A personal knowledge wiki that collects links, tweets, videos, articles, and web pages into .md files and answers questions using direct context instead of retrieval pipelines.
Architecture
Problem it solves
Bookmarking tools and note apps become graveyards — you save hundreds of links and articles but never retrieve the right information when you need it because search is shallow and RAG pipelines lose nuance.
How it works
Everything you consume — links, websites, tweets, images, videos, YouTube transcripts, blogs, articles — is collected and stored as .md files in a personal wiki. The AI answers questions by reasoning directly over this knowledge base with no vector DB or chunking layer in between.
Approach
Built on the belief that modern LLMs with large context windows and agentic tools can handle entire knowledge bases at once. No retrieval pipeline — just feed the relevant .md context directly and let the model build understanding.