Comment by XenophileJKO

So my current pet project is a to-do system with an app that you can look at, edit, or complete tasks in. But I have both a fully agentic interface and simpler LLM powered inputs.

I'm really enjoying it. I think it is a good example of how to leverage LLMs to reduce drudgery.

Things I can do now: - take a picture of a notice like a license renewal and a task is created and automatically filled with due dates and information extracted from the image and likely from online searches.

- turn a design document into a reasonable task plan.

- create classified and researched tasks with a sentence.

I'm just getting started on it but it already is kind of feature complete. Programmed with Claude Code, about 20k lines.

The key I think is to have something as easy to input as a text file, because it applies intelligence to remove friction.