The era of vibe-research is here
Vibecoding —> Vibe-research?
The coding world has been awash the past few weeks in discourse about "vibecoding." Andrej Karpathy, an AI researcher, coined the term in an early February tweet ("where you fully give in to the vibes, embrace exponentials, and forget that the code even exists"). Although Karpathy spoke for himself, he described a phenomenon that has become plain to anyone who uses AI to code, whether or not they had preexisting knowledge. As Kevin Roose, a technology writer for the New York Times, puts it:
Vibecoding...is useful shorthand for the way that today’s A.I. tools allow even nontechnical hobbyists to build fully functioning apps and websites, just by typing prompts into a text box. You don’t have to know how to code to vibecode — just having an idea, and a little patience, is usually enough.
Likewise, I am now seeing signs that the legal vibe-research era may be upon us.
Just last week, Iqidis AI, a legal AI company, released a free product ("Researcher") that lets anyone use an AI chat interface to perform legal research on the web. The screen just looks like this:
Researcher comes in three flavors: A standard, web-based search; a standard search for "academic research" (which pulls in articles, but also tends to find statutes and regulations if you ask about them); and a "deep research" feature akin to the one released by OpenAI recently.
Since reading the Trump Administration's executive order on preventing "abuses" of the legal system and the federal courts, I've been wondering about the part of the order where Trump:
[F]urther direct[s] the Attorney General to take all appropriate action to refer for disciplinary action any attorney whose conduct in Federal court or before any component of the Federal Government appears to violate professional conduct rules, including rules governing meritorious claims and contentions, and particularly in cases that implicate national security, homeland security, public safety, or election integrity. In complying with this directive, the Attorney General shall consider the ethical duties that law partners have when supervising junior attorneys, including imputing the ethical misconduct of junior attorneys to partners or the law firm when appropriate.
Although this is, by far, the most drastic decree in the order, it is also the least enforceable. Attorneys are regulated by state bar authorities, which are—in most cases—agencies of the state supreme court and completely outside the control of the President.
It occurred to me that I didn't actually know what rules govern the process for attorney investigations in Illinois, the state where I am licensed. So I asked Researcher: "If an attorney is referred for professional disciplinary action in Illinois, what happens next?"
I’ve posted the answers as an attachment. I think they are quite good, for a five-minute, off-the-cuff answer.1
Elsewhere at the frontier of legal research, this recent video podcast with Ed Walters, an executive at the legal research company vLex, he shows how their paid AI tool (Vincent AI) can now convert an audio transcript of an oral argument into text on which the user can perform point-and-click legal research:
And so we can see right away the entire transcript of the oral argument. Now I can copy that, paste it into a Word document or something. But from that, just like other places you’ve seen in Vincent AI, Vincent reads that transcript and says, what do you want to do with this?
...
And so you see these very specific suggestions, create a concise summary of the defense arguments presented by Mr. Funk on behalf of this judge focusing on the First Amendment issues, create a timeline, draft rebuttal points. And then at the bottom, these are legal questions. So if you want to do research, and find out what the legal precedents are for First Amendment protections for judges in Ohio, you can select that. And of course, we’ll pull that from the VLex collection. What are the standards for judicial discipline in Ohio? These aren’t baked in. These are suggestions that are created by reading through the transcript of that oral argument.
The Slow Collapse of the Research Gestalt
I started law school in 2014. At the beginning of our required first-year legal writing class, we spent 30 minutes in the library with a law librarian, who showed us how to look up a case in a printed book. To do this, we used "reporters" and "digests" that were first popularized in the late 19th century by a company called West Publishing (which was later bought by Thomson Reuters, a conglomerate, in the 1990s). With these books, when you look up an opinion, you get not only the opinion itself but also a list of subject headings and short synopses of key legal decisions within the opinion. Next to each heading/synopsis is a key symbol followed by some alphanumeric codes, like this2:
In the early days of the Republic, the database of published cases in the United States was so small that there was no need for an organizing system. But as the United States grew in size and publishing got easier—making the disorganized system untenable—a publisher named John West filled the void. West, a lay businessman, succeeded because he provided two things that lawyers wanted: Access to every case and a way of organizing it. West provided the former with the cases in the "National Reporter System," which published every case in a system of reporters for different courts and regions of the country. He provided the latter with the "American Digest System," which provided short "digests" of each case under key numbers that corresponded to a certain topic.
