140 lines
4.9 KiB
Markdown
140 lines
4.9 KiB
Markdown
# Curriculum
|
|
|
|
## Module Overview
|
|
|
|
| Module | Title | Hours | What You'll Learn |
|
|
|--------|-------|:-----:|-------------------|
|
|
| M1 | Foundations | 4-6 | What agents are, harness vs model, decision frameworks, trust |
|
|
| M2 | Architecture | 6-8 | Tools, loops, context, memory, skills, codebase patterns |
|
|
| M3 | Safety & Security | 5-7 | 6-level security ladder, hooks, verifier pattern, defense-in-depth |
|
|
| M4 | Orchestration | 7-9 | Multi-agent patterns, P-threads, delegation, P2P communication |
|
|
| M5 | Production | 5-7 | CI/CD, shadow deploys, observability, rollback, 5-tool stack |
|
|
| M6 | Economics | 4-6 | Model pricing, cascade routing, pass@k evals, cost optimization |
|
|
| M7 | Advanced | 5-7 | Autoresearch, meta-agents, beyond MCP, always-on agents |
|
|
| M8 | Capstone | 8-12 | Build a production-grade multi-agent system |
|
|
|
|
**Total**: 44-62 hours, 65 lessons, 13 labs, 56 quizzes
|
|
|
|
---
|
|
|
|
## Module 1: Foundations (4-6 hours)
|
|
|
|
| Lesson | Topic |
|
|
|--------|-------|
|
|
| 1.1 | What Makes an Agent? — Karpathy thesis, Software 3.0 |
|
|
| 1.2 | The Harness vs The Model — 5 subsystems, 4 dimensions |
|
|
| 1.3 | The Repository IS the Spec |
|
|
| 1.4 | Decision Framework: "Should I use an agent for this?" |
|
|
| 1.5 | The Agent Loop: Think, Act, Observe, Repeat |
|
|
| 1.6 | Tool Calling Deep Dive |
|
|
| 1.7 | Vibe Coding vs Agentic Engineering |
|
|
| 1.8 | Do You Trust Your Agents? — Hotz critique, Schmidt strategic view |
|
|
| Lab | Your First Agent (single-tool) |
|
|
|
|
## Module 2: Architecture (6-8 hours)
|
|
|
|
| Lesson | Topic |
|
|
|--------|-------|
|
|
| 2.1 | The Four Pillars: Tools, Loop, Context, Memory |
|
|
| 2.2 | Tool Design Patterns: MCP, CLI, Script, Skills |
|
|
| 2.3 | Agent Loop Variants: 5 levels |
|
|
| 2.4 | Skills System Deep Dive: path-scoped, plugins, LSP |
|
|
| 2.5 | Agent-Readable Workspace: init phase, feature lists |
|
|
| 2.6 | Context Window Management |
|
|
| 2.7 | Memory Patterns: mental models, scratch pads |
|
|
| 2.8 | The Reasoning Parameter |
|
|
| 2.9 | Codebase Architectures: 4 patterns |
|
|
| Labs | Multi-Tool Agent + Context-Aware Agent |
|
|
|
|
## Module 3: Safety & Security (5-7 hours)
|
|
|
|
| Lesson | Topic |
|
|
|--------|-------|
|
|
| 3.1 | Prompt Injection (L0) + Why Bash Is the Problem |
|
|
| 3.2 | The 6-Level Security Ladder |
|
|
| 3.3 | The L3 Marque Break |
|
|
| 3.4 | Damage Control: 3 access levels |
|
|
| 3.5 | Hook Architecture: 13 lifecycle events |
|
|
| 3.6 | The Verifier Pattern |
|
|
| 3.7 | Defense-in-Depth Stacking |
|
|
| Labs | Whitelist Hook + Verifier Agent |
|
|
|
|
## Module 4: Orchestration (7-9 hours)
|
|
|
|
| Lesson | Topic |
|
|
|--------|-------|
