PAUSE.

Inspiration

Online lectures and long youtube videos are a bit too good at turning into background noise. We wanted a way to force active listening without rewinding, note-spamming, or pretending we’ll “review later”. Pause. was inspired by the simple idea that being interrupted at the right moment can actually help you learn, not annoy you. Think pop quizzes, but make them friendly, unexpected, and a little chaotic.

What it does

Pause. is a chrome extension that randomly pauses a youtube video and quizzes you on what you just watched. Once activated on a video, it waits for a random timestamp, pauses playback, and overlays a clean, animated quiz directly on top of the video. You answer the question, get instant feedback, and then continue watching like nothing happened. The goal is to keep viewers engaged and accountable while watching educational or informational content (and are constantly expecting the next Pause. moment in fear).

How we built it

Pause. is built as a chrome extension with a python backend that works hand in hand with Gumloop’s API. On the frontend, a content script detects the active youtube video, tracks playback time, and schedules a quiz pause at a randomized but reasonable moment. When that moment is reached, the video pauses and a quiz overlay is injected directly into the page.

On the backend, we fetch the youtube transcript and align each transcript segment with precise timestamps. This was critical, since every quiz is generated based on the content watched up to that exact moment, not the entire video. We send the transcript slice and timing context to Gumloop, which structures the content and generates relevant quiz questions that actually reflect what the viewer should know by then. The generated quiz is returned to the extension in real time and rendered instantly, creating a seamless “pause and test” experience.

Challenges we ran into

The biggest challenge was reliably accessing and working with youtube transcripts. Between availability issues, formatting, and edge cases, this part took far longer than expected. We also had a steep learning curve getting familiar with Gumloop’s workflows and understanding how to structure inputs and outputs properly. Integrating multiple moving parts while debugging chrome extension quirks definitely tested our patience, sanity, and sleep schedule.

Accomplishments that we're proud of

We’re proud of building a system that tightly synchronizes video playback, transcript timestamps, and quiz generation without breaking the viewing experience. Getting the “material seen up to now” logic right required careful timing and transcript slicing.

We’re also proud of how effectively we integrated Gumloop into our pipeline. It became the backbone of our quiz generation workflow, allowing us to focus on logic and experience instead of prompt engineering. Turning long, messy transcripts into well-structured, context-aware quizzes felt like magic once everything clicked.

What we learned

We learned how important timing and context are when building AI-powered educational tools. A quiz is only useful if it matches what the user has actually seen, and getting that right forced us to think carefully about transcript alignment, timestamps, and edge cases.

We also learned how powerful Gumloop can be as an orchestration tool. Instead of treating AI as a black box, we learned how to design clear inputs, structure outputs, and build reliable workflows around them. Also, we learned that browser extensions are humbling, youtube is chaos, and that nothing builds character like debugging async logic at 3 a.m.

What's next for PAUSE.

We plan to support multiple difficulty levels and smarter timing based on content density. We also want to explore audio feedback and accessibility features, and eventually expand beyond youtube to other video platforms. Pause. is just getting started and it’s already refusing to let us zone out.