Recent AI Reading [15 July 2026]

Papers

Agentic AI

• Agentic AI and the next intelligence explosion
• SAGE: Multi-Agent Self-Evolution for LLM Reasoning
  • Cross Topics: Training Paradigms
• Sentinel Agents for Secure and Trustworthy Agentic AI in Multi-Agent Systems
• Tool Attention Is All You Need: Dynamic Tool Gating and Lazy Schema Loading for Eliminating the MCP/Tools Tax in Scalable Agentic Workflows
  • Cross Topics: Model Context Protocol
• Agentic World Modeling: Foundations, Capabilities, Laws, and Beyond
  • Cross Topics: Embodied AI
• PASK: Toward Intent-Aware Proactive Agents with Long-Term Memory
  • Cross Topics: AI Memory
• From Skill Text to Skill Structure: The Scheduling-Structural-Logical Representation for Agent Skills
• The Last Human-Written Paper: Agent-Native Research Artifacts
• Coordination as an Architectural Layer for LLM-Based Multi-Agent Systems
• AAFLOW: Scalable Patterns for Agentic AI Workflows
• The Bystander Effect in Multi-Agent Reasoning: Quantifying Cognitive Loafing in Collaborative Interactions
• Self-Distilled Agentic Reinforcement Learning
  • Cross Topics: Training Paradigms
• Code as Agent Harness
• SkillOpt: Executive Strategy for Self-Evolving Agent Skills
• Harness Updating Is Not Harness Benefit: Disentangling Evolution Capabilities in Self-Evolving LLM Agents
• The Consistency Illusion: How Multi-Agent Debate Hides Reasoning Misalignment
• Self-Harness: Harnesses That Improve Themselves
• HarnessX: A Composable, Adaptive, and Evolvable Agent Harness Foundry
• OpenClaw-Skill: Collective Skill Tree Search for Agentic Large Language Models
• Skill-MAS: Evolving Meta-Skill for Automatic Multi-Agent Systems
• The Hitchhiker’s Guide to Agentic AI: From Foundations to Systems
• The Harness Effect: How Orchestration Design Sets the Token Economics of Enterprise Agentic AI
• MetaSkill-Evolve: Recursive Self-Improvement of LLM Agents via Two-Timescale Meta-Skill Evolution
  • Cross Topics: Training Paradigms

AI Alignment (with Human Preferences, and other methods)

• Positive Alignment: Artificial Intelligence for Human Flourishing

Large Language Models

• Attention Residuals
• Debating with More Persuasive LLMs Leads to More Truthful Answers
  • Cross Topics: Agentic AI
• MIRAGE: The Illusion of Visual Understanding
  • Cross Topics: Model Evaluation
• Recursive Language Models
• Think in Sentences: Explicit Sentence Boundaries Enhance Language Model’s Capabilities

Training Paradigms

• PivotRL: High Accuracy Agentic Post-Training at Low Compute Cost
  • Cross Topics: Agentic AI
• Embarrassingly Simple Self-Distillation Improves Code Generation
• A Primer in Post-Training Reasoning Data: What We Know About How It Works

Model Evaluation

• AutoLab: Can Frontier Models Solve Long-Horizon Auto Research and Engineering Tasks?
  • Cross Topics: Agentic AI
• Human-on-the-Bridge: Scalable Evaluation for AI Agents
  • Cross Topics: Agentic AI
• Long-Horizon-Terminal-Bench: Testing the Limits of Agents on Long-Horizon Terminal Tasks with Dense Reward-Based Grading
  • Cross Topics: Agentic AI

Retrieval-Augmented Generation

• PathRAG: Pruning Graph-based Retrieval Augmented Generation with Relational Paths

Synthetic Data

Embodied AI

• ASPIRE: Agentic Skills Discovery for Robotics
  • Cross Topics: Agentic AI

AI Memory

• REMem: Reasoning with Episodic Memory in Language Agent
  • Cross Topics: Agentic AI
• MemCollab: Cross-Agent Memory Collaboration via Contrastive Trajectory Distillation
  • Cross Topics: Agentic AI
• GAAMA: Graph Augmented Associative Memory for Agents
  • Cross Topics: Agentic AI, Retrieval-Augmented Generation
• Artifacts as Memory Beyond the Agent Boundary
  • Cross Topics: Agentic AI
• Memory Transfer Learning: How Memories are Transferred Across Domains in Coding Agents
  • Cross Topics: Agentic AI
• StructMem: Structured Memory for Long-Horizon Behavior in LLMs
• MemPrivacy: Privacy-Preserving Personalized Memory Management for Edge-Cloud Agents
  • Cross Topics: Agentic AI
• MeMo: Memory as a Model
• Memory is Reconstructed, Not Retrieved: Graph Memory for LLM Agents
  • Cross Topics: Retrieval-Augmented Generation, Agentic AI
• AtomMem: Building Simple and Effective Memory System for LLM Agents via Atomic Facts
  • Cross Topics: Agentic AI
• Are We Ready For An Agent-Native Memory System?
  • Cross Topics: Agentic AI
• Neural Procedural Memory: Empowering LLM Agents with Implicit Activation Steering
  • Cross Topics: Agentic AI
• Always-On Agents: A Survey of Persistent Memory, State, and Governance in LLM Agents
  • Cross Topics: Agentic AI
• AutoMem: Automated Learning of Memory as a Cognitive Skill
  • Cross Topics: Agentic AI
• Remember When It Matters: Proactive Memory Agent for Long-Horizon Agents
  • Cross Topics: Agentic AI

Articles and Blog Posts

• The One-Person Unicorn 🦄 - by Linas Beliūnas
• Block - From Hierarchy to Intelligence
• Demystifying evals for AI agents - Anthropic
  • Cross Topics: Model Evaluation, Agentic AI
• Why Frontier Agents Fail at Reading Documents — and How We’re Fixing It - Databricks
  • Cross Topics: Agentic AI, Retrieval-Augmented Generation
• The AI-native interview - Sierra
• Claude Code Dreams: Anthropic’s New Memory Feature
  • Cross Topics: AI Memory
• The founder’s playbook: Building an AI-native startup - Claude
• AI 2040: Plan A

Miscellaneous

• GitAgent — The Open Standard for Git-Native AI Agents
  • Cross Topics: Agentic AI
• The Future of Everything is Lies, I Guess
• andrej-karpathy-skills: A single CLAUDE.md file to improve Claude Code behavior
• Memento: Fine-tuning LLM Agents without Fine-tuning LLMs (code)
  • Cross Topics: AI Memory, Agentic AI