Reporting Back from CALIcon

These are a few notes from CALIcon 2026, held in Ann Arbor, Michigan in early June. This was my first CALIcon, and I’ll definitely be back; lots of practical takeaways, new ideas, and plenty of fun too. The session recordings will hopefully be posted in this Youtube playlist. Only the keynote (well worth watching) and one session have been posted so far, but I assume there’s more to come.

The hands-on vibe coding session from David Colarusso, Rebecca Fordon, and Jennifer Wondracek was a real highlight. (You can find their materials at https://lawascode.com/.) David Colarusso shared some vibe-coded tools he’s used in his classroom, including a cite-checking simulation that demonstrates to students how they might over-rely on automated tools in high-stress situations. Jennifer Wondracek followed up with a general outline of a vibe-coding project and emphasized the constraints and precautions that any project should account for. Rebecca Fordon walked us through the project stages in more detail: identify a pedagogical pain point; pick your tool; use AI to develop a detailed plan; have AI build it; iterate and refine (and iterate, and reiterate, and…); and finally, deploy. Then they set us loose to create our own single-page HTML apps. In about 10 minutes, I had an interactive HTML app that helped illustrate the difference between federal slip laws, the Statutes at Large, and the U.S. Code. The presenters did a great job showing us both the why and the how, and it was pretty exciting to move from idea to prototype in just a few minutes.

On the more theoretical side, I found Alex Zhang’s presentation of a new framework for validating AI output very interesting. I’ve been thinking a lot about the need to go beyond cite-checking when working with AI output, and I’m excited to read her forthcoming paper on this framework. She identified a major problem with how the legal profession talks about validating AI output: we’ve focused narrowly on verifying individual citations, checking that a citation points to a real case that actually includes the quoted language or stands for the proposition for which it’s cited. However, generative AI output doesn’t just produce citations; it produces propositions synthesized from multiple cases and other input, and competent attorneys need to be able to validate not only the individual citations but also the synthesized propositions contained in generative AI output. Zhang pointed out that different research tools – digests, keyword searches, natural language searches, generative AI, agentic AI – produce different epistemic objects, each requiring a different method of validation.

For a more practical angle on validating AI output, I recommend Eli Edward’s session on using actual court filings to teach students about hallucinations in filings. Besides being useful as verification exercises, using the actual filings can show students how hallucinations might actually appear in practice, and the real consequences of filing unreviewed AI output. Edwards included many methods for incorporating filings into class exercises and assignments, and close with some tips for finding new cases involving AI hallucinations (unfortunately, they just keep coming).

It was also refreshing to attend a session that had nothing to do with AI at all – Dan Quiggen’s session on using Visual Basic for Applications to create useful macros for Excel and Word documents. There’s still some residual fear of macros, and there’s a bit of a learning curve to learning VBA itself, but it’s completely built in to Excel and Word – no additional installations required – and it has capabilities way beyond what you can do with the regular Excel functions. I’ve used VBA a couple of times, but only in the “copy/paste some code from Stack Exchange and hope it works” sense, so it was great to learn a little more about what I was actually doing and get some ideas for future uses.

Overall, I enjoyed the professional conversations with significant substance and depth on AI and legal education; it’s taken a little while, but I think we’re finally moving beyond the general or hypothetical conversations about how AI will impact our work, and into the specific, practical discussions about what we’re actually doing about it. And although the conference was definitely AI-heavy, it was also nice to see some room for other technologies and for pushback on reliance on AI.