Invisible ink: what an AI summary leaves out of an annual report
Machine summaries of annual reports silently drop conditions, negative disclosures and footnote context; the note shows how to notice what is missing before it misleads you.
The notebook, kept public
Research notes
The notes below are the substance of this journal. Each one takes a real task from an individual investor's reading week — summarising a filing, checking a model's claim, keeping notes that can be audited later — and rebuilds it as a workflow that treats AI as a fast, fallible assistant.
Read them in any order. If a subject you need is missing, the question queue is the shortest route to it: ask the desk, and recurring questions graduate into notes on this page.
Notes that survive a chatbot: file your reading so future-you can audit it
A note format that keeps verbatim quote, plain fact and dated interpretation in separate layers, so any line of research can be audited later without re-reading.
The verification debt: checking what a model tells you about a company
A four-question checking pass the desk runs on every model-generated claim, plus the red flags that make verification debt visible from across the room.