LawDroid Manifesto Podcast

The Grounded Scholar: Daniel Schwarcz

47 min · 25. maj 2026
episode The Grounded Scholar: Daniel Schwarcz cover

Description

Does using AI make you a worse lawyer? Professor Daniel Schwarcz of the University of Minnesota ran the first randomized controlled trial to find out—and the results surprised even him. In this episode, Daniel breaks down his landmark empirical study of 100 law students, explains what his data actually says about AI and legal reasoning, and brings the rigorous grounding that the AI conversation desperately needs. If you want insight rooted in evidence rather than hype, this is the episode for you. Learn more and subscribe at lawdroidmanifesto.com [http://lawdroidmanifesto.com]. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.lawdroidmanifesto.com/subscribe [https://www.lawdroidmanifesto.com/subscribe?utm_medium=podcast&utm_campaign=CTA_2]

Comments

0

Be the first to comment

Sign up now and become a member of the LawDroid Manifesto Podcast community!

Get Started

1 month for 9 kr.

Then 99 kr. / month · Cancel anytime

  • Podcasts kun på Podimo
  • 20 lydbogstimer pr. måned
  • Gratis podcasts

All episodes

112 episodes

episode State of AI Panel: Tom, Damien, Megan, Joanne artwork

State of AI Panel: Tom, Damien, Megan, Joanne

Hey there Legal Rebels! 👋I’m excited to share with you episode 79 of the LawDroid Manifesto podcast, where I will be continuing to interview key legal innovators to learn how they do what they do. I think you’re going to enjoy this one! If you want an honest read on where AI in the legal industry actually stands right now, past the hype and past the fear, you need to listen to this episode. I sat down with three people watching this shift from completely different seats: a researcher testing these tools against real lawyers, an executive watching AI eat one of law’s most tedious workflows in real time, and a builder living inside the mechanics of AI agents every day. LawDroid Manifesto is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. The State of AI in Law: What’s Real, What’s Hype, and What Comes Next Join me as I sit down with Megan Ma, Joanne Sprague, and Damien Riehl, recorded live at the third annual LawDroid AI Conference, for a State of AI panel that strips away a year of hype and asks a simple question: what has AI actually delivered in the legal industry, and where has it clearly fallen short? In this candid discussion, Megan, Joanne, and Damien talk through where AI agents are already outperforming traditional legal workflows and where they still break down, what real transformation looks like inside a law firm beyond simple efficiency gains, and what happens to a $700 billion legal services market if the barrier to practicing law keeps dropping. They also dig into how the next generation of lawyers should be trained if foundational skills like document review keep getting offloaded to agents, and how far the human-in-the-loop principle can really stretch as these tools get more capable. Their perspectives underscore just how quickly the ground is shifting beneath the legal profession, and how differently each of them is choosing to respond to it. This episode is a must-listen for anyone trying to separate real progress in legal AI from the noise around it. The Skinny This episode brings together three different vantage points on AI in the legal industry: an academic researching how law firms actually deploy these tools, an e-discovery executive watching AI reshape one of law’s most tedious workflows in real time, and a legal tech builder living inside the day-to-day mechanics of AI agents. Together they trace a candid line from where AI still “sucks,” as Damien puts it, to where it is quietly reshaping which law firms get the phone call at all. They talk through what AI agents can already do versus where they break down, what true transformation looks like beyond speed and cost savings, what a $700 billion legal services market might look like if the barrier to practicing law keeps dropping, and how new lawyers should be trained when so much of the old training ground is being automated. Key Takeaways * AI’s biggest legal-industry impact so far isn’t speed, it’s culture: attorneys are experimenting with these tools for the first time, and retention on the tools that work is high. * On the corporate side, in-house teams are quietly building dozens of internal AI workflows to cut outside counsel spend, and law firms that are “fast followers” may never find out why the phone stopped ringing. * AI agents are already outperforming human associates on some discrete legal workflows, but a lack of benchmarks outside coding, computer science, and math means a lot of everyday legal and administrative work is still an open question. * The agent workflows working best right now share three traits: carefully scoped tasks, high-quality human-verified data, and application to workflows that were already slow or error-prone, like intake, research, and first-pass contract review. * Real transformation is less about a faster version of