DillyDally

See your focus. A productivity self-surveillance tool meant to reclaim your focus and illuminate how you use your time.

Comment

See your focus. Improve your focus.

Instead of just measuring how long you worked, DillyDally helps you understand how well you worked.

Inspiration

We initially wanted to build a trip-planning app named DillyDally, but when that didn't pan out, we still found ourselves drawn to the name. While looking for inspiration, we stumbled upon RescueTime, and imagined what it could be with modern AI. That thought brought us to what we now know as DillyDally. DillyDally builds on the ideas of RescueTime, utilizing AI insights and analytics to level up your productivity and focus. With AI, DillyDally delivers personalized, insightful breakdowns of your focus — helping you spend more time working and less time dilly-dallying. Image Image

What It Does

DillyDally runs a Pomodoro-style focus session, tracking your focus and engagement through computer vision, screen captures, and LLM insights. Once the session ends, DillyDally provides extensive data and information, including: total time focused and productivity scores, a visual timeline of your activities, activity breakdowns, and more.

Computer Vision

DillyDally uses lightweight, on-device computer vision to track when the user's gaze is on the screen. DillyDally utilizes @tensorflow/tfjs together with the face-landmarks-detection model to track the direction and stability of key facial landmark points, rendered through react-webcam. Image All processing happens locally in the browser, in real time.
No images or videos are recorded, transmitted, or stored, and only the classification and timestamp are stored, ensuring DillyDally remains secure and lightweight. image of main screen with video

Screen Capture

It is obviously not enough to know just where they are looking, but where. This is where screen capture comes in. DillyDally will periodically take screenshots of the user's screen and sends it to our express server backend. Once there, our express server uses an LLM to extract: whether the user was productive, what tab they were on, the name of the activity, and a summary of the user's actions. This information is then stored in the Convex database.

MCP Server

We also expose an MCP server, allowing interested parties to utilize LLM to interact with their DillyDally information and insights. This allows for even greater control and flexibility over their data, deeper analytics, and greater focus improvements. We are committed to enabling a self-surveillance solution for users who want to reclaim how their data is used online and on their devices. Dedalus's drop in MCP server hosting allows us to expose the DillyDally API to all users for use in their own devices. Image

What’s Next

  • Historical insights panel, enabling long-term, deeper analytics and insights
  • Reward-based mechanisms encouraging effective focus and engagement
  • Scaled to enterprises, for increased worker productivity, greater analytics, and improved performance tracking

Built With

Share this project:

Updates