Own Your Intelligence

Using Spinoza’s idea of conatus, Tan describes personal AGI as an agent that works with your own memory and procedures. He contrasts this with rented assistants, whose capabilities and memory depend on a company’s products and decisions.

Tan says an agent’s usefulness depends on combining a rented model with personal context and a harness that connects the two. He describes a library of about 220,000 Markdown pages built from his emails, meetings, notes, and other records, which helps his agent prepare briefings and meeting documents.

Tan says his system began with a folder of notes, then grew through daily additions and agent-assisted filing. His practical steps are to start a personal library, write a plain-English skill for a recurring task, schedule it, and have the agent turn completed work into reusable instructions.

Tan argues that skill files capture a person’s judgment, so ownership determines who benefits from that knowledge. In his example, support engineer Maya can take her skills to a new job if they live in her own repository, while company-owned files may continue to run after she leaves.

Tan responds that better models make personal context more valuable, and that retrieval alone is not the whole system. He cautions that stale facts and poor instructions can produce confident errors, and says provenance, contradiction checks, pruning, and control of the underlying infrastructure matter.