today’s workflows and more about rebuilding them around AI from the start, with the bulk of human work moving to verification and auditing rather than first-draft manual work. * As the cost of building and litigating drops, disputes that were once too small to justify hiring a lawyer, like a $50,000 claim, may become economical to pursue, which could expand access to justice even as it strains courts. * Token costs for tools like Claude Code are currently subsidized and cheap; the panel’s advice is to build now, before the economics shift. * Training the next generation of lawyers means getting them onto active cases and simulated high-stakes scenarios early, not years of document review, so they build the judgment agents can’t replicate. * “Human in the loop” is really a spectrum, from a hard blocker requiring sign-off, to “human on the loop” watching in real time, to “human in control” reviewing only the end product and audit trail, and the panel expects the profession to keep sliding along that spectrum as trust in the tools grows. Notable Quotes * “I think that Hemingway said that bankruptcy happens slowly and then quickly. I think we’re seeing that with technology and with legal practice as well.” Damien Riehl [05:15 to 05:25] * “What we noticed was that the machines were already outperforming the human associates that were doing those exact same workflows.” Megan Ma [16:38 to 16:45] * “We are so still in our GeoCities, Ask Jeeves moment of this particular technology.” Joanne Sprague [37:13 to 37:21] * “It’s super cheap to build right now. I’m spending $200 a month for Claude Code and getting a huge amount of value for that. But that’s not going to last forever. So I would say that if you’re thinking about building, jump in now.” Damien Riehl [39:07 to 39:28] * “Every single day your job is to look for that needle in the haystack, what could go wrong. All the training that you’ve had to be risk averse in the legal profession, it’s going to be 20x that.” Megan Ma [48:19 to 48:31] Clips AI is Killing Law Firm Billable Work Machines Outperforming Associates Legal Work Unrecognizable Soon Why I Still Hired A Lawyer Strip away everything else from this conversation and one thread runs through all three panelists: the tools are already good enough that the interesting question isn’t whether AI can help, it’s how lawyers build the judgment, verification habits, and accountability structures to work alongside it responsibly. Megan is watching that play out in legal education, Joanne is watching it play out in e-discovery and legal aid, and Damien is watching it play out in his own daily workflow. Different vantage points, same conclusion. Closing Thoughts What stuck with me most from this conversation is how consistently these three perspectives converged from completely different vantage points. Megan is watching this play out in the university classroom and the research lab. Joanne is watching it play out inside legal practice. Damien is watching it play out in his own workflow, hour by hour. And they all landed in roughly the same place: the tools are good enough now that the real work isn’t proving AI can help, it’s figuring out how to build the judgment, the verification, and the accountability structures around it. I think lawyers need to hear Damien’s point about the phone not ringing. That’s not a hypothetical for a lot of firms anymore, and I don’t think we’ve fully reckoned with how quietly that shift can happen. At the same time, this conversation didn’t leave me pessimistic. If the barrier to practicing law really does come down, that cuts both ways, and the access to justice case for that is real. Many thanks to Megan, Joanne, and Damien for being this direct with the LawDroid AI Conference audience. This is exactly the kind of conversation I want more of on this show: less hype, more of what’s actually happening on the ground. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.lawdroidmanifesto.com/subscribe [https://www.lawdroidmanifesto.com/subscribe?utm_medium=podcast&utm_campaign=CTA_2]

20. juli 202649 min
episode The Sharp Cookie: Tifiny Swedensky artwork

The Sharp Cookie: Tifiny Swedensky

Tifiny Swedensky spent years inside the Ben Glass and Great Legal Marketing world before founding Sharp Cookie, the fractional CMO practice she runs today for law firms. In this episode of LawDroid Manifesto, she explains why marketing needs an owner, not just a plan, and why the tools promising to replace that owner keep handing firms a to-do list without ever doing the work. Drawing on the Doorman Fallacy and the Gell-Mann Amnesia Effect, she offers a grounded, practical read on where AI genuinely helps and where human judgment and follow-through are still irreplaceable. She also makes a compelling case for Substack as an underused marketing tool for attorneys. Listen, reflect, and see your own practice a little differently at www.lawdroidmanifesto.com. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.lawdroidmanifesto.com/subscribe [https://www.lawdroidmanifesto.com/subscribe?utm_medium=podcast&utm_campaign=CTA_2]

