Overview
I’m writing a Murder, She Wrote Quiz & Activity book. It’s in progress and has taken a lot of research — but it also pushed me into building new tools in transcription, voice-print matching, and reasoning. The book will be on Amazon once it’s done.
How It Works
The toolchain transcribes episodes in batch, matches the cast — using voice-print matching to pin down who is actually speaking — finds act breaks, inventories characters, and audits the transcripts, so a quiz question can be checked against what happens on screen rather than against my memory or a wiki. It also runs “closed-book” reasoning and compares how different models transcribe and reason about the same material, writing a report on each.
A Paper Personality Test
One of the activities is a personality test — in the spirit of the “which princess are you” Royal Disposition Inventory — that a reader can run entirely on paper, no computer needed, at least for getting their final result. The heavy pre-calculation that makes a paper-computable scoring scheme possible, on the other hand, took a lot of computer.
So… Is It Work?
It kept the big server busy all week. And the funny thing is: because it turned into a multi-model comparison of transcription and reasoning, with written reports on each, it’s actually quite work-related after all — a quiz book that accidentally became a model evaluation.
Current Status
The toolchain runs and has chewed through a lot of episodes; the book around it is still early days, headed for Amazon when it’s finished.