|
|
| 4.1 | Why One Agent Is Not Enough |
|
|
| 4.2 | Orchestration Patterns: Dispatcher, Pipeline, P2P |
|
|
| 4.3 | P-Threads: Parallel Agent Execution |
|
|
| 4.4 | F-Threads: Fusion, N Agents One Winner |
|
|
| 4.5 | Depth-2 Delegation |
|
|
| 4.6 | Agent Experts That Remember |
|
|
| 4.7 | Domain Locking |
|
|
| 4.8 | TillDone Task Discipline |
|
|
| 4.9 | Agent Chains |
|
|
| 4.10 | P2P Communication + MCP Agent Mail |
|
|
| 4.11 | Service Connectors |
|
|
| 4.12 | Conversation Awareness |
|
|
| 4.13 | CEO Board System |
|
|
| 4.14 | UI Agents System |
|
|
| Labs | Agent Chain + Multi-Team Config |
|
|
|
|
## Module 5: Production (5-7 hours)
|
|
|
|
| Lesson | Topic |
|
|
|--------|-------|
|
|
| 5.1 | What Production Means for Agents |
|
|
| 5.2 | CI/CD for Agents |
|
|
| 5.3 | The 5-Tool Production Stack |
|
|
| 5.4 | The Agent Manager Role |
|
|
| 5.5 | Shadow Deployments |
|
|
| 5.6 | Rollback Strategies |
|
|
| 5.7 | Observability + Cross-Provider Search |
|
|
| 5.8 | Alerting and Monitoring |
|
|
| 5.9 | Deployment Modes |
|
|
| 5.10 | Cost Control |
|
|
| Labs | Observability SQLite + CI/CD Pipeline |
|
|
|
|
## Module 6: Economics (4-6 hours)
|
|
|
|
| Lesson | Topic |
|
|
|--------|-------|
|
|
| 6.0 | The Compute Advantage Equation |
|
|
| 6.1 | Tokenomics: 3 Levels |
|
|
| 6.2 | LLM Pricing Landscape |
|
|
| 6.3 | Cascade Routing |
|
|
| 6.4 | Cost Per Session Math |
|
|
| 6.5 | Agent Evaluation Metrics |
|
|
| 6.6 | Automated Evaluation |
|
|
| 6.7 | A/B Testing Agents |
|
|
| 6.8 | Human Evaluation |
|
|
| Labs | Eval Harness + Cost Optimization |
|
|
|
|
## Module 7: Advanced (5-7 hours)
|
|
|
|
| Lesson | Topic |
|
|
|--------|-------|
|
|
| 7.1 | Autoresearch: Self-Improving Agents |
|
|
| 7.2 | Integrity: Keeping the Loop Honest |
|
|
| 7.3 | Meta-Agents: Agents That Build Agents |
|
|
| 7.4 | Beyond MCP: Choosing Tool Channels |
|
|
| 7.5 | Mac Mini Agent: Physical Sandbox |
|
|
| 7.6 | Always-On Agents |
|
|
| Labs | Autoresearch Loop + Meta-Agent |
|
|
|
|
## Module 8: Capstone (8-12 hours)
|
|
|
|
Build a production-grade multi-agent system. Choose from:
|
|
- **Brand Monitor** — Multi-LLM brand mention tracking
|
|
- **Code Review Pipeline** — Plan, Build, Review, Verify
|
|
- **Strategic Decision Board** — 8-agent CEO board
|
|
|
|
## Reference Documents
|
|
|
|
- [FIELD-MANUAL.md](/modules/field-manual) — Quick reference cheat sheet
|
|
- [FEYNMAN.md](/modules/feynman) — Core concepts in plain language
|
|
- [DEBATE.md](/modules/debate) — The great agent debate (Armin vs George)
|
|
- [COMPETITIVE-ANALYSIS.md](/modules/competitive-analysis) — vs Anthropic, TAC, ClaudeFAST
|
|
- [REFERENCE-STACK.md](/modules/reference-stack) — Production architecture
|
|
- [TOOL-REFERENCE.md](/modules/tool-reference) — 5-tool CLI reference
|