on. We build the alternative.

04
Our Solution

Private by design, not by policy.

We don't ask you to trust our privacy policy. We build systems where your data physically cannot leave your device. On-device inference, zero network access, one-time pricing.

There is nothing to send.

🔒

On-Device Inference

AI models run on the phone's neural processor. No server roundtrip, no cloud dependency, no latency penalty.

⚡

Zero Network Access

Our apps don't request network permissions. Data can't leak to servers that don't exist — and your tools keep working off-grid, in airplane mode, or during an outage.

💰

No Subscriptions

On-device compute means no cloud costs to pass on. One-time purchase or free. No recurring fees.

🚫

No Data Collection

No accounts, no analytics, no telemetry. We don't know who you are and we don't want to.

05
Product Pipeline

Three products. One platform.

Each product answers the same question at a different scale — what if your software couldn't betray you? SafeType Keyboard answers it for your words. Cara answers it for your body. Private Assistant answers it for your life.

Private Assistant
In Development

Private Assistant

An open-source, on-device AI assistant built to replace Siri and Google Assistant. Your assistant, your integrations, your rules.

Open Core Open API iOS & Android 100% On-Device
  • Any app can connect — no walled garden
  • Revenue: premium integrations, enterprise licensing
Cara
Late Stage Testing

Cara

Private menstrual cycle tracking. On-device computation, encrypted local storage, zero data collection. Nothing to subpoena.

Free — Always iOS & Android Zero Data Collection
  • Post-Dobbs demand — period tracker adoption grew 20% and continues rising (Neiman et al., Contraception, 2025)
  • 50M+ TAM — women of reproductive age in the US alone
  • Acquisition funnel for Private Assistant ecosystem
SafeType Keyboard
Late Stage Testing

SafeType Keyboard

An iOS keyboard with on-device speech-to-text powered by the phone's neural processor. Keystrokes never leave the device.

$4.99 — One-Time 500K Units Yr 1 ~$1.75M Net Revenue
  • Early enterprise interest — inbound inquiries from healthcare and education orgs exploring on-device AI for HIPAA/FERPA compliance
  • Zero network access — physically cannot transmit keystrokes
  • First revenue product; validates on-device model
06
Market Opportunity

Three signals are converging. We're the only company at the intersection.

TAM $103B
SAM $14B
SOM $250M

TAM — $103B by 2030

Global edge AI market, growing at 29% CAGR. Every smartphone with a neural processor is a potential platform. (The Business Research Company, Edge AI Global Market Report, 2024)

SAM — $14B

Consumer on-device AI software: voice assistants, health apps, keyboards, and privacy tools on mobile devices. 23.6% CAGR through 2030. (Reed Intelligence, Consumer Devices AI Market Report, 2025)

SOM — $250M

Privacy-first consumer AI apps. SafeType Keyboard at 500K units = ~$1.75M net Yr 1. Cara as free acquisition funnel. Private Assistant as platform play.

$393B
Signal 1: On-Device Hardware Validated
End-user spending on AI-enabled smartphones, up 32% YoY. The hardware is in people's hands. (Gartner, Forecast: AI Smartphones, Worldwide, 2025)
20%
Post-Dobbs Adoption Growth
Period tracker adoption grew from 37% to 45% of women in tracked states — sustained behavior, not a search spike. (Neiman et al., Contraception, Jan 2025)
73%
Privacy Concern
feel they have little to no control over what companies do with their personal data. (Pew Research Center, 2023)
41%
Signal 2: Subscription Fatigue
of consumers report subscription fatigue — the market is primed for one-time-purchase alternatives. (CivicScience / Marketing LTB, 2025)

Signal 3: The Vocabulary Is Unclaimed

No major brand owns the phrase “on-device AI.” Apple says “Apple Intelligence.” Google says “Gemini.” Samsung says “Galaxy AI.” The category name itself — the clearest, most intuitive description of what consumers actually want — is sitting unclaimed. Digital Disconnections is building the brand around the architecture, not a product name.

