Inspiration By the time my AI-generated study notes were ready, I’d already taken my first study break.
The part that frustrated me was everything around the prompt. I’d find lecture files in different folders, upload them, explain how I wanted my notes organized, download the result, and move it out of Downloads and into my course folder. Sometimes the PDF had formatting errors, so I’d ask for the LaTeX source and compile it myself in Overleaf. I built Late Night Pal because I wanted to spend that time studying. What it does Late Night Pal is a desktop study sidekick that works inside a folder you choose, called a workshop. You can select files across several subfolders, choose what you want to make, and save the finished PDFs directly where you need them.
My favorite feature is the daily study plan. If I have two days before an exam, I can select my course materials and ask Late Night Pal to split them across those days. I get a folder for each day with its own notes, practice questions, and checklist. When I sit down to study, I know what to open first. How I built it I built the desktop app in Python with PySide6 and connected it to the Gemini API. Each of the eight study tasks has a reusable prompt, so students can choose an action like Make sample questions or Create a one-page cheat sheet. Gemini returns structured study content. The app checks that the expected documents are present, compiles them into PDFs locally with LaTeX, and saves them in the workshop. Challenges and lessons One bug showed up just as I thought a generation run had succeeded. The PDF was saved correctly, but the window froze because the background job was waiting for itself to finish. Fixing it taught me to check the whole experience, including whether the student can start another task after opening the first result.
I also learned to measure where the time goes. In two study-plan runs, roughly 97% of the wait came from the Gemini request. Compiling six PDFs locally took about 14 seconds. The clearest improvement for me was how much less I had to do during that wait: one click handled the prompting, formatting, and filing that used to interrupt my study session.
I originally wanted to add tools for reorganizing entire course folders. With eight hours to build, I focused on the part I’d actually open the app for the night before an exam: turning the materials I already have into something I can study from.
