A four-week AI residency for lawyers.
You’re matched 1:1 with a mentor and build alongside a cohort of lawyers like you.
The outcome
Two weeks changed how they work. Here’s what it felt like, in their own words.
Four weeks. A ramp-up week, then three weeks of building. You’re matched with one mentor who already does this work, plus a cohort of peers, live teaching, and a platform you keep. You leave with something real: a workflow, tool, or system that proves the pattern has taken hold.
Cohort 1 did this in two weeks; Cohort 2 gives you four. A selection of what they built. Every one started where you’re sitting now.
Joey Strength 🇺🇸
Partner, Real Estate
Figured he'd have to hire an engineer. He hadn't used a MacBook in 25 years and was brand new to Claude Code and the terminal.
Now he ships his own software. His line went from “I need to find an engineer” to “I've got 40.”
“I could not do this three weeks ago.”
Title work that took ~3 hours now runs in ~20 minutes (his estimate).
Gillie Abbotts-Jones 🇬🇧
Senior Legal Director, AI & Data Privacy
Had never opened a terminal.
Two weeks later she'd built a working DSAR detection engine herself, from a standing start, proven on synthetic data.
“I feel like I'm down a highway on the moon. Just endless.”
On a blind 500-doc synthetic test, her deterministic core hit 98.4% precision and 83% recall, beating the AI-adjudicated path. Never run on real personal data.
Kenny Tam 🇭🇰
Senior Associate, Technology
Rated himself 4–5/10 and kept apologising about tech on day one.
Now he directs an AI agent end-to-end through a full build. The apologies have stopped.
“I think I'm learning a lot more than I did in the past year of law.”
Demoed end-to-end on synthetic fixtures; by the end he was directing his agent by remote control “from the couch.”
Alexis Werth 🇺🇸
Senior Corporate Counsel
By her own account, she'd always struggled to close out a first version.
She shipped a framework the whole cohort can build their own “second brain” from, and found her feet as a builder.
“It has fundamentally changed who I am as a person. And it's been 14 days.”
A local fail-closed privacy gate that already caught one real would-be leak. The framework ships structure only, never her data.
Peter Scripps 🇺🇸
Senior Privacy Counsel
Had only ever built inside a chat window.
Now he deploys his own apps. He built and deployed Arcus, a regulatory-enforcement signals tracker, on Firebase.
“I just built a front end I really like, in two days. That's insane.”
Built entirely on public agency enforcement records.
Sean Galvin 🇺🇸
Senior Associate, Litigation
Already had an AI workflow for drafting discovery objections, but it ran in a chat window, and he didn't yet think of himself as a builder.
Two weeks later it's a structured app with real guardrails — the AI only picks from approved objections, never invents one — and it's ready to serve as the backend for a Teams bot a whole team can use.
“My CLAUDE.md was 20,000 lines. Claude Code literally told me to cut it back. The fix was about 200.”
Demonstrated on synthetic data, under attorney supervision.
Tamur Shah 🇨🇦
Managing Partner
Stuck on the one problem he most needed solved: safely putting AI on privileged files.
Now he runs a privacy firewall he built himself. He can put a top model on confidential matters and trust what comes back.
“As long as you have the right tools, and the mentorship we got here, it helps you understand what we're capable of making.”
Proven end-to-end on a synthetic 555-document corpus; hardened through a mentor security review. Not yet run on real client data.
Repos stay in the cohort; builds run on synthetic or public data. Nothing’s public unless the resident says so.
Want to hear it straight from someone who did it? Ask Jamie for an intro to a Cohort 1 resident.
A mentor’s job is not to lecture. It’s to make a work pattern legible, and to be there for the questions that get harder to ask the further you climb.
Most lawyers bolt AI onto the way they already work. The harder, more valuable move is rebuilding the work around it: your own tools in Claude Code, agents that run while you’re doing something else, processes you’ll actually keep using.
Use · old pattern
Leverage · new pattern
The artefact you leave with is the visible trace of the shift. The shift itself is the product.