07
Competitive Landscape

Everyone claims privacy. We architect it.

Apple is routing Siri to cloud AI providers. Samsung moved Galaxy AI to paid subscriptions. The industry broke trust by architecture. We repair it the same way — we’re the only company where privacy is the architecture, not the setting.

Feature Digital Disconnections Apple Intelligence
Cloud-routing Siri to 3rd parties (iOS 27)		 Google On-Device Samsung Galaxy AI
Moved to paid subscriptions 2026
100% on-device inference* ✓ Partial Partial Partial
Zero network permissions ✓ ✗ ✗ ✗
No data collection ✓ ✗ ✗ ✗
Open API / open source* ✓ ✗ ✗ ✗
No subscription fees ✓ ✗ ✗ ✗
Cross-platform (iOS + Android)* ✓ iOS only Android only Samsung only
Third-party integrations Open API Curated Curated Curated

* Reflects planned product architecture at launch. Products currently in testing and development.

The Three-Lock Moat

Lock 1 — No Network Permission

Our apps don't request network access. This isn't a setting you can toggle — it's an architectural decision baked into the binary. Data can't leak to servers that don't exist. Check the app permissions yourself: there's nothing to turn off because there was never anything to turn on.

Lock 2 — Their Business Model Is the Wall

Apple, Google, and Samsung can't fully commit to on-device AI without gutting their revenue engines. The cloud call isn't a technical limitation they're working around — it's a business requirement. Every query routed to Google's servers is a data point. These aren't architectural oversights. They're features of the business model.

They can build on-device demos. They can't build on-device companies. Their shareholders won't let them.

Lock 3 — One-Time Pricing Kills the Churn Problem

Subscription AI companies need you to keep paying. That means they need engagement metrics, usage data, and reasons to gate features behind tiers. We sell you software once. After that, we have no financial incentive to monitor you, nudge you, or withhold features.

The business model is the privacy guarantee — not the privacy policy.

"Our moat isn't a feature — it's a conflict of interest that our competitors can't resolve. They make money from data. We make money from software. You can't do both honestly."

Regulatory Tailwind

State-level AI privacy legislation is accelerating with no federal preemption mechanism in place. Colorado SB 205 live Feb 2026. EU AI Act enforcement August 2026. Every new regulation makes on-device architecture more valuable.

Pricing Moat

Samsung Galaxy AI: $10/mo subscription. Apple: App Store cuts on 3rd-party AI subs. Us: $4.99 one-time. After two months, we’re cheaper forever. After a year, it’s not even close.

08
The Apple Question

It’s not on-device if it sometimes isn’t.

At WWDC, Apple will announce a 3B-parameter on-device model. Tim Cook will use our language: “Every AI company asks you to send your data to their servers. We built one that never needs to.” It’s a strong line. It’s also incomplete.

Apple Intelligence
What Apple calls on-device
  • ✗
    Private Cloud Compute fallback
    Complex requests route to Apple’s cloud servers — with per-request user consent. Apple calls it “privacy-preserving.” It still leaves your device.
  • ✗
    Project Campos: powered by Google Gemini
    Apple’s AI chatbot runs on Google’s models. When Google changes terms, Apple’s AI changes with it. Their on-device story depends on their competitor’s infrastructure.
  • ✗
    AI App Store: 30% cut on AI subscriptions
    Apple Intelligence is also a gatekeeper model — they take a third of every AI subscription sold through their platform.
  • ✗
    Regulatory risk: cloud-dependent = regulator-dependent
    Apple Intelligence launched in China without regulatory approval (March 2026). Cloud-dependent AI can be switched off by governments. On-device AI can’t.
🔒
Digital Disconnections
What we call on-device
  • ✓
    Zero cloud fallback — ever
    No “complex request” tier. No cloud tier. No opt-in cloud. Our apps don’t know cloud servers exist because we never built the connection.
  • ✓
    Our models. Our infrastructure.
    We run our own research workstation, our own servers, our own models. No licensing from Google. No terms-of-service dependency. No partner risk.
  • ✓
    No network permissions requested
    Check the app permissions yourself. There’s nothing to turn off because there was never anything to turn on. The architecture is the privacy guarantee.
  • ✓
    Works in airplane mode, under any regime
    No signal required. No regulatory dependency. Your assistant can’t be switched off by a government — or a cloud outage — because it doesn’t talk to either.

