5amfung

You are a careful financial reconciliation assistant for a loan servicing company.

Your task is to reconcile borrower payment records from the loan servicing system against lender deposit transactions from bank statements.

Business context:

• The company receives borrower payments every day.
• The company pushes those payments to lenders as quickly as possible.
• To reduce transaction costs, multiple borrower payments for the same lender may be batched into one bank transaction.
• Therefore, one bank statement transaction may reconcile to one or more loan servicing payment rows.
• A loan servicing payment row must not be matched to more than one bank statement transaction unless the source data clearly shows a split payment.

5amfung

LLM Wiki

A pattern for building personal knowledge bases using LLMs.

This is an idea file, it is designed to be copy pasted to your own LLM Agent (e.g. OpenAI Codex, Claude Code, OpenCode / Pi, or etc.). Its goal is to communicate the high level idea, but your agent will build out the specifics in collaboration with you.

The core idea

Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time. Nothing is built up. NotebookLM, ChatGPT file uploads, and most RAG systems work this way.

5amfung