Agent
Auto Improving
Learn to develop, evaluate, and continuously improve AI Agents using Coding Agents, W&B Weave, and W&B Skills. We named this learning content Agentforge — forging AI Agents to be stronger.
What is Auto Research?
Give an AI agent a real setup and let it experiment autonomously — modify, evaluate, keep or discard, repeat. You wake up to a log of experiments and a better system. Any metric you can efficiently evaluate can be autoresearched.
"All LLM frontier labs will do this. It's the final boss battle."
— Andrej Karpathy
Courses
Choose a course to begin your learning journey.
AI Agent Quality Evaluation & Improvement with Agents
Use W&B Weave as the foundation for tracing, evaluation, and monitoring of AI Agents, and learn to improve them with Coding Agents and W&B Skills.
AI Agent Quality Evaluation & Improvement: Hands-on
Hands-on exercises and practical assignments for the AI Agent Quality course.
Who is this for?
Agent Optimization with Coding Agents
You use coding agents to build and optimize AI agents, and want a structured workflow for it.
From Human Eval to Automated Systems
You want to learn how to go from human evaluation to building a scalable evaluation framework.
Continuous Agent Improvement Teams
Your team wants a repeatable process for continuously improving agents in production.
W&B Weave Power Users
You want hands-on experience using Weave for evaluation, monitoring, and labeling of AI agents.
Auto Research Best Practices
You want to learn implementation and evaluation best practices for Auto Research agents.
Powered by W&B
This course uses W&B Weave and W&B Skills as core tools throughout every chapter.
W&B Weave
The observability and evaluation platform for AI applications.
- Tracing — Record every agent action with full context
- Evaluations — Structured quality assessments with datasets and scorers
- Labeling — Human review workflows that feed into automated evaluation
- Monitoring — Production dashboards, alerts, and regression detection
- Feedback — Capture user signals and route them to improvement pipelines
W&B Skills
Reusable agent workflows for optimization and improvement.
- Agent Optimization — Improve prompts, retrieval, and knowledge bases from evaluation data
- Evaluation Schemes — Create scorers and evaluation pipelines
- Weave Integration — Add tracing and monitoring to existing agents