A clear route into AI.
Start with an explanation, follow a practical workflow, or build something small. You can read any chapter without committing to the whole curriculum.
Understand AI
How it works behind the scenes. For judges, lawyers, and curious newcomers: what a model does, what it can see, why a persuasive answer can be wrong, and what courts and regulators already expect. No code.
Use AI in Your Work
Move from a clear assignment to source-linked work products, focused revision, and repeatable quality checks.
Build AI Systems
Choose and test a local model, then build document workflows with Python, APIs, retrieval, and bounded agents. Finish with security, deployment, and maintenance. Local processing is not a privacy guarantee.
The 9 chapters
Expand a chapter to see its lessons01AI essentials
Understand what a model does, what it can see, and why a persuasive answer still needs support.
7 lessons
AI essentials
Understand what a model does, what it can see, and why a persuasive answer still needs support.
What actually happens when you ask AI a question?
A language model predicts text one piece at a time; the application around it decides what it sees. Knowing which part did what is how you check an answer.
Tokens, attention, and why wording matters
Models read tokens, not words. Legal citations can use more tokens than ordinary prose, and clear labels help the model link the right pieces.
What uploading a document does—and does not—establish
The context window is the model's whole working desk. Learn what fills it, what happens when it overflows, and why fitting is not the same as being read.
Why a persuasive answer can still be wrong
Recognize the kinds of errors that polished writing can hide.
How models learn from examples
Understand parameters, loss, and why a good training score is not enough.
A short history of language models
From counting word pairs to transformers and chat assistants: each step explains a strength or weakness you still see today.
Why the same question can produce different answers
Understand generation settings and evaluate reasoning by its support.
02Using AI effectively
Write a useful brief, preserve meaning, distinguish evidence from inference, and resume work reliably.
8 lessons
Using AI effectively
Write a useful brief, preserve meaning, distinguish evidence from inference, and resume work reliably.
What a prompt is, including the part you never see
A prompt is everything the model reads before it answers, including hidden instructions the application adds. Treat all of it as visible, and write explicit goals, source boundaries, and output requirements.
Give AI a brief it can actually follow
Turn a broad request into a reviewable assignment.
Keep evidence, instructions, and inference separate
Prevent a summary from quietly becoming a source of facts.
Package source material so the model can use it
Lists, tables, JSON, and tagged documents each suit different material. Choose the format deliberately, give every source an ID, and put long material before the question.
Ask for the working: chain of thought and reasoning models
Asking a model to reason step by step improved accuracy on multi-step tasks. Newer reasoning models do this on their own. Either way, written reasoning is output to check, not proof.
Improve a draft without changing its meaning
Review substance and expression in separate passes.
Does AI learn from your corrections?
Build continuity you can inspect rather than assume.
Build a reusable task pack
Save the brief, examples, and checks that make a recurring task work.
03Legal Workflows
Build source-linked chronologies, transcript work products, research tables, and drafting workflows.
5 lessons
Legal Workflows
Build source-linked chronologies, transcript work products, research tables, and drafting workflows.
Build a chronology that survives checking
Extract events, preserve conflicting dates, and audit coverage.
Turn a transcript into useful examination preparation
Preserve questions, answers, qualifications, and source locations.
Research with AI without outsourcing the authority check
Separate finding candidates from verifying propositions.
Build a drafting workflow around the record
Move from issues to an outline, then draft and verify one section.
Use AI to challenge an argument
Turn simulated opposition into source-linked questions you can investigate.
04Tools and choosing between them
Compare tools on complete tasks, design review tables, and turn dictation into usable work.
3 lessons
Tools and choosing between them
Compare tools on complete tasks, design review tables, and turn dictation into usable work.
05Agents, skills, and automation
Understand the components, design a first bounded agent, and make recurring work recoverable.
7 lessons
Agents, skills, and automation
Understand the components, design a first bounded agent, and make recurring work recoverable.
Prompts, skills, workflows, and agents
Understand the parts before designing an automation.
Give the model the facts, or give it tools
The most reliable AI systems hand the model the information it needs. When that is impossible, they give it a small set of well-described tools and check what it does with them.
Build your first agent around a document inventory
Define a small job, limited permissions, and an observable result.
When do multiple agents help?
Assign functional work and preserve a shared source record.
Make recurring work resumable and reviewable
Specify triggers, checkpoints, duplicate handling, and failures.
Maintain project instructions without accumulating clutter
Keep stable rules, changing state, and source material in their proper places.
