Research & Dialogue¶

April 28, 2026
in Legal Analysis, Research & Dialogue
16 min read

AI Review Got Cheaper. Bad Collection Got More Expensive.

Fragmented evidence — broken chains, shattered conversations, torn document fragments — funneled through a sleek portal into neat stacks of color-coded documents

AI review is getting cheaper, fast.

Relativity bundled aiR for Review and aiR for Privilege into standard RelativityOne packaging. Everlaw made single-document AI actions free. GenAI-assisted review pricing is now clustering in the $0.11–$0.50 range — challenging the $0.50–$1.00 band where human review still lives.

The industry is celebrating the new unit economics.

But nobody is asking what those AI models are actually reviewing — or how much of it should never have entered the pipeline.

April 23, 2026
in Research & Dialogue, Standard Development
11 min read

What One Tenant's Measurement Suggests for the RGR Staircase

A tape measure extended across a document on a dark desk, teal layout blocks beneath the tape — measurement applied to content that had previously been intuition

A first tenant-bounded measurement of linked-document version drift, on a platform-versioning measure — and what it suggests for the standard's conformance staircase.

April 7, 2026
in Research & Dialogue, Standard Development
8 min read

The Dog, the Tail, and the Staircase

Staircase ascending into darkness — the first step lit in teal

Craig Ball's engagement with the RGR Standard surfaced a real vulnerability in the conformance model. Here's the conversation — and what it led to.

April 2, 2026
in Research & Dialogue
9 min read

I Tried to Break the Preservation Gap Argument. Here's What's Left.

Steel beam under stress — what holds after pressure testing

Before I brought this argument to the profession, I tried to break it. Two AI systems argued the Preservation Gap under Rule 37(e) — and the objections that survived forced significant revisions.

March 26, 2026
in Research & Dialogue
7 min read

The Simplest E-Discovery Test Most Tools Would Fail

Two document versions — one sharp, one wrong

When a SharePoint link resolves successfully but the produced document is not the version that existed at send time, the evidentiary record is silently wrong.