
prof & Dir Law & Technology, @NorthwesternLaw & @NorthwesternEng; @CodeXStanford Affiliated Faculty; frmr #BigLaw partner; AI4Law; Law4AI; ppl+process+data+tech
Chicago, IL @NorthwesternU
Joined February 2013
- Tweets12.1K
- Following7.7K
- Followers10.2K
- Likes15.7K
Pinned Tweet
Law must be a part of computer science curricula
Agreed, @mireillemoret!
Excellent @CACMmag piece!
So much important CS+Law work to be done!
We're onboard at @NorthwesternU w interdisciplinary CS+Law courses in both @NorthwesternEng & @NorthwesternLaw!
cacm.acm.org/magazines/2021/…
Daniel W. Linna Jr. retweeted
Interesting comments from @Jack_Newton talking with @BobAmbrogi at @GoClio’s #ClioCon a few weeks ago.
Bob received the below answer as to Clio's moat now that the frontier AI companies are making a move into legal Newton's answer is data.
1) 1 billion legal documents from vLex, spanning 110 countries
2) Another 1 billion court filings through Clio Docket
3) 18 years of operational context from law firms
Newton also had an interesting comment about MCP connectors. He says stitching together 10 different systems through MCP connectors is "a bit of a pipe dream," and that it produces a fragile "Rube Goldberg machine," which makes it hard to govern securely.
I'm still one to believe that one of the moats is the insight of legal professionals, and with 20 years of that, millions of pieces of data.
lawnext.com/2026/11/on-lawne…
Daniel W. Linna Jr. retweeted
People ask me how an AI-native law firm actually works.
Spector Louden Law is one lawyer, me, running a team of AI agents I built and supervise, organized the way a company organizes departments.
Claude is the engine, working through Claude Duet as my chief of staff: a coordinating agent that triages my inbox, routes work to the right desk and flags what needs me.
Under it, every function a firm would normally staff has its own agent. Intake and Conflicts screens new leads, runs conflicts checks, drafts engagement letters and sends them out through DocuSign.
Billing captures my time, catches double entries and feeds Clio through a sync script I built.
Firm Ops maintains the routines and custom skills that encode how my firm does things.
Business Development scans the news for post ideas and tracks referrals, and Events handles speaking invitations.
Every client matter gets its own agent too, doing the work of an associate and a paralegal: reading the whole file, building timelines and facts sheets that tie every number to its source, drafting letters and redlines, watching client mail and prepping me for calls.
Google Workspace is the system of record, Granola transcribes client calls, and Lexis and Midpage back the research and cite checks. Each department sends my chief of staff a one-line status, so I see at a glance what needs me.
That's the work a small firm would hire five or six people to do.
What the agents never do is decide. I clear conflicts and choose which clients to take. Advice, strategy and every number in a negotiation are mine, and nothing leaves the firm until I've read it and said send.
So the machine does the reading, the routing and the first draft, and my clients get a senior lawyer who knows their file cold. BigLaw craft, tech-firm speed.
This post is for the five lawyers on X who don’t want VC funding to start an AI native law firm. carolynelefant.substack.com/…
Daniel W. Linna Jr. retweeted
Our firm spends over $10M a year on legal fees. Almost all of it on M&A transactions that are all pretty similar.
The funny thing is we have very few disputes to point to where you say this is the value we got for that legal spend.
My understanding is that AI is not there yet. It can't really do a good job marking up a legal document while understanding the context for the transaction. But I am pretty sure in the next 5 years it will be.
We could still have a lawyer look things over but have AI do 90% of the work. And we would be willing to put up with a little lower quality if we could significantly cut our spend.
Change in the legal world is coming....
It is becoming incredibly clear to me that VC and AI people fundamentally do not understand the legal industry or what lawyers do. Like, not even a little.
Here’s a basic hint: people don’t pay lawyers for their precedent library, and running with a template document that you don’t understand is akin to buying a $5,000 suit that isn’t tailored to you.
I used to assist startups relatively pro bono about a decade ago when these “buy your LLC docs” were incredibly popular, and there are just so many examples of how these things go incredibly wrong.
You pay a lawyer for their legal counsel, which is wisdom from experience. You aren’t paying them to regurgitate their precedent library unless you really just want a legal opinion attached to the docs.
Daniel W. Linna Jr. retweeted
Update: Can AI agents use Pangram to beat Pangram? In small tests, I found simple prompts to the Codex allowed it to beat Pangram. The final output was 100% human generated. Prompts that helped:
- Have the agent develop a style guide based on past writing
- Have it develop an "anti-style guide" based on AI writing tics that people have complained about
- Instruct it to get to a seed sentence that Pangram doesn’t flag (or at least flags as AI-assisted rather than fully AI-generated). Then the agent knows what strategies work and can iterate from there
I tried exactly this back in February using Claude Code. It worked pretty easily, though ~all of the important ideas had to come from me and CC itself couldn't figure them out.*
I tried it again a week ago. Complete failure, despite Claude being much better and not needing as much guidance from me. Pangram seems to have been dramatically hardened in the last few months. The API kept returning scores > 0.99 for AI writing so there wasn't even a gradient for it to figure out what's working and what isn't.
* Things it couldn't figure out on its own a few months ago, even after a bunch of iterations:
1) Remove well-known tells like em-dashes (!)
2) Do controlled experiments instead of stuffing a bunch of ideas at once
3) Start from known-best text and surgically make edits. Much better than trying to rewrite the whole essay (or even paragraphs or sentences) because the that just produces text from the LLM distribution all over again.
In my more recent experiment, 1) and 2) were obvious to Fable but 3) still wasn't. Pretty surprising.
Anyway, hats off to @pangram. I'll probably give it another shot at some point but I'm more-or-less ready to change my mind as an AI-detection skeptic.
