A Korean solo developer in Seoul building AI apps with the honest split. Your saves stay on the phone. No user accounts. No sync. Only a small metadata payload travels for AI sorting — never the content body of your saves.
See Our First App →Share from any app. CatchNote grabs it. Done in under 1 second.
AI reads your content and files it into the right category. No manual tagging. No folder setup.
The save itself — article body, screenshot, full note — stays in SQLite on your phone. For categorization, the app sends a small metadata payload (URL, title, snippet) to an inference API. No accounts, no sync, no telemetry. We do not say "100% on-device" because it is not — and we would rather defend the true line.
Resurfaces forgotten saves when they become relevant. Your past self helps your future self.
Currently in private beta. Store launch coming soon.
See the full pipelineSave anything from any app. AI sorts it. Saves stay on your phone — only metadata travels for the AI sort.
Point your camera at food. Get instant calorie and nutrition data. Vision AI, no manual logging.
Phone lock for sleep. Stop doomscrolling before bed. Out soon.
Camera-based halal food analysis. Read ingredient labels in seconds. Built with cultural review.
One person. One mission. Just shipped.
OrangeAI is the label for apps built by a Korean solo developer in Seoul. The data posture is the honest split: user content stays on the phone, and only minimal metadata travels to an inference API for AI features. No accounts. No sync. No user-data backend.
Our philosophy: ship fast, validate with real users, and double down on what works. Every app starts as a PoC with a clear gate — 1,000 installs, 4.3+ stars, 50% activation in 90 days. Pass the gate, and we invest. Miss it, and we maintain.
We picked the smallest privacy claim we can defend for years over the maximalist one we would have to walk back. We do not say "100% on-device" because some inference is remote. We do say: the content body of your saves never leaves the device, and we have no user-data backend. That sentence survives a packet capture, which is what matters.
Saves stay on the phone. AI categorization is a metadata-only inference call. We do not run a user-data backend, and we do not claim what we cannot defend. Privacy by what we ship, not by marketing.
Every app launches as a PoC. Real users give real feedback. We iterate based on data, not assumptions.
No feature bloat. Each app solves one specific problem and does it well. Complexity is the enemy of daily use.