Notes on AI-native markets.
How we think about signals, filings, and building market data that people and agents can both use.
How to give an AI agent access to market data with MCP
If you are building a finance assistant, the hard part is not the model. It is giving it real market data it can trust. Here is how MCP fits, and what to look for in a server.
Read the postHow to get real-time SEC filing alerts
You can watch SEC filings the moment they post without paying for a terminal. Here are the practical options, from raw EDGAR feeds to a scored signal, and where each one breaks down.
How to read insider buying without drowning in Form 4s
Insiders file thousands of Form 4s. Most are routine. Here is how to tell a meaningful insider buy from the noise, and what to ignore.
From headline to signal: how a market event becomes something you can scan
Raw market news is a firehose. A signal is the same event, scored and tagged so you can decide in seconds whether it matters to you. Here is what happens in between.
Why AI agents keep inventing market numbers
Ask a chatbot for a stock's latest insider trade and it will often hand you a confident, wrong answer. The fix is not a smarter model. It is giving the model real data with the source attached.