It’s been a few days of trial and (mostly) error.

I fell into the classic trap: I built a "God Agent."

One massive AI prompt to rule them all—save the data, check for duplicates, summarize, and report. Total disaster. It’s hallucinating reports and forgetting the database logic.

Note to self: just because an LLM can do everything doesn't mean it should.

Back to the drawing board.