End of day worklog: voice-driven daily retrospectives with AI

Here's something I've noticed about my own work habits: the easiest way to remember what I'm working on is to just keep working. But that's a terrible organization system, and it leads to working way longer than intended. I needed a way to cleanly "park" the workday — capture where I am on everything, and pick up smoothly the next morning.

The concept

End of Day Worklog is a voice-driven workflow designed around a simple routine: at the end of the day, you record a voice memo describing what you worked on, what progress you made, and what's blocking you. The audio gets processed through Gemini's multimodal AI, which generates an organized markdown document (or Google Doc) with projects categorized by type. The next morning, you get an email with a "start-of-day brief" that links to both the document and the original audio.

Why voice?

At the end of a long work day, the last thing I want to do is sit down and type out a structured summary of everything I did. But talking through it? That takes two minutes and feels natural. Voice memos capture the nuance and context that you'd normally lose — the "oh, and I also noticed that thing is broken" observations that you'd never bother writing down but are actually important.

Using Gemini's multimodal capabilities means the AI can work directly with the audio rather than relying on a separate transcription step, which tends to preserve more of the original meaning and context.

Current status

This project is in early planning stages. The repo includes a technical specification, the original voice note that sparked the idea, and both verbatim and cleaned transcripts. It's one of those concepts that emerged from a real friction point in my daily workflow, and I'm looking forward to building it out. If you have a similar problem with end-of-day context capture, keep an eye on the repo.