~/curriculum

Six modules on designing, testing, and trusting AI agents. Most study a real agent team — a syndicate — from the Melchizedek repo: the YAML that defines it, the behavior it produces, the design pattern underneath; the last studies the system that built this course. You read the theory, read the actual config, then exercise it — scripted traces, terminal checkpoints, one hands-on exercise per module.

Who this is for: the practitioner who has outgrown the chat window. You already use AI daily — for work, for research, for the thing you are building — and you can run a terminal command and edit a config file. You do not need to be a software engineer, and the course will not try to make you one. It teaches the layer above the code: how to design, instruct, test, and trust a team of agents that serves your own domain — your patients, your portfolio, your research, your product, your cause.

Why this course exists: individual autonomy in the AI era — becoming better at building AI, and better at building with AI. Met through a chat window, generative AI is a superb way to search; commanded as agents, it drives efficiency, accelerates research, and makes new things under your direction. AI offloads the iterative work — which frees you to question paradigms, pursue an understanding, and build something new. But it is only as good as your ideas, and only as powerful as the hands that mold it. Every module here is a mold: a way of putting your judgment into a form the machine must obey.

Melchizedek is the specimen, not the syllabus. By the end you'll navigate the repo comfortably — but every pattern here (delegation contracts, critic loops, memory doctrines, inventory protocols, agentic coding molds) is framework-independent. The takeaway is what you can design on your own, with whatever stack you already use.

what you need

To take the course: nothing. No account, no keys, no install. Every exercise runs on this page; progress is a checkmark in your browser, set by you.

To run the specimens yourself (optional, recommended by module 3):

• Node.js 22+ — Melchizedek is a headless Node CLI; npm install, copy .env.example to .env, then npm run chat:syndicate starts a REPL.
• A Gemini API key — free — from aistudio.google.com; the free tier runs every text syndicate in the course, memory embeddings included. Module 05's image-generation model alone requires the paid tier. Each agent's model is one YAML line, so substituting another provider is a supported exercise, not a workaround.
• A Supabase project — free (only for the memory module's specimens) — two tables plus the pgvector extension; the repo ships the exact SQL, including the hardening script.
• Optional, paid: an Anthropic key — only if you point an agent at a claude-* model; usage-billed, no free tier. Nothing in the course requires it.

The full key-by-key cost table — including the paid market-data keys the private financial syndicates add — lives on the melchizedek tool page. Prefer to delegate the setup? melchizedek-agent-setup.md is a paste-ready prompt for your coding agent, with the human steps kept separate.

The repo is public: github.com/jhwadman/melchizedek-agents. Every syndicate studied here also ships as a verbatim YAML download inside its module, so you can follow the architecture line-by-line — and rebuild it anywhere — without cloning anything. Each module states what its specimen needs.

modules

[□□□□□□] 0/6 marked complete · progress lives in this browser only

1. [□] 01 Generative AI agent design ~30 min

   What an LLM actually computes, why it fails, and how grounding plus a persona turns a general model into a specialized instrument.

2. [□] 02 Agent workflows & conversation styles ~35 min

   Four workflow topologies, the delegation contract that wires them, and conversation style as an engineered surface rather than an emergent accident.

3. [□] 03 Testing & refining an agentic workflow ~30 min

   Prompting functions as an iterative dialogue. Master the five debug variables, deploy adversarial personas, and encode quality as machine-checkable critic loops.

4. [□] 04 Building memory systems for agents ~35 min

   Understand session state versus long-term memory, the critical judgment of distillation, and why fabricated recollection represents a system failure, not a convenience.

5. [□] 05 Multi-modal agents: predictable images, objective eyes ~40 min

   Spec-first image generation and the strict division of visual labor. We separate a blind inventory of what is present from the audit that judges it.

6. [□] 06 AI builds the feature ~35 min

   The agentic coding loop relies on rules files, action-driven skills, knowledge documents, and plan artifacts that strictly gate the build process.

try the controls

Checkpoints throughout the course look like this: type the letter of your answer and press enter. Every option gets a real explanation — wrong answers teach; nothing is scored.