Agent rot: the silent degradation of deployed AI
Why deployed AI agents quietly decay, how the failure stays hidden, and the governance model that keeps them honest.
Language models & AI agents
We build AI agents that survive production.
Featured
Exploring the hidden thoughts of language models
Anthropic found a small, shared workspace inside a language model where it holds the concepts it is about to use. What the result shows, and what it does not.
The agent index
A field guide to 70+ autonomous AI agents and agent frameworks: what they cost, where they run, who's winning, and what each is good for.
Latest
Per page
Agentic AI and agentic workflows
A practitioner's walk through agentic AI — what makes a system an agent, how to analyze whether you need one, and how to design workflows, memory, tooling, evaluation, observability, safety, and security for agents in the enterprise.
AI-Augmented Lean Enterprise
How Lean and autonomous AI agents reinforce each other — adaptive standardization, digital Kaizen, and just-in-time intelligence — plus the benefits, risks, and roadmap for an AI-augmented Lean enterprise.
AI Fundamentals for Product Leaders
A whiteboard tour of the AI literacy product leaders need — from the shift to self-learning machines and the three learning paradigms, through how LLMs and hallucinations actually work, to RAG and fine-tuning trade-offs, prompt engineering, multimodal AI, and the move from chatbots to agents.
In a world full of agents: are we heading toward a future designed by AI?
A tour through the agentic AI landscape — what agents and multi-agent systems are, how they reason from prompts, how to architect and govern them, and whether the future ahead is one designed by AI.
Virtual employees at scale: productivity gains vs systemic risk
Deploying AI agents as virtual employees across the enterprise, and the governance, structure, and workforce changes that keep the productivity gains from turning into systemic risk.
Claude Code folder structure
A reference to how Claude Code organizes its files and folders based on the official docs.
GitHub Copilot folder structure
A reference to GitHub Copilot's file and folder structure based on the official docs.
OpenAI Codex folder structure
A reference to OpenAI Codex's file and folder structure based on the official docs.
Agent rot: the silent degradation of deployed AI
Why deployed AI agents quietly decay, how the failure stays hidden, and the governance model that keeps them honest.
Episode 8: AI and the great imbalance — the future of human work
The eighth episode of our Agentic AI and agentic workflows series, on the uncertain future of human labor as generative AI replaces tasks faster than it creates sustainable new roles — the great imbalance, the broken professional pipeline, and the choice between replacing people and augmenting them.
Episode 9: Deep research — the AI research revolution
The ninth episode of our Agentic AI and agentic workflows series, on the emerging AI research revolution — reframing research as an end-to-end, agent-driven process where systems plan investigations, gather evidence, and synthesize findings into structured, verifiable outputs.
Episode 7: Evaluation, observability, safety, platforms, and future directions
The seventh episode of our Agentic AI and agentic workflows series, on what it takes to move agents out of the lab and into production — how they are measured, monitored, and governed once live, and the safeguards and operational discipline that keep them aligned, auditable, and resilient over time.
Episode 2: Foundations of agentic AI
The second episode of our Agentic AI and agentic workflows series, on what agentic AI really means — reframing agents as coordinated systems that reason, remember, act within clear boundaries, and pursue complex goals on their own.
Episode 5: Implementing specialized agents in code and frameworks
The fifth episode of our Agentic AI and agentic workflows series, on what it takes to move agents from paper designs to reliable production systems — how tooling choices shape flexibility and scalability, and how core behaviors like reasoning, memory, decision-making, and safety become real implementations.
Episode 4: Need analysis and agentic workflow design
The fourth episode of our Agentic AI and agentic workflows series, on how to tell whether an agent is the right tool, how that answer shapes architecture from simple automation to advanced agents, and why workflow structure, decision boundaries, and memory strategy drive reliability at scale.
Episode 6: No-code and low-code agentic workflows
The sixth episode of our Agentic AI and agentic workflows series, on how visual no-code and low-code platforms are reshaping the way agentic systems are designed and deployed — from event-driven starts and knowledge retrieval to decision loops, tool use, and the moments where humans step in to guide or approve outcomes.
Episode 3: Reasoning methodologies for agents
The third episode of our Agentic AI and agentic workflows series, on how agents think, plan, and decide — the reasoning styles behind everything from quick answers to complex multi-step workflows, and the trade-offs of speed, depth, adaptability, and reliability.
Episode 1: The agentic shift
The first episode of our Agentic AI and agentic workflows series, on the move from passive AI assistance to goal-directed agents — why it is gaining urgency now, and where it leaves traditional automation.
Google's Open Knowledge Format
A vendor-neutral way for agents to share knowledge: plain markdown files with YAML frontmatter, no SDK or backend required. OKF is to knowledge what MCP is to APIs.
Knowing your limitations is the foundation of expertise
Agentic AI is rediscovering Minsky's 1986 community of agents and calling it revolutionary. The scarce thing isn't novelty, it's the humility to know what we don't know.
The neocognitron and AI's rediscovery habit
Old ideas in AI keep getting rediscovered, repackaged, and presented as new. Sparse attention, MCP, and A2A all have decades-old roots. The shortage isn't creativity, it's historical awareness.
SkillOpt: self-evolving agent skills
Microsoft Research's SkillOpt improves an agent's instructions automatically through trial and error, like gradient descent for text, with no extra cost at deployment.
The ELIZA effect
The ELIZA effect warns us not to read a mind into a simple program. With today's systems the more rigorous-seeming error may be the opposite: explaining the behavior away as mere mechanism.
The real danger isn't that AI is too advanced
The real danger of AI is not excess capability but our tendency to overestimate it, and the misplaced trust, premature deployment, and hype-driven policy that follow.
We didn't solve Moravec's paradox, we rebranded it
Broad capabilities, brittle consistency: today's agents reason like PhDs while lacking the reliability of a toddler stacking blocks.
Being smart is not the same as wanting something
Steven Pinker on why superintelligence need not bring the will to dominate: intelligence is the ability to reach a goal, not a goal in itself.
The agentic design patterns directory
A working reference of the agentic design patterns, from single-agent building blocks to the controls that keep production systems safe.
Designing a hierarchical skill directory
A design guide for building a hierarchical skill directory — the Agent Skills open standard, and how different agent tools implement it.
Exploring the hidden thoughts of language models
Anthropic found a small, shared workspace inside a language model where it holds the concepts it is about to use. What the result shows, and what it does not.
Moving toward AGI: bridging the gap between text and understanding
Fluent token prediction is not understanding.
The sophistication gap: what misbehaving AI reveals about the limits of artificial intelligence
AI companies claim we are near artificial general intelligence. This is not true. A recent open-source software incident proves we are far from that goal.
The agent index
A field guide to 70+ autonomous AI agents and agent frameworks: what they cost, where they run, who's winning, and what each is good for.