Hi HN — I built QuickSummarize, an open-source Chrome extension for working with YouTube videos through captions. The core idea is to make video Q&A transcript-first instead of treating a one-time summary as the whole memory. Current workflow: - summary - transcript-first chat - timeline browsing - subtitle export It runs in the Chrome side panel and currently focuses on YouTube. A few implementation choices I cared about: - transcript is the main source of truth for follow-up questions - bring-your-own API instead of another hosted subscription - support for both OpenAI-compatible APIs and Anthropic-style APIs - English / Chinese UI Still early, but it already feels more useful to me than the usual one-shot “summarize this video” flow. Would especially love feedback on: - whether transcript-first chat is actually the right UX - how people think about long-video memory/retrieval - caption reliability edge cases