The valuation compendium
What it does
Before the night projects, I spent twenty years valuing companies and assets. All of it was sitting on a cloud drive: 66,000 files, 372 GB, with copies of copies and folder names only I could decode. The 10,000 documents I could parse hold about 155 million words, a fair share of them spreadsheet cells. The compendium turns that pile into something I can read. It is a small private website with 55 topic notes in nine sections, from cost of capital to purchase price allocations to impairment testing, plus flashcards and a glossary of 868 English terms with their Portuguese equivalents. Search is instant and ignores accents.
Each note shows where its claims come from, lists the questions only I can answer, and has a "Your take" box for my own view. Guests read the notes and the methods. The source files stay with me, because many of them are client work.
Why I built it
Experience lives in your head and in folders you never open again. I wanted it in one place, searchable, in my own words, before the details fade. It is also a reference for a handful of people I trust, so it sits behind a login and is not linked from this page.
Stack
- A crawler that indexed every file and folder on the drive
- Parsers for modern Office files and for the old 1997-era Word, Excel and PowerPoint formats
- SQLite full-text search over 448,000 passages from 10,350 documents
- Claude to draft the notes from that corpus, me to check them
- Two views of every page, one for me with sources, one for guests without
- A small Python web server, run from home behind a login, that also collects answers from guest readers