Lawyers snapped up the West digests and reporters, and West Publishing became synonymous with the gold standard for legal research. In his 1994 article Collapse of the Structure of the Legal Research Universe: The Imperative of Digital Information, the Robert C. Berring, a law librarian, argued that the West Reporter + Digest system not only made legal information accessible, along with the case method of instruction created by Christopher Columbus Langdell, a Harvard Law School professor, it defined the "gestalt" of how lawyers actually thought about the law through most of the 20th century.
The information system based on these West products, supplemented in an important way by the Shepard's Citations System, allowed the growing database of cases to be controlled. It also set the tone for legal research training. The legal information system intertwined itself with the organization of law itself. Given Langdell's stress on categorizing the law into inflexible subject areas, and the focus on law as finding primary sources, the legal information system played an especially dominant role. How one organizes the law became the center of what the law could and did mean. While this was a conscious process for Langdell and for West, as time passed legal scholars forgot that choices had been made and began to see the existing categories as inevitable; thus the gestalt of case law was created.
Six years after Collapse, Berring published Legal Research and the World of Thinkable Thoughts. Reprising similar themes from his earlier essay, Berring writes about how Langdell, West, and William Blackstone—an 18th century Oxford lecturer on the common law—invented the world of "thinkable thoughts" available to American lawyers through the 1990s. However, by 2000, Berring noted that the users of the information systems—law students—were coming to law school as more sophisticated users of information.
So mix a technology that provides wide-ranging information, a new breed of users, and an academic setting that is separating itself from the actual practice of law, and one creates information anarchy. This jumble is fast coming to resemble the world of chaotic legal information that Blackstone found. Sound crazy? Think it through. The old classification system of West topic and key numbers can be an important element in research, but they no longer define the reality of legal thinking. The new generation of researchers is governed by the algorithms of its search engines. There is simply too much stuff to sort through. No one can write a comprehensive treatise any more, and no one can read all of the new cases. Machines are sorting for us. We need a new set of thinkable thoughts.
When a lawyer learns law in the tradition of Blackstone, Langdell, and West, they learn how to solve problems by "applying the law to the facts"; the canonical law school exam presents you with a "fact pattern" and your job is to figure out how a court might apply the law to those facts if it came up in a real case. The entire premise of this approach is that the information contained in the facts and the information contained in "the law" are mediated by a human brain (the lawyer or lawyer-to-be), and that the brain can be trained to map legal concepts on to facts. Thus, even though it was anachronistic, in 2014, for a law librarian to teach me how to look up cases in a physical book, the exercise was useful at that early stage to train my mental map of how "the law" is constructed.
Likewise, learning how to program is an exercise in learning how to provide a computer with a set of instructions to carry out a computational task. Although programming education has always been more democratic and diffuse than legal education, both operated on the premise that a lawyer or programmer's brain must be trained in the proper method for carrying out the task. Although opinions about the proper approach may have varied, it was assumed that competent practitioners maintained some arrangement of "thinkable thoughts" that they carried with them to think through a problem.
In the vibes era—where "an idea, and a little patience, is usually enough"—anyone can get a (pretty good) answer to a legal question without the need for any thinkable thoughts, other than the vague notion that they have a question in need of an answer.
A major value-add of Iqidis (over a general-purpose chatbot like ChatGPT or Claude) is that it provides links to the web resources used in the search. The PDF does not include the images of all those links.
Thomson Reuters has imported the key digest into Westlaw, its digital research product, which is where that screenshot is taken from.