DATA HAS GRAVITY
Every transfer adds a network, processor, log, policy, and failure point. Obsidian Ridge Labs keeps core AI close to private data whenever supported Apple hardware can do the work.
- Core path
- On-device
The glasshouse isburning.
We are living in a surveillance economy built at planetary scale. It did not arrive as a cage. It arrived as convenience. Conversations, finances, memories, habits, and relationships became the price. Convenience was the bait. Private life became the catch.
Read the countermeasuresObsidian Ridge Labs is building the exit: powerful Apple apps with core intelligence on your device.
01 / Threat modelThe convenience trap
The panopticon
The cloud is someone else's computer. Your private life should not be its inventory.
A remote AI workflow can create more copies, more logs, more retention questions, and more companies to trust. That may be a fair trade for some work. It should not be the default price of transcribing a meeting, understanding your finances, writing in a journal, or remembering someone you love.
Cloud-first pathPrivate data takes the long way around.
Four trust points- 01Your inputCreated on your device
- 02Network transitSent beyond the device
- 03Remote processingHandled on outside infrastructure
- 04Result returnsDelivered back across the network
Local-first pathIntelligence moves to the private data.
One clear boundary- 01Your inputCreated on your device
- 02On-device intelligenceProcessed on supported Apple hardware
- 03Your resultReady without sending the content to us
Fewer copies. Fewer unknowns. More control.
If you think you have nothing to hide, you are not looking closely enough.Privacy is not secrecy. It is the right to decide who gets to look.
02 / CountermeasuresFour rules for private software
The counter architecture
Four refusals.
One private standard.
These rules decide what belongs on the device, what may connect, and what the user must always control.
THE CLOUD MUST EARN ITS PLACE
A connection should exist only when it delivers a capability the device cannot provide well on its own. Core work stays local. Optional services are named before they are used.
- Network access
- Purpose-bound
OFFLINE IS THE TEST
The network should add a specific capability, not control the core experience. After required setup, important work should continue without Wi-Fi or cellular service.
- Core workflows
- Offline-ready
MEMORY BELONGS TO YOU
Personal software should not keep more than it needs. Storage, export, retention, and deletion should be clear and controlled by the person who created the data.
- Retention
- User-controlled
03 / Counter architectureIntelligence that stays close
Your AI. Your phone. Your business.
Private intelligence.
On your terms.
On supported Apple hardware, app-bundled models and Apple frameworks process private inputs close to where they were created. The result is private, resilient software that can keep working offline and does not need our servers to understand your content.
01Private inputAudio · writing · records
02Local intelligenceApple silicon
03Useful resultReady where you are
- Core intelligence
- On-device where supported
- Core experience
- Offline-ready after setup
- Advertising profiles
- None
04 / The boundaryConnection is exceptional, never hidden
No hidden exits
If data crosses the boundary, you should know why.
Some features may connect to Apple, Plaid, or another named service for a specific job. Model downloads, purchases, optional iCloud sync, and optional bank sync are disclosed in context. The core product does not become an excuse to collect your content.
Read the privacy model- Core intelligence
- Processed on supported Apple devices
- Local by design
- Optional services
- Named before you enable them
- Your choice
- Diagnostics
- Where offered, opt-in and off by default
- Controlled
05 / ProofA standard you can experience
Proof in use
Privacy should be felt in the product.
Core work should remain useful when the network disappears. Optional connections should have a clear purpose. Personal content should never become the raw material for an advertising business.
06 / Philosophy in practice9 private tools. One standard.
Built by Obsidian Ridge Labs
Your life is not raw material
for someone else's cloud.
07 / Straight answersQuestions private software should answer
Before an AI app earns your data
It should answer these questions.
“On-device” and “private” matter only when the boundary is specific. These are the questions Obsidian Ridge Labs answers so people can choose with confidence.
01Why is on-device AI more private than cloud AI?
When AI runs on your device, the core input can be processed without creating a second copy on a remote AI server. That reduces the number of systems, transfers, and retention policies you must trust. It does not remove every risk, but it creates a shorter and easier-to-explain data path.
02Is Apple Intelligence completely on-device?
No blanket answer applies to every Apple Intelligence feature. Apple explains that many requests are processed on-device, while more complex requests may use Private Cloud Compute. We document the behavior of each Obsidian Ridge Labs app separately instead of treating a platform label as a universal privacy guarantee.
03Does local-first software ever send data off the device?
It can, when a person chooses a connected feature or when setup requires it. Examples include model downloads, App Store verification, encrypted iCloud sync, or optional Plaid bank sync. Local-first means the core path is designed around the device and every exception is disclosed in context.
04Is offline AI less accurate or less capable than cloud AI?
A focused local model can be faster, more private, and available offline while still outperforming a larger cloud model on a specific task. Echo Chamber is one example: its Parakeet TDT model records lower WER than Whisper large-v3 on the cited public English benchmark. Obsidian Ridge Labs chooses narrow workflows where local intelligence can deliver excellent results rather than chasing general-purpose breadth.
05Why build private AI apps only for iPhone, iPad, and Mac?
A focused Apple platform strategy lets us design around Apple silicon, the Neural Engine, Secure Enclave-backed controls, native accessibility, and system privacy APIs. It also lets us publish precise compatibility requirements instead of designing to the lowest common denominator.
06How can someone verify an app’s privacy claims?
Read the product-specific privacy documentation, inspect permissions and network disclosures, review public source where it is available, and check whether the app still performs its core job offline. We make those verification paths visible because trust should survive inspection.
07Do private AI apps require an account?
The core Obsidian Ridge Labs experience does not require an Obsidian Ridge Labs account. Apple services, purchases, optional sync providers, or support channels may have their own identity requirements, which are separate from an account with us.
08 / Further readingThe thinking behind the work
Go deeper.
The Obsidian standard
Move the intelligence.
Not the private life.
Obsidian Ridge Labs builds Apple apps that keep core intelligence on the device, keep optional connections visible, and keep control with the person doing the work.