I am Lucas.
Aegis V1.0 is now live.
It is a 100% offline, dual-AI voice input method based on a Unidirectional Communication Architecture, running entirely on local mobile compute.
We have cut the network connection to ensure Zero Upstream Data and Absolute Isolation. It runs only on your device.
Language Protocol
V1.0 currently supports ENGLISH ONLY.
To achieve the highest efficiency in local neural recognition, our current engineering is hyper-focused on the English language structure. We refuse to compromise performance for broad language support.
For custom neural model requests (Arabic, French, Russian, German, Chinese, etc.), please contact our development team for specialized commercial deployment.
Current Capability
We officially advertise a baseline accuracy of 82% on the local NPU. However, driven by our proprietary C++ FuzzyMatch engine and the 1-million-word domain lexicons, real-world accuracy consistently scales up to 92%.
Physical Boundary
Running offline means we cannot compete with the massive cloud computing power of GPT-4. This is not a defect; it is the physical reality we must face for Absolute Privacy.
Iteration Roadmap
V1.5: Deep kernel optimization to enhance local efficiency.
V2.0: Offline Workstation batch-processing architecture is in progress.
On Payment & Refund
Payment is your endorsement of the 'Offline Freedom' concept and a direct funding to our open-source research.
We strictly adhere to Google Play's standard automated refund policies. If the experience does not meet your hardware capabilities or expectations, simply use Google Play's built-in refund mechanism within their designated window. The system will process it automatically.
To Engineers
If you want to dance on the edge of mobile computing limits, welcome to the optimization challenge. Source code is available on GitHub.
— Lucas