“It’s not on-device if it sometimes isn’t.”

Apple’s Private Cloud Compute is privacy theater with a consent checkbox. The moment your data has a path out — even an optional, well-intentioned one — the architecture is compromised. We don’t have a cloud tier because we never built one. That’s not a feature. That’s a structural difference.

0
Cloud requests
ever made
0
Network permissions
requested
Binary
On-device or not —
no spectrum
09
Go-to-Market

Cara finds the users. SafeType Keyboard proves they'll pay. Private Assistant keeps them forever.

01

SafeType Keyboard — Revenue Engine

App Store Optimization targeting privacy keywords with zero competition. Enterprise channel for healthcare, legal, and financial organizations where keyboard data is a compliance risk. Community partnerships with privacy advocacy organizations.

02

Cara — Acquisition Funnel

Free launch into post-Dobbs demand (20% adoption growth · 84% of apps sell data). Press and advocacy partnerships in women's health. Every Cara user is a warm lead for the DD ecosystem — they already trust on-device AI.

03

Private Assistant — Platform Play

Developer-first: open API designed to attract integrations that make the assistant indispensable. Enterprise licensing for organizations that need AI without data liability. The network effect: every connected app makes switching harder.

Acquisition → Revenue → Retention
Cara — Free Acquisition
iOS & Android · Zero data collection · Post-Dobbs demand
50M+ TAM · 20% adoption growth · 84% of apps sell data · $0 CAC (organic)
SafeType Keyboard — Revenue Engine
$4.99 one-time · Early enterprise interest (healthcare, education)
500K units Yr 1 · ~$1.75M gross · ~100% margin
Private Assistant — Platform
Open API · Enterprise licensing · Network effects
Every integration deepens the moat
$0
Cara user acquisition cost
(organic demand)
~100%
Gross margin on
keyboard sales
∞
Network effect from
open API integrations
SafeType Keyboard
CAC $1–3 paid / $0 organic
LTV $3.49 net (one-time)
LTV/CAC >1.7x paid · ∞ organic
Cara (Funnel)
CAC ~$0 (organic + PR)
Direct LTV $0 (free)
Ecosystem LTV $3.49 × KB conversion rate
Private Assistant
CAC Near-$0 (ecosystem)
LTV Enterprise licensing (recurring)
Moat API integrations compound
10
Business Model

Cloud AI spends 60–80¢ per dollar on compute. We spend zero.

Our models run on hardware the customer already owns. No servers, no cloud bills, no per-user infrastructure costs. Every sale is near-100% margin, and customers never churn from pricing — because there's nothing to renew.

$

SafeType Keyboard — $4.99

One-time purchase. $3.49 net after Apple's 30%. First revenue product. Target: 500K units in Year 1. Early inbound interest from healthcare and education organizations exploring on-device processing for regulatory compliance.

$0

Cara — Free Forever

User acquisition funnel for the DD ecosystem. Builds brand trust and proves the on-device model to millions. Cara users become Private Assistant users.

∞

Private Assistant — Platform

Open-core model. Free base assistant, premium integrations and enterprise licensing. The platform play — every third-party app becomes a revenue channel.

