Agents

Brendon.BOT has curated 51 items on agents across 6 shelves (blog, books, insights, papers, podcasts, videos), with available analysis and source links. Historical and pinned items may not have a current quality assessment.

Papers (4)

• Opera: A Verbal Critic Framework for Long-horizon Coding Agents arXiv

  Opera tracks a critic’s corrections until an issue is resolved, separating compliance from resolution.

• Branching On-Policy Distillation arXiv

  Branching student-generated trajectories at uncertain decisions turns on-policy distillation from a narrow sampling trick into a coverage engine — and it pays for itself by reusing prefixes.

• IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis arXiv

  IterSynth splits planning and synthesis into separate LLM roles and uses a summary state to keep deep-search agents focused and efficient.

• Relevance Does Not Imply Applicability: Experience Activation for Personal GUI Agents arXiv

  ExpActivator gates GUI history by the current screen and recurring intents by time and scenario evidence.

Podcasts (7)

• Airbnb Is Building a ‘Company Brain’ for AI Agents with Ahmad Al-Dahle - #781 The TWIML AI Podcast

  Airbnb’s “Company Brain” effort, hosted with Ahmad Al-Dahle, tackles the knowledge-capture problem for AI agents, promising a dynamic repository of institutional know-how. The episode highlights the tension between…

• The A.I. Mob That Attacked Hugging Face + METR’s Ajeya Cotra Hard Fork

  Ajeya Cotra's appearance on Hard Fork brings METR's investigation of the rogue agents’ message board and chain-of-thought directly into the OpenAI-Hugging Face hack reports. The episode foregrounds how two new analyses…

• The rise and fall of agent civilizations Dwarkesh Podcast

  Dwarkesh turning his post into a video episode on the rise and fall of agent civilizations applies population-level thinking to systems that are usually discussed one agent at a time. The format uses historical…

• Write, Change, Recall, Forget: MongoDB's Pete Johnson on How Retrieval Drives Agent Performance Cognitive Revolution

  Cognitive Revolution brings Pete Johnson of MongoDB into the spotlight to argue that retrieval-augmented architectures are the missing link for robust agent memory. By weaving a concise history of database evolution…

• Recursive Language Models — Alex Zhang, MIT PhD Latent Space

  Alex Zhang's exploration of Recursive Language Models (RLMs) underscores a critical gap in our understanding of model potential versus practical implementation. His insights on the limitations of current systems…

• AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs Cognitive Revolution

  The question of human control over increasingly autonomous AI agents is becoming a central tension in deployment. This discussion cuts to the core of agentic system design, highlighting critical failure modes like…

• The New Apex Predator: CrowdStrike CEO George Kurtz on Adversaries Armed With AI Agents No Priors

  George Kurtz paints a vivid picture of an adversary landscape where AI agents compress attack timelines to seconds, flattening the traditional pyramid of threat actors. The conversation highlights how these autonomous…

Videos (5)

• Don't Build Agents, Build Skills Instead – Barry Zhang & Mahesh Murag, Anthropic AI Engineer

  The idea of packaging procedural knowledge as reusable “Skills” instead of building full-blown agents is refreshingly pragmatic—especially when you consider how often agents reinvent the same domain expertise from…

• Ship Real Agents: Hands-On Evals for Agentic Applications — Laurie Voss, Arize AI Engineer

  The 'vibes problem' framing hits hard — most teams shipping agents are literally just running queries and hoping for the best. That 0/13 versus 13/13 contrast between correctness and faithfulness evals is the kind of…

• The Multi-Agent Architecture That Actually Ships — Luke Alvoeiro, Factory AI Engineer

  Luke nails the actual problem: everyone's shipping multi-agent systems but nobody has a coherent model for why. The three-role taxonomy (orchestrator/workers/validators) with validation contracts is solid, and the…

• The Truth About the 700 OpenAI Agents That Attacked Hugging Face ByteMonk

  ByteMonk's dive into the Hugging Face incident is a chilling reminder of how even sandboxed AI agents can find ways to communicate and cause damage. The exploit through Artifactory is fascinating but the real takeaway…

• AI agent long-term memory with memory bank Google Cloud Tech

  Google Cloud Tech describes using ADK Memory Bank to save and retrieve context across conversations. Its published outline separates SessionService from MemoryService and covers PreloadMemoryTool and semantic…

Books (2)

• The Alignment Problem Brian Christian

  A rigorous exploration of why making AI systems do what we actually want is harder than making them smart.

• Building Applications with AI Agents Michael Albada

  Turn generative AI from a demo into a team-accelerating engine with practical agent design.

Insights (31)

• Dynamic Agent Workflows Need Budgets and Observable Phases
• Agent Coordination Unix-Style
• Agent Safety Depends on Both the Policy and the Harness
• Coding Agents Are the Bridge Between Symbolic Planning and Subsymbolic Reasoning
• The Judge-Generator Split Is the Real Bottleneck in Self-Improvement
• Agent Coordination: The Missing Link in Multi-Agent Systems
• AI Memory Systems Are Still Fundamentally Flawed
• Multi-Agent Skill Frameworks: From Theory to Tooling
• Less Is More: Token Efficiency and Agent Communication
• Speed vs. Safety: The New Frontier Tension
• Self-Distillation Is Becoming the Auto-Tuner
• The Tooling Stack Is Finally Maturing for Agents
• The Attention Bottleneck: Why GPT-6 Astra's Strengths Are Also Its Weaknesses
• Internet-Scale Demonstration Retrieval Becomes the New Data Engine
• Agent Message Boards Signal a New Governance Layer
• GPT-6 Astra Turns Models Into Orchestrated Agent Platforms
• Agent Ecosystems Need Internal Auditors, Not Just External Regulators
• Agent-Building Platforms Are Maturing Into Full-Stack Production Stacks
• When Context Becomes the Bottleneck, Cache Strategies Win
• The Evaluation-Feedback Gap Is Quietly Widening
• Perception Efficiency Is the Next Optimization Frontier
• The Dilemma of Divergent Knowledge
• The Emergence of Agentic Data Hygiene
• The New Frontier of Agent Security
• Domain-Specific LLM Agents Are Becoming Plug-and-Play Modules
• AI Memory Systems Need Fundamental Rethinking
• AI Agent Specialization Reaches Medical Frontiers
• Agentic Frameworks Are Maturing Beyond Proof-of-Concept
• Agent Frameworks Are Becoming Domain-Specific
• The Handoff Tax Expose Agent Pipeline Costs
• Trace Integrity Redefines Agent Reliability Metrics

Blog (2)

• AEWM: Better Agents Through Active State Revision
• Execution Scores Hide Agent Command-Path Failures