The library
11 pieces. Narrow them by what you are trying to solve.
Sorted newest first by default. Filter by subject or by the shape of the piece, or order by length when you have twenty minutes rather than an hour.
A realistic agent evaluation needs more than a benchmark and a score. It needs explicit scope, least-privileged tools, observable actions, stop conditions, and a recovery plan.
02Beyond the Prompt: The Agent System Stack ExplainedAgent systems912mContext, tools, memory, harnesses, loops, and evals are different layers of an AI agent system. Here is how they fit together—and why the model alone does not decide whether an agent works.
03Context Engineering: Why More Context Can Make an AI Agent WorseContext engineering411mThe model can only make its next decision from what is in front of it. Context engineering is the practice of keeping that working set current, relevant, and deliberately small.
04AI Agent Evals: How Do You Know an Agent Actually Works?Agent evaluation212mA convincing final message is not proof of completion. Agent evaluations check the task, the actual environment, repeated trials, and the evidence left behind.
05Harness Engineering: Why a Better Model Is Not the Whole AnswerAgent systems511mA capable model still needs a well-designed runtime around it. Harness engineering is the work of giving an agent useful context, tools, boundaries, state, and evidence of completion.
06Loop Engineering: What It Means—and Why It Is More Than Letting an Agent Run ForeverAgent systems510mLoop engineering is an emerging label for designing repeated agent work: its trigger, state, feedback, boundaries, verification, and explicit ways to stop or hand off.
07Why an Agent Eval Can Pass on Monday and Fail on FridayAgent evaluation211mA passing eval is evidence about one version of an agent system. When the model, context, tools, policies, or environment change, the evidence needs to be renewed.
08Autonomous Is Not Helpful: How Should We Evaluate Proactive AI Agents?Agent systems210mAn agent that can act without being asked is not automatically useful. The hard question is whether it noticed the right thing, had evidence, chose the right moment, and knew when to stay quiet.
09AI Agents, Explained: The Loop Behind the LabelAI agents39mA beginner-friendly map of what makes a system agentic, how the model–tool loop works, and when a simpler workflow is the better design.
10How Tool-Using AI Agents Work: A Weather Assistant, Step by StepTool use510mFollow one tool call from a user request to a validated result, then see what changes when an agent can take actions instead of only reading data.
11Agent Memory, Explained: Context, Retrieval, and StateContext & memory310mSeparate four mechanisms that are often called memory, then use the distinctions to design and test an agent that remembers the right things.