Welcome to the Lyceum
Aristotle’s Lyceum taught while walking: knowledge pursued in motion, with the student shaping the dialogue alongside the teacher. This site borrows that arrangement for software that no longer only calculates but acts. Lyceum Agents teaches the mechanics of agentic AI systems, and it treats technical design and ethical responsibility as one discipline rather than two.
Once software acts on its own, structure decides consequence. Which tools an agent may reach, what it has to confirm with a person, and what it may never do unattended are architecture questions with someone on the other end of them, and they are answered in the same file that says how the thing works. That is why both subjects sit on the same page here.
the architecture of this space
Three choices shape how you read here. The first is interaction in place of static media: where another site would embed a diagram, this one hands you a live dialogue to steer or a terminal to command. The second is the shell as the front door. Press / anywhere on the screen and type help, and from there you can ls the posts, open any article by its slug, and find tools that no menu lists. The third is that the material compounds. Each post is one day of a curriculum still being written, and the essays accumulate into a textbook you can read in order.
Watch the machinery answer a reader’s request:
the curriculum ahead
Three threads run through the site:
- Foundations: We examine the core mechanics of AI agents. This includes planning loops, tool integration, memory structures, and the orchestration of multi-agent systems.
- Ethics: We define the boundaries of delegation and accountability. We explore alignment in practical terms and establish what we owe the individuals positioned on the receiving end of an agent’s actions.
- Tools: We construct open-source instruments parallel to our writing. This begins with melchizedek, a framework for multi-model, multi-agent orchestration.
Start with the curriculum. Part 2 opens with Generative AI agent design, which builds the definition of an agent through a conversation you steer rather than a summary you read.