the studio.
Studio DNA, privacy floor, HR-driven data security standards, and non-negotiable engineering guardrails governing every app built at calmdownoscar.
Apps Built to Put the Phone Down
Software exists to reduce screen time and decision fatigue. Success for a calmdownoscar app means the user finishes what they came to do and puts their phone down — not maximizing session length, infinite feeds, or screen time.
Business Rails: Zero ads. No third-party trackers. One-time or lifetime purchases whenever possible. Never use urgency bait, notification spam, or artificial streak counters to hook users.
Personal Data Protection as a Priority
Coming from a background in Human Resources, safeguarding personal confidentiality is not a legal checklist — it is a foundational duty of trust. Handling sensitive personnel records translated directly into how software and AI architectures are designed here.
100% On-Device Processing
Sensitive operations (photo scanning, disk maps, voice notes) execute locally on Apple silicon. Data stays strictly on your device.
Zero 3rd-Party Analytics
No SDK trackers, no telemetry scripts, no third-party data broker libraries. Crash visibility uses Apple's native MetricKit.
Server-Side Auth Enforcement
UI hides controls, but server endpoints enforce strict per-record authorization. Client claims are never trusted blindly.
Zero Secrets in Client Binaries
API keys never ship inside a client binary. Cloud LLM calls route through a stateless relay (see Wedding Concierge) that holds the credential server-side instead.
How AI Features Are Kept Honest
When a model gets wired into an app, these are the constraints it has to work inside:
Single Source of Truth
The model and the native UI read and write to the same state — not two systems that can drift apart. Wedding Concierge was built specifically against that failure mode.
MCP Ground-Truth Memory
Unli Rice runs over Model Context Protocol (MCP) so every connected agent — Gemini, Claude, Codex, Antigravity — reads from one vault instead of re-explaining context each session.
Manual Pre-Mortem Review
Before an App Store submission, I walk through likely edge cases and regressions by hand. It's a manual pass today, not an automated eval suite — worth saying plainly rather than dressing it up.
Cloud Where It's Simpler, Local Where It Matters
Not every app needs both. Wedding Concierge currently runs cloud-only, no local fallback — that's a real tradeoff, not a hybrid claim I'm dressing up. ClearSpace goes the other way: photo review runs 100% on-device — a perceptual hash and Apple's Vision framework, not a cloud model — because that task is privacy-sensitive enough to warrant it.