What changes for lawyers when an AI model stops writing and starts deciding?
In Episode 3 of the LegalQuants podcast, host Victor Wang talks with Dan Sito, head of the Tax practice at General Legal and former partner and first AI Innovation Partner at Perkins Coie. Dan explains Jev, TypeSafe AI's classifier model that answers with a probability instead of generating text, using a Magic 8 Ball he built to understand it. Victor describes how Jev turned his agent-swarm approach to discovery and due diligence into a near-instant first-pass filter.
The conversation moves to legal design: why interactive HTML deliverables beat the email-then-call loop, Dan's /legaldesign skill, and Victor's interactive client questionnaires. Then come voice-driven charts, a whiteboard that draws while you talk, software built for one person, and an 83(b) election filled in two seconds. They close on bionic lawyers, why not using these tools is now a choice, and why we still ask a human when the risk is real.
Dan Sito at General Legal: https://general.legal/attorneys/dan-sito
Jev by TypeSafe AI: https://typesafe.ai/
Dan's /legaldesign skill, in the LegalQuants plugins: https://www.legalquants.com/lqplugins
LegalQuants Residency: https://residency.legalquants.com/
LegalQuants: https://www.legalquants.com/
00:00 Cold open
00:50 Welcome: Dan Sito of General Legal
01:40 Dan's ChatGPT moment
05:04 More legal work, more creativity
07:17 Every model is a paradigm shift
11:51 What is Jev?
12:43 Jev example #1: the Magic 8 Ball
18:25 From agent swarms to Jev
22:44 Legal design: HTML instead of email
30:30 Victor's client questionnaires
33:23 Jev example #2: voice-driven charts
36:52 Jev example #3: the live whiteboard
39:23 Software for one
42:24 Jev example #4: an 83(b) in two seconds
46:33 Speed, focus, and cognitive load
49:54 Bionic lawyers
53:02 Still ask the human
