Notes
Research and field notes.
Speculative research, event studies, and working notes. Nothing here is investment advice — data, code, and known limits.
Getting better with AI.
Stop chasing the newest tool. Pick one and master it.
Each tool has a shape. The market shifts perception every three months. A workflow for using AI to actually get smarter — eval criteria, context, steering docs, and why Projects is a layperson's RAG.
AI Rush.
The picks-and-shovels watchlist for $400B/yr in hyperscaler capex.
Fifty public companies that capture hyperscaler AI spend, organized by where they sit in the stack and scored against my 7-step framework. Live and updated as I deepen each name.
The RIF detector.
What AI layoffs actually do to stock prices.
Most AI-blamed layoffs don't move the stock — but the ones that do pop 15 to 40 percent in a day. The pattern, a 12-company watchlist, and the honest limits.
Kimi 2.6 — cheap, fast, real questions.
Ten days with Moonshot's coding model.
A B+ that moves fast and ships surprisingly solid output for the price. The jailbreaks are real, the jurisdiction is Beijing-adjacent, and the privacy story deserves more attention than the benchmark scores.