$ claude code --skill nda-review reading 28 pages… flagging risks… cross-referencing house standard… ✓ review-2026-05.md
A repeatable Claude Code process for a matter you handle weekly: clause review, redline pass, first-pass diligence, regulation tracker.
┌─ INTERNAL · CLAUSE COMPARE ─┐ │ │ │ ⌘ drop NDA file │ │ ⌘ select house standard │ │ [ COMPARE → ] │ └─────────────────────────────┘
A custom project management dashboard. A web crawler. A Slack bot. An email agent. A contract-review assistant. Internal apps your team can actually use.
$ claude code --agent --hours 7 reading 200 documents… cross-referencing precedent… drafting initial findings… runs unattended overnight
Running an agent autonomously for 5–7 hours. Standing up an MCP endpoint. Configuring your own OpenClaw or NanoClaw. A repeatable move you can demonstrate live.
monitor stock prices ─► detect drop > 15% ─► agent scans SEC + news ─► surface fraud indicators ─► draft Statement of Claim
A new way of working: the practice itself, redesigned. Monitor stock prices for unusual drops, dispatch an agent to scan SEC filings and news for fraud indicators, auto-draft Statements of Claim for potential plaintiffs.
Week 1 · Ramp-up · July 2026
The Mirror
Self-assessment · do this first
Kickoff & intro call
Meet your matched mentor
Environment setup
Terminal, Claude Code, Git/GitHub, Vercel + Railway accounts
Reading list · foundations
Scope your build with your mentor
Weeks 2–4 · Building · July – August 2026
Mentor 1:1s
60 min each
Webinar · Judgment
Picking the right problem
Webinar · Judgment pt 2
Own your data + verify AI output
Weekend
Mentor 1:1s
60 min each
Webinar · Craft pt 1
CLAUDE.md, Skills, Hooks, Plugins
Webinar · Craft pt 2
Pipelines, QA, shipping well
Weekend
Mentor 1:1s
60 min each
Webinar · Career & sustaining
Compounding, permissionless building, burnout
Webinar · Local models
AI on your own hardware
Submit
Repo · demo · writeup
Peer review
Paired in-cohort
Demo Day
10 min each + live Q&A
After
From Aug 3
Alumni community
Slack + platform, playbooks and the LegalQuants community · for life
Your self-assessment, the curated library, the Foundation Tapes, the recorded office hours, and every build the cohort ships. One logged-in home, yours to keep after the residency ends. The day-to-day conversation happens in Slack.
80 lawyers · two hackathons · 38 public builds
Miri Nae
Robert Kuter
Jesselyn Tham
Kong
Jae Kim
Kira Ahveninen Kuha
Claude Bothe
Sergey Kinchin
Adam Masser
Alexios
Ken Priore
Hemma Lomax
Nick Hayes
Ryan Malek
Muhammed Bhayat
Lana Kelly
Nicolas Panigutti
Patrick Munro
Polina Chtchelok
Andrew Cooke
Amit Bhanot
Kiarash Kiani
Nicole Diaz
Kaycee Hulet
Hannah Lim
Alvaro Tejada
Paimon Abedi
Jason Chang
Alex Latham
Julian Kie
David Liu
Jo Sauer
Chloe Chan
Richard Backhaus
Kevin Pratt
Callum Obrien
Jason Knapp
Benjamin Jacobs
Miri Nae
Robert Kuter
Jesselyn Tham
Kong
Jae Kim
Kira Ahveninen Kuha
Claude Bothe
Sergey Kinchin
Adam Masser
Alexios
Ken Priore
Hemma Lomax
Nick Hayes
Ryan Malek
Muhammed Bhayat
Lana Kelly
Nicolas Panigutti
Patrick Munro
Polina Chtchelok
Andrew Cooke
Amit Bhanot
Kiarash Kiani
Nicole Diaz
Kaycee Hulet
Hannah Lim
Alvaro Tejada
Paimon Abedi
Jason Chang
Alex Latham
Julian Kie
David Liu
Jo Sauer
Chloe Chan
Richard Backhaus
Kevin Pratt
Callum Obrien
Jason Knapp
Benjamin Jacobs
Lawyers from AmLaw100, Magic Circle, and Fortune 500 in-house teams. One chat community, hundreds of messages a week, some of the most AI-native lawyers anywhere. When you join the cohort, you’re in. After the residency, your stuck point is usually someone else’s solved problem.
