The Overnight Cookbook
Patterns for longer-running agent work: durable state, review, recovery, and human decisions.
Read the Cookbook →HOPE Lab
Hands-On Projects
and Experimentation
Learn to work with agents across multiple stages, with clear goals, boundaries, and evidence of what worked.
The HOPE Lab format
Give your agent a project.
Not just another prompt.
We work through practical examples together and explore how the same patterns apply to your own public, non-confidential project.
No programming experience required. This is hands-on project work, not a beginner chatbot tutorial or an installation clinic.
- Define the outcome.
Turn a broad idea into concrete deliverables. - Set up durable context.
Provide instructions, files, and authority boundaries that carry across stages. - Guide longer-running work.
Use checkpoints and human decisions without constantly starting over. - Check what happened.
Evaluate quality, completion, traceability, and correction burden.
Bring your agent.
Bring your curiosity.
Lawyers and other professionals are welcome. Bring a laptop and a working agent environment you already use: Codex, Claude Code, Claude desktop/Cowork, Cursor, or a comparable tool.
Bring a non-confidential project idea, or use one of our examples. Please do not bring confidential client or organizational information into shared exercises.
HOPE Lab sessions are free. Any charges for your own agent tools remain yours.
Featuring Descrybe
Integrating Legal Research with AI Agents
Source-grounded research · Inspectable authorities · Human judgment
Learn to connect source-grounded legal research to your AI agent projects, inspect the supporting authorities, and apply your own judgment.
For the September 16 lab, Descrybe offered limited-time free access to registered participants who opted in. Access arrangements for any future session will be announced separately.
Start exploring now
Open materials. Practical experiments.
You don’t need to read these in advance or buy anything to participate.
HOPE exercises
Try the starter prompts, project structure, and practical exercises at your own pace.
Explore the exercises →Want to go further with your own project? Explore ways to work with Dazza →
About your guide
Dazza Greenwood
Dazza Greenwood is a lawyer-technologist who develops methods and tools for agent workflows, authority boundaries, and practical evaluation. His work includes Interlateral, the Overnight Cookbook, and hands-on HOPE sessions for putting those ideas into practice.
Writing & background →LinkedIn profile →Explore the open-source work →
HOPE is part of the Leave No Lawyer Behind initiative.