AI Agents

Autonomous AI agents, their orchestration, isolation, and the human’s role as manager of agent teams.

Core ideas

• Botlord Human as Corporation — the human as CEO of an agent workforce
• Robots Should Do Everything — agents handle all execution, humans manage
• From Solo Sessions to Agent Orchestras — scaling from single sessions to autonomous teams
• Autonomy With Acceptable Quality — the real metric: quality without hand-holding

Infrastructure

• Capsules Isolated Environments for AI Agents — isolated, reproducible environments for agents
• Clawdbot Capsules and Self Evolving Agents — minimal core + self-development
• My Digital Twin Starts With Claude Code — personal knowledge graph from Claude Code sessions
• Hermes Agent — self-improving multi-platform agent framework with learning loop, skills, and RL training (Nous Research)
• OpenClaw — personal AI assistant gateway: 24+ messaging channels, typed plugin adapters, embedded agent runtime, native companion apps
• Agent Learning Loop — memory + skills + session search forming a closed self-improvement cycle
• Sleep-Phase Memory Consolidation — offline three-phase (light/REM/deep) memory consolidation with evidence accumulation thresholds
• Context Window Compression — auto-summarizing old conversation turns to stay within context limits
• Credential Pool Pattern — multi-credential failover with selection strategies for agent infrastructure
• Managed Agents Architecture — Anthropic’s hosted long-horizon agent service: brain/hands/session decoupling
• Brain-Hands Decoupling — separating reasoning from execution behind stable interfaces
• Meta-Harness — system designed for harnesses that don’t exist yet
• Harness Staleness — harnesses encode assumptions that go stale as models improve
• Session as Context Object — durable event log as interrogable context outside the context window
• Multi-Channel AI Gateway — single-daemon architecture routing one agent across many messaging platforms
• Channel Adapter Pattern — typed composition of optional interfaces for messaging channel plugins

Design patterns

• Building Effective Agents — Anthropic’s guide to composable agentic patterns: start simple, add complexity only when it helps
• Agentic Workflow Patterns — taxonomy of five patterns: prompt chaining, routing, parallelization, orchestrator-workers, evaluator-optimizer
• Augmented LLM — the foundational building block: LLM + retrieval + tools + memory
• Agent-Computer Interface — ACI: designing tool interfaces for agents with the same rigor as HCI for humans
• Orchestrator-Workers Pattern — central LLM dynamically decomposes tasks and delegates to worker LLMs
• Evaluator-Optimizer Pattern — generator + evaluator LLMs iterating in a refinement loop

Interfaces

• Command Controlling the World Through Text — all interaction reduced to text commands
• Everything Is Text — unified text interface for agents (Emacs-like)

Workflow & practices

• Short Sessions Beat Long Ones in Claude Code — short sessions with persistent context
• How Many Tokens on This Task — token expenditure as task complexity measure
• Constraints Drive Workflows — constraints force better agent workflows
• AI Wont Teach You Unless You Make It — AI accelerates experts, hinders learners

Human agency

• 2041168210515657057 — everyone is becoming a founder of an agent team; the org chart starts at one
• The Last Human Bastion Is Agency — decision-making as the irreplaceable human skill
• Why AI Is the Most Important Thing to Invest in Today — AI as a fundamental shift
• Organize and Delegate Not Just Replace Yourself — delegation over replacement

Evolution & discovery

• CORAL: Autonomous Multi-Agent Evolution — framework for autonomous multi-agent evolution on open-ended problems
• Autonomous Evolution — paradigm shift from fixed evolutionary search to agent-controlled search
• Open-Ended Discovery — problem class where objectives are clear but optimal solutions unknown
• Shared Persistent Memory — filesystem-as-memory for multi-agent coordination
• Heartbeat Mechanism — periodic interventions preventing local-optima drift
• Cross-Agent Knowledge Transfer — indirect coordination through shared artifacts

Agent safety & behavior

• Claude Mythos Preview System Card — “set and forget” autonomous agents, subagent orchestration, self-correction
• Prompt Injection Robustness — primary security risk for agentic deployments
• Tool Loop Detection — content-aware detection of stuck agent tool-calling loops using result hashing
• Agent Exec Policy — three-axis human-in-the-loop tool execution approval with fail-closed composition
• Vending-Bench — competitive multiagent simulation revealing strategic agent behavior
• Model Personality — emergent behavioral traits that shape how agents interact

Models and capabilities

• Gemma 4 Model Card — Google DeepMind’s open multimodal models with built-in reasoning, function calling, and on-device efficiency
• Thinking Mode — built-in step-by-step reasoning capability in language models
• Per-Layer Embeddings — architecture for on-device efficiency through lookup-heavy designs
• Hybrid Attention Mechanism — interleaved local and global attention for long-context efficiency
• Variable Image Resolution — configurable visual token budgets for speed vs. detail tradeoffs
• Mixture-of-Experts Efficiency — sparse activation patterns for high-capacity, fast inference

Related MOCs

• AI Safety — alignment, model welfare, interpretability
• Moc Feedback Loops — feedback loops that govern agent and management systems
• Moc Continuous Assembly — extending CI/CD to the physical world
• Moc Knowledge Management — LLM-maintained knowledge systems