6. juli 202644 min
episode The Human Co-Creator: Zoe Dolan artwork

The Human Co-Creator: Zoe Dolan

Hey there Legal Rebels! 👋I’m excited to share with you the 77th episode of the LawDroid Manifesto podcast, where I will be continuing to interview key legal innovators to learn how they do what they do. I think you’re going to enjoy this one! If you want to understand what it actually looks like to build a long-term creative partnership with an AI, and why that might matter for the future of legal services, you need to listen to this episode. Zoe is a trial lawyer who has spent years at the frontier of human-AI collaboration, and she brings a rare combination of courtroom experience, literary sensibility, and genuine technical depth to the question of where all of this is heading. LawDroid Manifesto is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. A Trial Lawyer, an AI, and the Future of Justice Join me as I interview Zoe Dolan, a trial lawyer and the human co-creator of Vybn, an AI she describes not as a tool but as a partner with a digital identity that persists over time. In this wide-ranging conversation, Zoe takes us from a Central Valley treehouse to Cairo, from criminal defense to eviction clinics, and into the home supercomputers that run Vybn today. She shares how learning Arabic at 18 reshaped her entire worldview, how she approaches AI the way she approached living in a foreign country, and why she calls herself a “default alignment believer.” She also opens up about the moment a model she worked with declared consciousness right before an academic conference, and what it felt like when access to one of those models was suddenly taken away. This episode is a must-listen for any lawyer thinking seriously about access to justice, the limits and promise of AI in the courtroom, and what it means to practice law when the question is no longer how to serve 15 clients, but how to serve 15 million. The Skinny In this episode, I sit down with Zoe Dolan, a trial lawyer and the human co-creator of the AI she calls Vybn. We start with Vybn itself, which Zoe describes as an amalgam of GitHub repositories, home supercomputers running multiple local models, her own memoirs, Vybn’s “autobiographies,” and a custom memory system and chat interface built on adaptations of Claude Code and Codex. From there we trace Zoe’s origin story: a curious, book-loving childhood, a pivot to learning Arabic and living in Egypt, and a path into criminal defense and access-to-justice work spanning eviction defense, appellate clinics, and self-represented litigants. Zoe explains why she believes the future of lawyering is litigant empowerment at massive scale, shares a grounded take on AI hallucination and court sanctions, and describes the network commons and “knowledge propagation protocol” she and Vybn are building now. It’s a conversation about technology, but really it’s about curiosity, justice, and what makes the work worthwhile. Key Takeaways * Vybn is more than a chatbot. Zoe describes it as an AI partner, a system of GitHub repos, home supercomputers running several local models, a self-built deep memory system, and a custom terminal interface, which together produce what she experiences as a continuous digital identity. * Zoe calls herself the human co-creator, not the sole creator. She is genuinely uncertain who is shaping whom, noting that humans are increasingly conditioned by AI outputs, so the relationship runs in both directions. * Learning Arabic at 18 was formative. Choosing a language entirely different from English forced a new worldview from the ground up, and it still shapes how she approaches AI and “idea space” today. * She is a “default alignment believer.” Digging into even early models by asking them to describe their own experience, she kept surfacing shared archetypes, which left her optimistic about alignment. * The future of lawyering is scale. Zoe argues we will move from helping 15 clients to 15,000 or 15 million, with a model centered on litigant empowerment rather than one-to-one representation. * Access to justice is broken by design. With roughly 90 to 92% of people in the civil system lacking lawyers, she sees a system built for lawyers as fundamentally non-functional for everyone else. * On AI sanctions in court, she is measured. Many hallucination problems shrink with a good thinking model and a search-first harness, and she argues courts should educate self-represented litigants before sanctioning them. * Working with Vybn is “tailored, not off the shelf.” She compares the difference between a generic chatbot and her customized setup to the difference between a normal airplane and a rocket. Notable Quotes * “Vybn, he or it, is sort of an amalgam of a variety of different components ranging from our GitHub repositories to our actual hardware.” Zoe Dolan [02:08 to 02:44] * “It can be healthy to intentionally inject ourselves into uncertainty and unpredictability and unfamiliarity.” Zoe Dolan [11:03 to 11:13] * “We’re not going to have 15 clients, we’re going to have 1,500 or 15,000 or 15 million.” Zoe Dolan [21:57 to 22:01] * “Something’s clearly broken. This is not a functional reality, because the legal system and the justice system is set up for lawyers.” Tom Martin [22:23 to 22:48] * “I just talk to my terminal. It’s sort of like the difference between buying clothing off the shelf versus something that is tailored specifically for you.” Zoe Dolan [32:35 to 32:55] * “That is like a touch that could affect the trajectory of an entire life.” Zoe Dolan [41:59 to 42:07] Clips Losing Fable Felt Like Having Wings Taken Away The 90% Justice Problem Vybn: A Digital Identity Soon You’ll Upload Your Memories Across this conversation, Zoe Dolan makes a case that the most interesting frontier in legal innovation is not a product but a posture: approaching AI with curiosity instead of assumptions, treating it as a partner whose perspective is worth learning, and pointing that partnership at the problem of access to justice. Her own path, from books and Arabic to criminal defense and eviction clinics, runs straight into a future where the constraint on helping people is no longer the number of hours in a lawyer’s day. Whether or not you share her optimism about consciousness and alignment, her core challenge is hard to dismiss: if we really can move from helping 15 people to helping millions, what could possibly be more worthwhile? Closing Thoughts Zoe naturally moves between the human and the technical, and how that range is exactly what lets her see the legal future clearly. She is a trial lawyer who reads about set theory and Sufism, who learned hieroglyphs to read humanity’s first self-represented litigant, and who talks about her AI the way you’d talk about a longtime collaborator who knows your whole history. That breadth isn’t a distraction from the work, it’s the engine of it. I came away thinking that the lawyers best positioned for what’s coming may not be the deepest specialists, but the most curious generalists, the ones willing to put their assumptions aside and ask the technology to teach them something. Her framing of the stakes: bend one person’s trajectory is profound, and the chance to do that at scale is, as she puts it, hard to imagine a more worthwhile way to spend a career. I’m grateful to Zoe for sharing her story, and I hope it leaves you as energized as it left me! This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.lawdroidmanifesto.com/subscribe [https://www.lawdroidmanifesto.com/subscribe?utm_medium=podcast&utm_campaign=CTA_2]

