BE THAT LAWYER

The episode contains a handful of genuinely useful operational insights - offline local LLMs for client-data safety, the irony anecdote of a firm locking its own lawyers out of Claude - but these are surrounded by extended throat-clearing, motivational clichés, and sponsor reads that dilute the useful-ideas-per-minute ratio considerably.

“I pulled down more than 150 filings. I think it was 161, instruct AI about, you know, read them all, figure out the order that they were filed, and retitle the documents so that we can have a common language...And it literally took five minutes for me to get these answers.”

“the only people in the room that were not allowed to access Quad, that were literally locked out of CLAUDE on an IT basis were the hosts of the event”

The 'nobody fires the starting gun' framing and the offline air-gapped LLM approach for confidentiality are genuinely fresh angles, but the broader thesis - billable hour is dying, AI will displace junior lawyers, get literate or get left behind - is some of the most recycled commentary in legal tech discourse right now.

“nobody is coming to save them. Nobody's going to say, now is the time for you to educate yourself about AI. No one's going to fire the starting gun.”

“if you have sufficiently robust hardware to download open source AI models from the Internet and then you can completely disconnect your computer from the Internet and it can run”

Baker is a legitimate practitioner with 25 years of experience including partnerships at AmLaw 100 and 200 firms who has actually built AI tooling himself - credible and relevant. His candid admission that he has no original thoughts and is purely synthesizing others caps his ceiling as a source of insight.

“I was in big law for the first 19 years of my career. The last job I had before I started my law firm in 2019, I was uh, a partner in AmLaw200 firm where I was for 14 years.”

“there probably won't be any original thoughts from me on this, uh, podcast. I'm mostly synthesizing what I'm hearing from others”

The episode punches above average on specificity for this genre: named dollar figures ($18K bill, $10M/$8.5M flat-fee example), named tools (Replit, Claude Code, Cowork), concrete targets (161 filings, 475 website pages, 3 hours to 3 minutes), and a named external authority (Dan Schnerbush). Several claims, however, remain unsubstantiated predictions.

“a couple weeks later I got a bill that was almost $18,000”

“the goal is to take a three hour contract review, redline task, take it down to three minutes”

The host consistently validates rather than probes: he restates the guest's points back, offers his own analogies (tractors, sandboxes), and never challenges the bold predictions or asks about failure modes, cost-benefit specifics, or workflow detail. The interview functions as a warm conversation between acquaintances rather than a rigorous knowledge extraction.

“Yeah, uh, yeah. And I think we're all sort of like playing in the sandbox and trying to figure this out as we go.”

“Yeah, well, that's really cool.”