AI Synthesis & Source Evidence
Creator Advisors
Private Desktop App
A private reader that organizes ideas from long-form videos into topic pages, with quotations linked to their source timestamps.
Local Knowledge App · Private Source
Key DecisionCompute quotation locations from the source instead of trusting generated timestamps.
The Case Study
- Problem
- Useful ideas are scattered across long videos. Summaries are easier to skim, but can lose context or point to a timestamp that does not support the claim.
- What I Built
- A pipeline that turns transcripts into source-linked topic pages, with a desktop and browser reader. Different transcription adapters feed the same pipeline; matching quoted wording to the transcript produces the citation locations.
- How It Works
- AI assists with organization and synthesis. Import checks require quoted wording to match the transcript and calculate timestamps from those matches. The reader keeps the source accessible beside the interpretation.
- Where It Stands
- Implemented for personal use with multiple creator libraries. This case study describes the workflow without publishing the private library, transcripts, or source code.
- Evidence and Limits
- An import dry run caught altered capitalization in quoted wording and required correction before import. Source checking constrains the AI output.
- A matching quote establishes traceability, not the truth of the speaker’s claim or the completeness of a summary.
- No claim of measured learning improvement or fully autonomous research is made.
- Product Stack
- TypeScript · Node.js · Electron · Transcript Processing · Markdown
- Evidence As Of
- 2026-10-27