For twenty years, corporations have run classification pipelines, longitudinal pattern detection, and intent inference on your browsing data. They cluster your visits, detect your purchase intent, build attention profiles across sessions, and sell the output. The analytical techniques are well-understood. The constraint was always access to the data and the cost of running inference.

Local LLMs remove the cost constraint. Browser extensions provide the data access. Memento tests whether these corporate analytical techniques, applied to data you collect about yourself, produce intelligence you can actually use.

The test domain is browser sessions. A Chrome extension captures your open tabs. A four-pass LLM pipeline classifies them, maps cross-category connections, and generates a structured session artifact. A longitudinal layer tracks patterns across sessions. A correction system feeds your disagreements back into future classification.

Whether these techniques transfer to other self-collected behavioral data — git commit patterns, communication logs, reading history — is an open question that motivated the architecture. The pipeline is designed to be domain-agnostic. Browser tabs are the first test case, not the point.