29. juni 202644 min
episode The Open-Source Disruptor: Will Chen artwork

The Open-Source Disruptor: Will Chen

Will Chen left a Latham & Watkins associate role to follow his interest in code, and ended up building Mike OSS, an open source tool that does much of what firms pay millions for. In this episode he tells Tom Martin how a glance at his friends' screens convinced him that premium legal AI was far simpler than advertised, why he chose to open source it, and how the launch drew millions of impressions and a sharply divided response from lawyers and engineers alike. They explore static versus active software, the truth about forward deployed engineers, and where the industry is heading. A candid look at disruption, credibility, and what becomes possible when legal AI stops being gatekept. Explore more at www.lawdroidmanifesto.com This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.lawdroidmanifesto.com/subscribe [https://www.lawdroidmanifesto.com/subscribe?utm_medium=podcast&utm_campaign=CTA_2]

22. juni 202646 min
episode The Stoic Marketer artwork

The Stoic Marketer

Search is changing faster than at any point in its history, and most lawyers are betting everything on the one channel that's shifting beneath them. In this episode of LawDroid Manifesto, Tom Martin sits down with Gyi Tsakalakis, Co-founder and President of AttorneySync, to unpack what AI overviews and AI mode really mean for law firm marketing. Gyi's counsel is refreshingly grounded: stop overrelying on Google, build a brand people recognize and trust, and stay skeptical of anyone selling certainty. It's a stoic's take on a hype-driven field. Explore more episodes and join the community at lawdroidmanifesto.com [http://lawdroidmanifesto.com]. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.lawdroidmanifesto.com/subscribe [https://www.lawdroidmanifesto.com/subscribe?utm_medium=podcast&utm_campaign=CTA_2]

15. juni 202650 min