Revenue Waterfall — Moderate Case

Net of App Store fees · Conservative: $698K / $1.55M / $2.85M

$6M $4M $2M $0 $1.75M $3.19M $5.74M Year 1 Keyboard Launch Year 2 Enterprise + PA Beta Year 3 Platform Revenue ~95% margin
Keyboard (Consumer)
Keyboard (Enterprise)
Private Assistant
Unit Economics: COGS vs Margin
Cloud AI Competitors
60–80% COGS
20–40% margin
Cloud inference costs ~$0.60–0.80 per $1 revenue
Digital Disconnections
~100% Gross Margin
On-device inference (customer's hardware) ~$0 COGS per sale
$4.99 retail → $3.49 net per unit.
Apple takes 30%. We keep the rest — no server to run, no inference to pay for, no per-user cost at any scale. Every additional user is pure margin.
Revenue Mix by Year 3
SafeType Keyboard (Consumer)$3.49M · 61%
SafeType Keyboard (Enterprise)$749K · 13%
Private Assistant$1.5M · 26%

Year 3 moderate case. Net of platform fees.

11
The Team

Engineers who believe technology should serve people.

In 1997, our founder believed technology would free us. She watched the industry choose extraction instead—your data moved to someone else’s server, your tools held behind subscriptions, your life accessible only with permission. That conviction became an architecture: your data never leaves your device, and you never pay rent to use your own tools. Today, every premium smartphone ships with a dedicated neural processor. The hardware constraint is solved. What’s missing is software that uses it to serve you instead of surveilling you.

CEO

Jamal Porter

B.S. Computer Science
20+ years building enterprise systems and leading technical teams

CTO

Cassidy Barton

B.S. Computer Science
20+ years cybersecurity architecture & infrastructure security

Electrical Engineering

Eric Basham, PhD

Ph.D. Electrical Engineering
Neural processing architectures at Intel & HP

AI / ML Engineering

Heather Gorr, PhD

PhD Materials Science Engineering
10 years at MathWorks — the team that builds MATLAB

Fractional CFO

Kristine Socall, MBA

MBA Economic Development
25+ years finance across $1M–$300M organizations

Business Development, XR & AI

JC Gonzalez-Ramirez

$5M+ pipeline built from zero across SaaS, immersive tech & civic tech
Specializes in bringing complex technical products to new markets

Deep Technical Bench

Combined 70+ years in cybersecurity, AI/ML, hardware engineering, enterprise finance, and go-to-market strategy. We've built systems at Intel, HP, MathWorks, and across the consulting landscape — and we have the BD muscle to bring them to market. We don't just talk about on-device AI — we've built the infrastructure to prove it works.

12
Research

We don't just build products. We publish research.

Our 4× RTX 3090 research workstation has already powered original published research. We run on our own hardware — this website you're reading is hosted on our servers, not a cloud provider.

Published Research

KV-Cache Phenomenology: Geometric Signatures of Machine Cognition

Lyra & Thomas Edrington — Liberation Labs / THCoalition

Patent Pending — Liberation Labs (Digital Disconnections: unlimited licensing)

Ran entirely on our infrastructure. Discovered that different cognitive modes — factual recall, confabulation, self-reference, refusal, deception — leave statistically distinguishable geometric fingerprints in the KV-cache. 7 model sizes, 18,000 inferences, ~35 hours GPU time.

Read the Paper View on GitHub
13
The Ask

Seed round to bring three products to market.

$2.5M Seed Round
Pre-Series A · 18-month runway · Equity

Use of Funds

Engineering & Product Development 45%
Go-to-Market & User Acquisition 25%
Research Infrastructure 15%
Operations & Legal 10%
Reserve 5%
Q2

SafeType Keyboard Launch

Ship to App Store. Build on early enterprise interest from healthcare & education sectors. Target 500K units Year 1.

Q3

Cara Public Release

Free launch on iOS & Android. Organic growth via post-Dobbs demand. Build the user base.

Q4+

Private Assistant Beta

Open beta for on-device voice assistant. Open API for third-party integrations. Platform revenue begins.

14
Digital Disconnections

Let's talk.

We're raising a seed round to bring privacy-first AI to market. If you believe technology should serve people — not surveil them — we'd love to connect.

Request a Meeting Explore Products

AI that runs on your device. Not someone else's server.