Zeke
Research workstation that scores content by signal-to-noise ratio and links every insight to its source.
Overview
A full-stack platform forked from Midday's finance codebase. It ingests YouTube, arXiv, RSS, and blogs; runs AI extraction jobs; and surfaces prioritized insights in a dashboard.
Problem
AI and marketing content is mostly recycled hype. Finding the few minutes that matter means wading through hours of noise.
Solution
A hype scoring algorithm ranks content by keyword relevance, highlight type, source authority, and freshness, with every insight linked back to a source timestamp.
Lessons learned
- Building content intelligence is easier than building distribution.
- Forking a mature product saved weeks of multi-tenant infrastructure work.
- Useful scoring models still need clear user-facing explanations.
- Technical completion and market conviction are different things.