Engineering
How we build legal AI
Internals of the product: what verification actually checks, how retrieval works, and what we test.
24 of 24
Cutting a BM25 Index Build From Four Hours to Ninety Minutes
How We Think About Hallucination in Legal AI
The Deterministic Layer Under the LLM
How We Keep a Legal-AI Product From Regressing
One Retriever Is Not Enough
Knowing When You Do Not Know
Teaching the Reviewer How Your Firm Negotiates
Draft the Fix, But Earn It
The Tracked Changes a Document Is Hiding
Real Word Redlines, Not a Diff View
The Law Is Not Clean Data
Extract Once, Read Many: The Shared Evidence Spine Behind a Matter Workspace
A Spreadsheet Where Every Cell Is a Cited, Verified Extraction
Catching a Fabricated [N] Before It Reaches the Lawyer
Research a Hard Question the Way a Lawyer Would: In Pieces, Then Synthesized
Hybrid Retrieval for Law, and the Citation-to-a-Subsection Problem
What Our Citation Verifier Actually Checks
Call US Law From Your Own Code: The Statutes and Regulations API
How to Verify an AI's Citations in Code, Not With Another Model
We Ran 1,341 Prompts Through Our Router to Find Out What Misroutes
What Our Citation Verifier Checks
Who in Legal AI Actually Shows Their Work?
How to Write Real Word Tracked Changes in Python