glimpsh is an experimental project exploring gaze input inside the terminal. The idea is to treat eye tracking as a secondary input signal alongside keyboard and mouse.

One motivating use case is multi-agent management. Using gaze to quickly switch focus between running agents or processes, inspect state, or trigger context-sensitive actions without constant keybindings.

I’m also interested in pushing on the quality of commodity eye-tracking software as an HCI tool. Webcam-based eye tracking is widely available but often inaccurate and underexplored in real interfaces. This project is a way to stress-test those tools in a demanding UI environment.

I’m curious about combining gaze with other high-bandwidth input, especially voice. Systems like Wispr suggest that voice plus gaze could work well together, for example in terminal multiplexors where gaze establishes focus and short commands trigger actions.

This is not meant to replace efficient keyboard-driven workflows. Many users are already extremely fast with keybindings. The goal is to explore whether gaze and voice can act as complementary tools when managing many concurrent agents or streams of information.

This is very early and experimental. I’d especially love feedback on:

whether gaze as a secondary input makes sense in terminal workflows

multi-agent or TUI use cases that might benefit

accessibility implications

thoughts on combining gaze and voice in developer tools

Happy to answer questions or dig into implementation details.