Research-driven writing on investing, markets, and AI — evidence-first, deep over frequent, and written to sharpen how you think rather than to tell you what to buy. No signals, no urgency: the reasoning is the point.
- 2026-06-06Market structure
SPY high-volume red-day analogs and ES tactical levels
What sixty completed historical analogs of a high-volume red day actually say about the days that follow — a worked example of reasoning from a base rate instead of a headline.
Read essay → - 2026-05-27AI & research tooling
Indexing a 10-K with Claude vs. a local LLM
Five indexing backends on the same filing, then a statistical rerun across eight filings and 48 queries — showing where the model choice matters for reading filings and, more interestingly, where it barely moves the answer at all.
Read essay → - 2026-05-18AI infrastructure
Leopold Aschenbrenner's 13F is not "short AI." It is a bottleneck map.
A primary-source read of a widely misread 13F: the reported put book against semis and infrastructure sits alongside major long exposure to power, memory, and data-center capacity. Reading the filing, not the narrative about it.
Read essay → - 2026-05-14Process, worked example
A week of portfolio commentary, start to finish
One week read through the whole process — regime, primary-source filings, and a hold/add/reduce judgment on each position. A demonstration of the method, not a set of calls to follow.
Read essay →
Clarion Research Public Sandbox →
Worked equity theses, the quarterly investor letter, and the live S&P 500 regime dashboard — the same surfaces the Clarion engine renders for its own monitoring, published openly to read.
Memory Market Monitor →
Daily DRAM retail prices with a rolling news feed — a small system built end-to-end (scraper, scheduled agent, API, and UI) as an example of turning a data question into a running tool.