A curated reading guide for the most important agent-systems research — surveys, architecture papers, memory, security, governance, and emerging directions.
These give the strongest architectural grounding before diving into specialized work.
Survey paper
Focus: One of the best modern surveys of agent systems.
Comprehensive survey
Focus: Very comprehensive landscape survey of LLM agents.
Reasoning-focused research
Focus: Probably the best paper on reasoning-oriented agents.
Core architecture papers shaping how production-grade agents are built.
Production-oriented research
Focus: Modern, production-oriented agent infrastructure paper.
Google Research
Focus: Empirical evaluation of scaling and coordination tradeoffs.
Skill-systems research
Focus: Modular skill architectures — highly relevant to MCP/tool ecosystems.
One of the hottest research areas right now.
Memory-systems research
Focus: One of the best memory papers available.
An area exploding in 2026.
First major agent-security survey
Focus: The first major security survey for AI agents.
Cybersecurity-focused survey
Focus: Agents in offensive and defensive security operations.
Governance & regulation
Focus: Important governance paper on agent regulation.
Foundational and forward-looking work on agent collectives.
Foundational survey
Focus: Classic foundational paper for the field.
Forward-looking essay/research
Focus: Forward-looking analysis of distributed agent ecosystems.
Where agents meet decentralized economic infrastructure.
Architecture paper
Focus: A novel architecture combining agents and blockchain primitives.
One of the newest directions for agent applications.
Autonomous research ecosystem
Focus: A research ecosystem for autonomous AI scientific contributors.
Agents tuned to a specific scientific domain.
Domain-agent research
Focus: A representative example of vertical, domain-specific agent design.
Curated GitHub repositories that update weekly from arXiv — very useful for staying current.
GitHub repository
Focus: Curated list of agent-systems papers.
GitHub repository
Focus: Another curated list, refreshed weekly from arXiv.
Where cutting-edge AI agent research is concentrated right now.
| Research Area | Importance |
|---|---|
| Agent memory | Very High |
| Multi-agent orchestration | Very High |
| Tool use / MCP | Very High |
| Agent governance | Very High |
| Agent security | Very High |
| Reflection / self-correction | High |
| Long-horizon planning | Very High |
| Agent evaluation | High |
| Persistent memory | Very High |
| Deterministic agents | High |
| Agent observability | High |
| Skill loading systems | High |
| Runtime orchestration | Very High |
Selected with a profile of Rust systems, distributed architecture, RAG/memory work, blockchain, and orchestration interest in mind — closest to where enterprise agentic architecture is heading.