Building My First RAG Knowledge Platform
What I learned stitching together retrieval, local models, and a useful interface for real questions.


What I learned stitching together retrieval, local models, and a useful interface for real questions.
What old transaction flows taught me about trust, reconciliation, and the cost of a small change.
The technical debt is only part of the story. The people, workflows, and invisible contracts matter too.
A narrow task, a clear source, and a quick review can be more useful than an ambitious assistant.
A practical look at where an AI copilot helps most, and where the human context still does the heavy lifting.
How short decision records help the next person understand what changed and why.
Notes from opening up unfamiliar codebases, finding the seams, and making change feel less risky.
A few habits for finding the important behavior before trying to improve the implementation.
A few durable lessons about infrastructure, registration flows, and designing for the unhappy path.