Something big is happening
Heading borrowed from Matt Shumer, somethingbig.ai. His essay in early 2026 took the internet by storm and captured the sense of a tectonic shift in the world of knowledge work. Read his essay here.
You’ve probably felt it already. A colleague mentions they used AI to summarise a record overnight. A client asks whether you “ran it past ChatGPT”. The ground is shifting under our profession, and it’s happening faster than most of us expected. If you feel a little behind, you’re in good company.
We are tiny language models
Lawyers are tiny language models. We spend years reading law reports and textbooks, absorbing patterns and turning them into words and sentences (hopefully persuasive ones). That’s what legal training and legal work do to us. For a long time our edge was that we could read more, remember more and write faster than most. Now there are machines that do all three at a scale we struggle to imagine, let alone match.
That’s (hopefully) not the end of the story for lawyers. It just means that the value we are supposed to add is more exposed and more visible. Judgment, good advice and strategic thinking, empathy, and persuasive writing remain with us.
This site is about finding where AI tools can help us focus on these things and relieve us of the burden of the rest. Without getting us struck off the roll or being mentioned in the law reports for all the wrong reasons.
The AI curve
Back in 2015, Tim Urban of Wait But Why drew a picture that stuck with a lot of people. A little figure stands on a timeline. Behind him, progress looks gentle and flat. Just ahead, the line rockets upward. His point was simple: we judge the future by the past, so we keep underestimating how fast change arrives. Ten years later, it’s hard not to smile at how right he was.
Read the article: The AI Revolution: The Road to Superintelligence (Wait But Why, 22 January 2015).
This site is probably out of date
By the time you read this, some of it will already be old. Hell, by the time I wrote this it was probably outdated. The tools change every few weeks, and what couldn’t be done in January is routine by June. Hopefully some of the principles here survive to December. Think of the site as a friendly starting point rather than a rulebook. Or as a historical record of how AI once worked. Before it ate the world.
Wherever you are on your journey
Maybe you’ve never opened a chatbot and aren’t sure where to begin. There is a bit of learning to do to get from “Hey chat, why is my toenail green?” to producing polished legal work. This site is about sharing what I learnt in the process. Start at the beginning, or jump straight to what you need.
Which AI, and which model?
I have no clue. I am totally unqualified to advise on this. I started using Claude Cowork when it first came out. It was a revelation to non-coders. I’ve been using it ever since. I recently started using ChatGPT Work, and it is incredible. Open source models are apparently amazing. As are Gemini and Grok. I really don’t know which is best.
The principles here should carry across to whichever platform you prefer. Memory and skills can now be moved between platforms, though not always perfectly. How to give clear instructions, how to check the output and where to draw the line on judgment won’t change much from one tool to the next.
The choice is yours. Try a few, see what fits the way you work, and check each one’s privacy terms before you trust it with client material.
What I have benefited from is this: pick a tool and fight with it every day to help solve your actual work problems. Read this PDF, learn my writing style, help with this memo, do it again, try again, iterate, complain, shout at the screen, and so on. Don’t flit about between models; just pick one, give it real work, and apply the Festivus method: complain to it about how much it sucks. Tell it to try to write that letter again. Over and over. Iterate, change your mind, think hard about what you actually want to achieve, and just keep at it. Watch it learn your ways. Over time, you’ll be amazed at what you can achieve.
The blank chat box
Every AI tool greets you the same way: an empty box and a blinking cursor. It looks so simple. It’s also where most people get stuck. If your first few tries felt disappointing, it probably wasn’t you or the tool. It was the blank box.
How to fill in the blanks
The good news: getting better results isn’t hard. Tell the AI who you are, what you need, what material to work from and how you’d like the answer. Give it the documents instead of asking it to remember the law. Then check what comes back. It helps to picture a bright, eager junior on their first day. Quick and willing, but they need clear instructions and someone keeping an eye on their work.
Remember, it isn’t a magic answer machine
Sometimes AI produces results that are so astonishingly good, it becomes tempting to think of it as an oracle. It isn’t. It predicts words that sound right, and most of the time that’s genuinely useful. But now and then it gets things confidently wrong, including inventing cases that don’t exist and quotes no judge ever wrote.
This catches lawyers out. Prior to LLMs, it was almost impossible to accidentally cite a case that didn’t exist. That required deliberate mischief or a genuine copy and paste disaster. And all online legal research tools did was return a binary set of results based on the actual database of cases.
In 2023, two New York lawyers were caught out by this new and sometimes alien technology, and made legal history in the Avianca case: Mata v Avianca, Inc., 678 F Supp 3d 443 (SDNY, 22 June 2023). Wikipedia has a good account of what happened.
Since then, similar disasters have been repeated all over the world. Damien Charlotin, a Paris-based academic and lawyer, maintains the best-known list of reported cases documenting the misuse of AI: the AI Hallucination Cases database. At last count, there were over 2,000 cases on his list, searchable by country and issue type. His website also keeps lists on AI evidence and AI in discovery. It makes for fascinating if somewhat chilling reading.
My intense interest in improving my own legal practice with the help of AI is matched by my interest in not being caught out by it. Hopefully this site will help you do the same.
The best articles on AI in legal practice that I’ve read
Shortly after I started using Claude Cowork, I randomly came across a blog post on Twitter by Zack Shapiro (@zackbshapiro), on how to use Claude Cowork in a small or solo practice. This was transformative to my practice, and I recommend that you read it (and the others that followed it) before you get into the weeds on this site.
Zack Shapiro, managing partner of Rains LLP in New York, is a corporate lawyer who runs a small practice on Claude and writes about how.
Articles on X, 2026 (they open without an X account): The Claude-Native Law Firm (27 February); The Judgment Premium (2 March); The 10x Lawyer (11 March); The Input Layer (25 March). These are all worth reading again and again. If you do nothing else on this site, read these articles.