Agentic Coding

301: The Step Up

How does one human run a fleet? The first two years teach single-agent agentic coding (SAAC): one human and one agent, first as habits and then as a method at project scale. This chapter steps up to multi-agent agentic coding (MAAC), where one human runs several agents at once. It teaches the step as a change in the human's job, from doing the work to arbitrating it, with fluency with one agent as its entry condition. From there it composes a fleet out of lanes of the work and prices it in commits, in disk and in the machine it runs on. The last insight follows one method as it moves between the estates one human runs. An estate is a project's repository and the nodes that work in it, and each estate's own validation node carries the method in.

The chapter opens the third year. Its exercises add the evaluation framework's seventh dimension, multi-agent coordination, to the six scores kept from 104: Checking the Agent's Work, and it carries the Lamplight fleet case study.

Materials

This Chapter's Insights

Browse every insight on the course map.

Downloads