Blog
Writing on local-first knowledge bases, document compilation, and the limits of RAG.
Elicana Is Available Today
Elicana compiles your document archive into a knowledge base you can query in plain language. It runs on your own machine, and it is available today in open beta.
The argumentNadella Says You're Paying Twice. Do You Know If You Are?
You pay for intelligence twice: once with money, once with the context you rebuild every session. You do not need to be an enterprise for that to be true of you.
The data proofI benchmarked plain RAG against corpus compilation on my own vault, here's what the numbers show
Four arms, one 181-document corpus, same writer, third-family judge. Compiled arms roughly doubled raw-chunk RAG on cross-document questions. On dated-document chains, raw RAG collapsed to 0.03.
Founder storyI Read the GBrain Post. Here's What I Built Instead.
GBrain is the builder's answer. Elicana is the same move, compile before you retrieve, for people who want the result without owning the pipeline.
Vocabulary driftWhy AI Tools Fail at Your Document Archive
Enterprise search broke on dirty data for two decades. RAG is breaking on the same rock, and for the same reason.
Consultant's guideThe Consultant's Document Audit: 5 Questions Your Knowledge Base Should Answer (and Why It Can't)
Five questions your knowledge base should answer instantly, a diagnostic you can run for each, and the structural reason most can't.
Founder storyDreaming teaches agents to learn from sessions. What about everything you wrote before it?
Dreaming is scoped to sessions. Most of us carry a much larger body of material, accumulated over years, that stays untouched no matter what any agent remembers.