This was the LLM summary from my first experiment BTW. Still gives me a chuckle.
Daniel W. Linna Jr. retweeted
Artificial intelligence will transform more jobs than it eliminates outright. The biggest productivity gains will come not from incorporating AI into existing workflows but from redesigning those workflows entirely. @MITshapingwork mitsloan.mit.edu/ideas-made-…
Daniel W. Linna Jr. retweeted
A few years ago, “legal engineer” would have gotten blank stares. Now it’s in @nytimes.
Why? AI made it clear that the hard part isn’t buying technology. It’s redesigning the work around it. nytimes.com/2026/10/31/busin…
Daniel W. Linna Jr. retweeted
The CSLaw Workshop is starting up for 2026-2027. This month, on Friday, Oct 30, 1-3pm Central we will explore parallels in AI use in academic peer review and in federal courts, and aiming to find connections between these two research communities. cslawworkshop.org
Daniel W. Linna Jr. retweeted
This is an important perspective that in my view is much more useful than running taste tests between human and AI reviews.
Human-only review is over. It's all about designing the hybrids correctly.
Daniel W. Linna Jr. retweeted
New paper: Rather than automating review, can we target AI to known weaknesses of human review? Consider lottery variance, where differnet reviewer draws change outcomes.
We show how LLM-implemented checklists learned from prior review can reduce variance w/limited added bias 🧵
Daniel W. Linna Jr. retweeted
What would the writing of superintelligent AI sound like? You might expect it to have the characteristics of good human writing — figurative language, high information density, coining new terms, and punchiness — but dialed up to 11. Well, that’s what Claude did, until Anthropic dialed it back recently because of user complaints. It didn’t sound superintelligent to most of us. Instead, the neuralese writing style sounds like a “stupid person’s idea of a smart person”.
Our response to information density and metaphor-laden writing looks something like the cartoon graph I've drawn. Beyond a point, it completely backfires. The reason human writers don’t overdo it isn’t because they aren’t capable, but because they are aware of the reader’s cognitive burden. One of the hardest skills to learn as a writer is to simulate the reader’s experience.
It’s easy to train AI to mimic mediocre human writing. But optimizing it to be (perceived as) good is really hard, and made much harder by the fact that when everyone uses the same few AI models we get tired of their tics very quickly. Most Claudeisms are individually pretty good. I hate that I can’t use “load bearing” in my writing any more. If it weren’t for AI overuse it would be a neat metaphor.
A broader question is whether most tasks in the economy are like chess, where there is a clear performance metric that can be optimized, or whether they’re like writing, where quality is a multidimensional balancing act without clear training signals beyond human mimicry.
Daniel W. Linna Jr. retweeted
Lots of debate this week on whether journals should ban AI in peer review.
We actually tested this in our new preprint, and the results were pretty clear to me.
For context, I've reviewed for most of the big journals and evaluated 1,000 human reviews as an editor. I love to read and evaluate science.
TLDR: From now on I won't do any peer review without AI. For clinical trials, I'd even call human-only review dangerous 🧵
Daniel W. Linna Jr. retweeted
Legaltech friends: VC money is NOT the only way to get rich. But, when it works … oh boy!
Many of my friends in Legaltech made $5-$30m in exits involving very little (but not $0) outside capital. They then go on to run other companies, and make more. Pretty nice life if you ask me.
If you want to be a billionaire (see my pinned post) you gotta take massive outside capital.
If you raise big and don’t exit, which happens, it can be pretty painful.
Daniel W. Linna Jr. retweeted
For anyone interested in the regulation of #clinicalAI, I'm hosting a webinar next week w/ reps from the Federation of State Medical Boards and the AMA to hear how they think about clinical AI regulation and surrounding policy. Registration link here: wsgr.com/en/events/clinical-…
Daniel W. Linna Jr. retweeted
Yesterday, @thehoffprof asked how we can manage law review article selection in times of the slopalypse
Today, a new study offers some evidence for fighting AI flooding with AI. In head-to-head comparisons, AI-generated reviews beat reviews by paid human experts. AI judges overwhelmingly preferred them. Three former journal editors also preferred them in 26 of 30 judgments.
That doesn't mean that AI should select our future scholarship. The downsides are obvious. The study doesn't really test selection decisions, much less questions of scholarly impact. But it makes the case for using AI-assisted triage to direct scarce human attention toward interesting, rigorous, and creative articles
Complain about it, by all means. But the flood is coming either way. If you don't have a better AND realistic solution, the complaint won't get us very far.
Daniel W. Linna Jr. retweeted
By every measure, Opus 4.8 beats human experts as peer reviewers in a study of 124 Harvard social science papers covering a range from statistical analyses to formal models to qualitative work.
As academics, we need to figure out how to adapt the practice of science to AI, fast.
Law folks: If you’re not using Claude Code or ChatGPT’s Codex, you have no idea what you’re missing. So I’m distributing my research tools that provide a huge boost to legal scholarship. Research assistance, cite-checking, proofreading, and more (thread):
danepps.github.io/legal-rese…
Daniel W. Linna Jr. retweeted
🚨CFP: Computer Science & Law 27 🚨
Tomorrow is the last day to submit to CS&Law 2027! We welcome papers at the intersection of these two fields and have both an archival and non-archival (for previously published and working papers)!
computersciencelaw.org
"Frontier AI is itself a firewall" and "slowing its development can postpone new defenses" and "cost protection as well as progress."
Excellent @WSJ op-ed by my @NorthwesternLaw colleague, @joldmcginn.
The safest course on AI is not necessarily the slowest. More capable models may be our best defense against malicious uses of AI. A slowdown can cost us protection as well as progress, as I write in the WSJ today: wsj.com/opinion/the-best-def…






































