- Display size: 4.2-inch E-ink display
- Display technology: Custom high-refresh panel (contrast-optimized version)
- Resolution: 400 × 300
- Visual characteristics: Paper-like reading experience, partial refresh support, clearly visible in sunlight, monochrome black and white
- Main controller: ESP32-S3
- Memory: 16MB Flash, 8MB PSRAM
- Connectivity: Dual-mode Wi-Fi + BLE 5.0
- System architecture: Deep integration of standalone hardware with a cloud-hosted LLM
- PCB process: HASL finish + black solder mask
- Enclosure options:
- ☁️ Minimal Premium Edition: Injection-molded enclosure for individual users with high aesthetic expectations; designed to blend into the home with a refined “calm technology” look
- Everyday Edition: Identical functionality, suitable for DIYers who want to customize the software
- 🌈 Color Edition: Open-source model files are provided for enthusiasts to print themselves
- Battery: 2000+mAh rechargeable battery
- Buttons: Front voice-interaction/confirm button; side page-turn/function/power buttons
- Port: Type-C (supports charging and data communication)
- Other peripherals: Speaker, microphone, NFC, and status indicator
- Exposed expansion header: 3.3V/GND/TX/RX/2 GPIO
- Mounting: Magnetic backplate; desktop stand design to be determined
- Overall dimensions: 97mm × 97mm × 8.75mm
- Standard mode: Ready to use out of the box, with AI voice memos and information dashboards
- Developer mode: Open-source low-level drivers and interfaces for secondary development; the C++ version will be prioritized, followed by a C version
- Power strategy: Microamp-level deep sleep for extra-long standby time; supports press-and-hold power on/off
- Software updates: A relatively convenient, ongoing software update method will be provided for standard users
- To-do processing: Uses a Qwen3.5 model agent by default
- Pricing policy: Standard users in the first crowdfunding batch receive enough credits for one year of free use; subsequent devices include lifetime AI access
- The platform currently does not allow users to connect custom models. An open-source version for self-hosting by enthusiasts is planned for the future