Between the Briefs

The episode contains a handful of genuinely useful ideas—the 'find enough, not all' framing for discovery obligations, the observation that every major provider now shares the same foundational LLMs, and the inversion of the lawyer-technologist dynamic—but these are embedded in long biographical narrative and general enthusiasm that dilutes the per-minute idea rate significantly.

“The right objective, I think should be find enough of it so that you have a reasonable basis for taking the next step.”

“we're pretty much all working with the same foundational models. Now there's going to be a question of to what extent people can build their own small language models”

The 'sufficient vs. perfect' discovery framing and the observation that lawyer demand is now pulling AI adoption rather than vendor push are genuinely counterintuitive, but most of the content recycles familiar 'AI is like the early days of eDiscovery' analogies and broad optimism about access to justice without a real contrarian edge.

“history doesn't repeat itself, but it definitely can rhyme. You know, we're in a rhyming stage right now”

“I never once had this strategy fail me... What I have done is gone through your request and tried my best to identify those documents I think you are asking for”

Socha is a genuine practitioner with rare credentials—co-founder of EDRM, nearly 20 years of litigation practice, hundreds of expert witness engagements—and the conversation draws on lived experience rather than theoretical frameworks, though his current role as SVP of Brand Awareness layers in some vendor positioning that limits pure operator credibility.

“I do have both the grounding of all those years as a practicing lawyer and then the grounding of all those additional years as a consulting expert witness and consulting expert and then testifying expert witness”

“I got a court order that required them to turn over the entire contents of the server that ran their whole company”

The episode is punctuated with genuinely specific details—two named 1987 providers, the exact St. Paul meeting date, a $1M project cost, 25-second query response times, and 1 million document demo sizes—but these are interspersed with plenty of hand-waving ('various types of short messages,' 'roughly 20 different pathways') and some figures are hedged or approximate.

“as of, say, 1987, as far as I know, two service providers that focused on eDiscovery. One EED Electronic Evidence Discovery out of Seattle, and another, IBIS Consulting out of Providence, Rhode Island”

“That was a million dollar project doing that five or so years ago.”

The hosts ask a few substantive questions—notably the AI recall/completeness challenge—but mostly serve as biographical prompters, allowing extended monologues without follow-up on interesting claims (e.g., the million-dollar case that was never used at trial, or what the 20 agentic pathways actually are), and the closing 'hottest take' question is deliberately soft.

“How do you account for that from just a user experience perspective? How do you let a user feel like, yes, it might be returning meaningful information to me, but how do I know it's returning all of the meaningful information to me?”

“Are there things that scare you? Are there things that sort of excite you tremendously?”