Choose scripts, models, and human review deliberately
Route work by observable requirements and enforce boundaries outside the prompt.
06Documents and knowledge systems
Prepare source material, understand retrieval and RAG, and create reusable precedent and matter knowledge.
4 lessons
Documents and knowledge systems
Prepare source material, understand retrieval and RAG, and create reusable precedent and matter knowledge.
Prepare documents before asking AI to analyze them
Keep originals, inspect extraction, and build a coverage manifest.
Keyword search, semantic search, and reranking
Embeddings turn text into coordinates so a system can find passages with related meaning. Learn how that works, a trick for short queries, and what a search result does not prove.
How document-grounded AI answers work
Locate failures across ingestion, retrieval, and generation.
Build a knowledge library you can actually reuse
Give precedents, matter facts, and reusable methods different homes.
07Local AI and building systems
Learn the hardware, run a model, write small utilities, connect an API, and preserve structured facts.
7 lessons
Local AI and building systems
Learn the hardware, run a model, write small utilities, connect an API, and preserve structured facts.
Understand the hardware a local model needs
Estimate memory before considering an upgrade.
Run a local model and verify the actual setup
Complete a small experiment before adding documents or agents.
Your first useful Python script: a document inventory
Read a practice folder and print a CSV without changing its files.
APIs and JSON: connect a model to a small program
Ask for a defined output format, enforce it where the API allows, and remember that valid JSON can still contain wrong facts.
Use a database to preserve facts and sources
Learn a small SQL example before designing a matter system.
Prompting, retrieval, or fine-tuning?
Choose the intervention that matches the failure. Fine-tuning has training and maintenance costs; use public or synthetic examples while learning.
Turn a prototype into a service you can maintain
Plan access, storage, monitoring, and recovery before adding users.
08Reliability and oversight
Evaluate results, verify support, recognize prompt injection, and map the full data journey.
7 lessons
Reliability and oversight
Evaluate results, verify support, recognize prompt injection, and map the full data journey.
Test an AI workflow on work you can assess
Measure omissions, unsupported additions, and total review effort.
Decide whether an AI tip deserves your time
Separate a transferable method from an unsupported claim or product pitch.
Verify a work product before relying on it
Review source support and completeness in both directions.
Recognize prompt injection, jailbreaks, and leaks
Instructions in a prompt guide normal use; they do not stop a determined user or a hostile document. Assume anything the model can see can be revealed, and enforce limits outside the model.
Map where information goes before using a tool
Turn confidentiality questions into a concrete system review.
What courts expect when you use AI
Check the rule, verify the source, and keep a record of the review.
Reading AI-assisted material from the bench
For judges and those who appear before them: what to look for, what to ask, and how courts themselves are using AI.
09Local AI, in depth
Optional depth: run a model on hardware you control. Check memory, setup, data flows and results before using client files.
10 lessons
Local AI, in depth
Optional depth: run a model on hardware you control. Check memory, setup, data flows and results before using client files.
Why run AI on your own machine, and what you give up
Keep client material on hardware you control, and understand the trade against frontier cloud models.
How a model actually runs on hardware
Weights, bits, the KV cache, reading versus writing, offloading, and mixture-of-experts, with the two formulas that predict fit and speed.
Every part of a computer, through the lens of local AI
What each component does, how much it matters for running models, and how to find out what is already in your machine.
Building a PC for local AI
Choose parts by budget, check they fit together, assemble step by step, and get through the first boot.
Devices other than a custom PC
Apple Silicon Macs, unified-memory mini PCs, dedicated AI boxes, and multi-GPU workstations, compared in one table.
Engines and front-ends: the software layers
What an inference engine does, what a front-end adds, and which combination to choose.
Choosing a model: families, sizes, formats, and licences
Pick an open-weight model for your memory, read its model card, and check its licence.
Setting it up: three tracks, easy and expert routes
Windows with NVIDIA, Windows with AMD, and Mac, each by application and by command line, and how to confirm the model is on the GPU.
Proving it works: offline, measured, and checked against the formula
An offline test anyone can follow, a benchmark, and a comparison between predicted and measured speed.
Keeping it current: engines, models, and hardware
Update the engine safely, swap models on evidence, and know when hardware is the real limit.
How to use the lessons
Each lesson explains one idea, shows it with a worked legal example, and ends with sources. Read a chapter in order, or open a single lesson when a question comes up. The technical chapter has additional setup instructions.
When you try any of this on your own work, start from public or fictional material.