We Document

What we learn from the work.

Short notes from building and testing AI systems for legal work — what worked, what failed, and what we are still figuring out.

We Build → We Test → We Document → Tutorials

Legal AI Documenting

Four areas of legal AI work: legal reasoning, legal workflows, AI systems, and attorney + AI.

Observing · Recording · Learning · Sharing

01 · Categories

What do we document?

We document what we learn across four areas: legal reasoning, legal workflows, AI systems, and attorney + AI collaboration.

01
Legal Reasoning
How AI systems reason through legal problems — analysis, judgment, and where reasoning breaks down.
Explore notes
02
Legal Workflows
How legal work is structured into workflows that AI systems can perform and attorneys can review.
Explore notes
03
AI Systems
The systems themselves — architectures, models, multi-agent orchestration, and failure modes.
Explore notes
04
Attorney + AI
Where attorney judgment enters the process, what it changes, and what AI systems need from it.
Explore notes

02 · Notes

Latest Notes

Each note captures something specific we learned — what worked, what failed, or what we are still figuring out. These are practical observations from building and testing AI systems for legal work.

Latest · AI Systems

Harvey Tenet: Engineering AI for Long-Horizon Legal Work.

Understanding the model, the agent, the training environment, and the benchmarks behind Harvey Tenet.

Sep 2026Read Note

Aug 2026

Attorney + AI · Legal Workflows

How Should Legal AI Collect Attorney Feedback? It is a UX Problem.

If attorney feedback is part of a legal AI workflow, how should an AI system collect, interpret, and act on that feedback?

Read Note

Aug 2026

Legal Reasoning · AI Systems

Why Legal Reasoning Cannot Be Reduced to One AI Prompt

Prompt engineering matters, but difficult legal work also requires engineering the context, loops, and structures around language models.

Read Note

Aug 2026

Legal Workflows · AI Systems

Why Intermediate Work Products Matter in Legal AI

Intermediate work products make complex legal AI workflows visible enough to review and revise, and make the chain of legal reasoning easier to follow.

Read Note

Aug 2026

AI Systems · Legal Workflows

Why an AI Legal System Needs More Than One Agent

A single model can answer a question. A legal system needs to manage a workflow with specialized AI agents.

Read Note