$4,725 $5,250
Early-bird, until 28 June. Standard $5,250 after.
All in. No upsells. Everything above is included: mentor, live teaching, the platform, and the community. Employer reimbursement and corporate sponsorship available, with invoice, course summary, and certificate of completion provided.
Cross as one lawyer
Cohort 2 · July 6 – Aug 2 · 20 seats
$4,725 early-bird until 28 June ($5,250 after). Secure checkout.
Bring your team across
Book a 20-min scope callSponsorship + certificates of completion. We’ll scope it with you.
Yes. That's the most common starting point, not a disqualifier. Several of Cohort 1's strongest builds came from lawyers who, two weeks earlier, didn't know what the terminal was (one bought a second-hand MacBook to start). The Claude-era stack is built for non-coders, and your mentor calibrates to exactly where you are. The bar is judgement, not engineering.
No. Picking what to build is half the work, and the first week is partly for finding it with your mentor. A good test: the thing you'd use yourself, daily. If you wouldn't, neither would anyone else. You don't need an app, either; a workflow, a document pipeline, a monitoring routine, or just a repeatable capability all count.
Yes, and it's designed around exactly this. You build on your own machine, your own plan, and on de-identified or public material. Cohort 1 used synthetic data, public filings, public title records, and the like. Nothing confidential goes near a model or a public repo. Your mentor never needs to see anything privileged. The judgement transfers even when the data can't leave your laptop.
It's educational, not an audition. LegalQuants isn't a studio or a VC, the residency isn't a tryout for a job, and we claim no IP in what you build. It's yours, on your own terms, and it travels with you. What you keep afterwards is the build, the habit, alumni access to the community, and the platform for life.
A hackathon is a day and a course is a video. This is a mentor beside you, most days, for four weeks: watching your actual screen, scoping with you, and getting you to a shipped thing you can prove. The hackathons are where many people meet us; the residency is where the capability actually takes hold.
No. The goal isn't software engineering. It's scoping, building, and verifying AI-enabled legal work. The point is your judgement encoded in a tool, not lines of code. Your mentor handles the rest.
Two to three hours a day, more if you go deep. The fixed live points are two webinars a week (Tue/Thu) and a 60-minute mentor 1:1 every other day (Mon/Wed/Fri). Everything else is async, on your own time.
Six sessions across the build weeks, taught by the whole mentor bench: Judgment (picking the right problem; owning your data and verifying AI output), Craft (CLAUDE.md, Skills, Hooks, Plugins; then pipelines, QA and shipping well), Career & sustaining, and Local models (running AI on your own hardware). All recorded for other timezones.
Cohort 1 ran across timezones from the US West Coast to New Zealand; some calls landed at 1am. Webinars run at a fixed slot and are all recorded, so you never lose the content, and your mentor works asynchronously through your private channel, so you're never blocked waiting.
Most lawyers we see have publication restrictions, so nothing is public by default. Your build stays internal, the Demo Day recording is cohort-gated, and peer review stays inside the cohort. You decide what, if anything, ever leaves the room.
Cohort 2 is $5,250 all in, with an early-bird rate of $4,725 until 28 June 2026. After that it's $5,250.
Yes. We provide an invoice, a course summary, and a certificate of completion for corporate sponsorship or reimbursement. For a whole team, use the corporate door above to scope a private cohort.
Two paths. You cancel: 50% refund up to 29 June 2026; after that, no refund except force majeure. Or we review your enrolment within 48 hours of payment and decide it isn't the right fit: full refund, no reason required.
Early-bird ends June 28. Starts July 6.
