← Back to curriculum
Phase 15 · 22 lessons
Autonomous Systems
Long-horizon agents, self-improvement, and the 2026 safety stack.
01The Shift from Chatbots to Long-Horizon AgentsLearn02STaR, V-STaR, Quiet-STaR — Self-Taught ReasoningLearn03AlphaEvolve — Evolutionary Coding AgentsLearn04Darwin Godel Machine — Open-Ended Self-Modifying AgentsLearn05AI Scientist v2 — Workshop-Level Autonomous ResearchLearn06Automated Alignment Research (Anthropic AAR)Learn07Recursive Self-Improvement — Capability vs AlignmentLearn08Bounded Self-Improvement DesignsLearn09The Autonomous Coding Agent Landscape (2026)Learn10Claude Code as an Autonomous Agent: Permission Modes and Auto ModeLearn11Browser Agents and Long-Horizon Web TasksLearn12Long-Running Background Agents: Durable ExecutionLearn13Action Budgets, Iteration Caps, and Cost GovernorsLearn14Kill Switches, Circuit Breakers, and Canary TokensLearn15Human-in-the-Loop: Propose-Then-CommitLearn16Checkpoints and RollbackLearn17Constitutional AI and Rule OverridesLearn18Llama Guard and Input/Output ClassificationLearn19Anthropic Responsible Scaling Policy v3.0Learn20OpenAI Preparedness Framework and DeepMind Frontier Safety FrameworkLearn21METR Time Horizons and External Capability EvaluationLearn22CAIS, CAISI, and Societal-Scale RiskLearn