You are reading a single comment by @stweedo and its replies. Click here to read the full conversation.
  • There was talk of being able to use your phone's mic for input. Would need some kind of speech to text on the phone then.

    In lieu of a mic or speaker, it might be useful to make a predictive text input that contextually predicts chunks of words it thinks you might want to enter into your prompt and present them as a list of options to append to your prompt.

    Initially it could start with generic "What, when, why, how, where" etc. and probably would need at least one or two other seed words that are input by the user with a keyboard app to give better context.

    Here's the basic thought process on how it might work:

    1. The Watch: Display the initial list of options. When an option is selected, send the selected word to the server.
    2. The Server: Receives the selected word from the watch. This selected word is used as a seed to generate new potential words/phrases using an LLM. The server then sends the generated words/phrases back to the watch.
    3. The Watch: Receives the new words/phrases from the server and displays them as new options. Once the user is satisfied they can send the whole prompt to be answered like normal.

    Back on topic... I thought it would be nice for Bangle.js 3 to have an LED indicator that could double as an ambient light sensor. Similar to how the LED is being used on this open source chirp device.

    From the description:

    Light sensing
    
    A simple LED is used as a light sensor. Capacitive properties of diodes are used. LED is driven forward-biased for some time, then it’s driven reverse-biased to charge the internal capacitance. The time it takes for this internal diode capacitance to discharge depends on the amount of light that falls into the diode. Microcontroller timer is used to measure this time and estimate how much ambient light is out there.
    
About

Avatar for stweedo @stweedo started