I've talked to ~20 developers over the past week and the pattern I keep seeing is: everyone writes things down somewhere, almost nobody can find it when they need it. The problem isn't capturing — it's retrieval.
Notion, Obsidian, plain files, GitHub wikis, Confluence — all have the same issue: search is keyword-based and only works if you remember exactly what you wrote.
Curious if others have found something that actually solves this, or if you've just accepted that you'll Google the same problem twice.
> search is keyword-based and only works if you remember exactly what you wrote.
For whatever reason, this has never actually been an impediment to finding what I'm looking for. I almost always remember roughly where I put whatever it is that I'm seeking.
My Obsidian markdown notes are continuously indexed into a private Qdrant vector database (segmented on sectional headings). I also have a local Gemma3 LLM which takes any given markdown note and generates a series of theoretical questions that could be answered by the note.
Embeddings are created for those and placed into the vector database as well to create more valuable linkage. Lots of others probably use a similar combination of keyword, semantic, and RAG to alleviate this issue as well.