About LexED2

This site follows a simple principle expressed in the Tao Te Ching, Chapter 2: 生而不有, “to create without possessing.”

The educational materials here are shared in that spirit. Where I hold the necessary rights, they are released under CC0 so others may use, adapt, and build on them freely.

How LexED2 was made

I built LexED2 with Claude, the AI model made by Anthropic (opens in a new tab). I designed the courses and decided how every page is sourced and checked, and Claude drafts the pages to that design. I read and approve every word, and I make edits where they’re needed.

I decided what each course covers and in what order, and I read and approved every course outline before a page was built. I designed the activity format, and I shaped each activity around something a student learns best by doing. I set the rules every page follows, from where its sources may come from to how it cites a case. The voice is mine in a narrower sense. Claude writes to a style guide built from my own writing: the essays I wrote for Sudbury Beach School (opens in a new tab) from 2015 to 2017, and the appellate briefs I wrote as a lawyer. The guide rests on Bryan Garner’s rules for legal writing. Claude also wrote the code behind the site, and I decided how it looks and works.

I use AI this way on purpose, and I see no reason to hide it. I ask the same of my students: most activities end by asking you to share one of your AI conversations and say what it shows. Consider this section my chat link.

I’ve written on my own for decades, in briefs, course materials, and essays, and I still enjoy it. Anyone who has done it knows how hard and slow good writing is, and how rewarding. But that wasn’t my goal here. LexED2 needed a great deal of explanatory prose written to a fixed format, and that kind of writing suits AI better than personal or creative writing does. So I spent my time where my expertise counts: the design, the sources, and the checking.

I think AI will likely come to dominate how most people write, and writing done without it will grow rarer. That doesn’t make it any less worth learning and practicing. I admire every writer who still writes every word. I’ve read hundreds of books of every kind, and a skilled author still leaves me in awe.

Why the courses work the way they do is a separate question, and my Teaching Philosophy answers it.

My background, from two decades of practicing law in Atlanta to teaching at Daytona State College, is on my CV.

Where the sources come from

A page doesn’t cite whatever a search turns up. Sources are ranked, and a page uses the highest rank that covers its point:

The text of constitutions, statutes, and court rules. Every Florida statute comes from the Florida Legislature’s site (opens in a new tab), and every Florida court rule from The Florida Bar’s rules page (opens in a new tab).
Court opinions, from the court’s own website. When the court keeps no stable free copy, the link goes to CourtListener (opens in a new tab), the free archive run by the nonprofit Free Law Project.
Regulations, agency guidance, and official databases.
Reports of the Congressional Research Service.
A short, named list of secondary sources, such as law-library research guides and plain-English explainers, used only when no primary source covers the point.

Law-firm blogs, Wikipedia, and commercial aggregators are left out unless nothing else covers the point. News stories and law-firm client alerts appear mainly in the Current Context sections, to report that something happened. When a Current Context says what a court held, it links the opinion or a neutral summary of it.

How the pages are checked

Claude doesn’t answer legal questions from memory here. It works through connectors to four legal research tools. The connectors use the Model Context Protocol (opens in a new tab), an open standard that lets an AI model search an outside database directly.

Descrybe, a case-law research service: finding cases by the rule they apply, pulling an opinion’s passages on a point, checking quotations word for word, and seeing how later courts treated a case.
DingDuff, a legal research service: reading the full text of opinions and statutes, and following federal dockets filing by filing.
CourtListener: checking reporter citations in batches, and supplying the link a page gives for a case.
Oyez, through a connector of my own to the Oyez Project’s (opens in a new tab) Supreme Court records: finding a case’s dates and vote, and what each Justice asked at oral argument.

These tools find the law, but none of them gets the last word, because each one is sometimes wrong. A summary misstates a holding. A record gives a decision date a month late. A citator shows a case as good law months after a later decision undercut it. So every statement on a page is checked against its primary source:

a holding, quotation, vote, date, or dollar figure, against the opinion itself;
an event in a pending case, against the entry on the court’s docket;
a statute or rule, against its official text;
a case link, by opening it and confirming it’s the right case.

Whether a case is still good law takes three checks, because citators lag behind new decisions. The first is the citator’s signal. The second is a search for newer cases on the same point. The third is the watch list each course keeps of the cases it follows. The pending cases on those lists get a docket check when I ask for one, which pulls their new filings so a page can change when a ruling comes down.

Scripts check the pages too. A format checker holds each rebuilt activity and reading to its template, to the site’s writing rules, and to the citation form of Florida Rule of Appellate Procedure 9.800. The site won’t build if a page or a quiz is malformed. Every Monday, one script opens every link on the site and reports any that has broken, and another tests every page against the WCAG 2.1 AA accessibility standard.

Then I read it myself. Once a page is published, I read the whole page on the live site and click every link on it. I’m checking that each link opens what the page says it does, and that the page teaches what I meant it to. Anything wrong goes back through the same process, and the corrected page passes every check again.

What still gets through

This process catches a lot. While the courses were being built, it caught wrong docket numbers and holdings stated more broadly than the court stated them. It caught one case’s rule credited to another case, and sources that didn’t say what a page’s plan claimed they did.

Some pages still carry text from the first version of LexED. They’re being rebuilt a course at a time, and each goes through this process as it is.

No process catches everything. If you find a mistake, tell me. Students can reach me through their course in Falcon Online: name the page and what’s wrong, so it can be fixed.

License

Except where otherwise noted, educational materials on this site for which I hold the necessary rights are dedicated to the public domain under CC0 1.0 Universal (opens in a new tab).

They may be copied, adapted, shared, and reused for any purpose without permission.

Third-party materials, trademarks, logos, and other content for which I do not hold the necessary rights are not included in this dedication.

The fonts are among them. The site is set in Charter, designed by Matthew Carter and released by Bitstream, and in IBM Plex Sans and IBM Plex Mono, released by IBM under the SIL Open Font License (opens in a new tab). Bitstream Charter is a registered trademark of Bitstream Inc.

The photograph at the top of the home page is Ansel Adams’s Evening, McDonald Lake, Glacier National Park (opens in a new tab), made for the National Park Service between 1933 and 1942. It is in the public domain as a work of the federal government, and the National Archives holds it.

My photo on the CV is not part of the dedication either.