One professor's life work — consolidated into a living practicum where a human + AI apprenticeship teaches lawyers to be the finest advocates they can be.
The work of John O. Sonsteng · with Roger S. Haydock · open & MIT-licensed
For fifty years, John Sonsteng has documented a stubborn gap: graduates emerge fluent in doctrine yet unprepared to practice — to interview a client, take a deposition, negotiate, try a case. He didn't merely assert it; he measured it. A 53-question instrument went to 19,077 Minnesota attorneys; 1,398 responded, scoring each skill for importance against how prepared law school left them. The verdict was consistent, and unflattering — the evidence is below.
The remedy has always been known — repetition, performance, and immediate expert critique (the apprenticeship the trades never abandoned). It simply never scaled: one master can only watch so many reps. That is the constraint AI just removed.
It does not train students to be lawyers.Sonsteng, Ward, Bruce & Petersen · A Legal Education Renaissance (2007)
Sonsteng named the work of a practicing lawyer precisely: 17 legal-practice skills and 9 law-practice-management skills. For each, the survey asked the same four questions — is it important, were you prepared, could it be learned in school, and where did you actually learn it. Every skill was rated important. Most could be taught in school. Yet law-school curriculum was the leading source of training for only 7 of the 17 practice skills — and for none of the 9 management skills.
Importance vs. preparedness — the seven widest gaps
The 17 legal-practice skills — the full instrument
A law practice is a business — yet law school is credited as the leading trainer for none of the 9 management skills. Where curriculum shows up at all, it is in the low single digits: billing 7.9%, marketing 3.9%, capitalization 2.3%.
All 26 skills matter. School leads on seven.The 2018–19 Minnesota study · one of a longitudinal series (1997–99 · 2013 · 2018–19)
Sonsteng's corpus isn't a pile of articles — it's a single sustained case, built across three decades. The Magnum Opus simply finishes the sentence.
The diagnosis. A practical, client-centered model for 21st-century legal education — the manifesto the profession cited but never fully built.
The solution, proven in a live classroom: his General Practice Skills Practicum — a working-law-firm simulation with real reps, memos, and graded advocacy.
Open-access exercises & case files (Trialbook; the "Midstate" jurisdiction), run jointly by Mitchell Hamline's C-LAB and IGUL at Bahçeşehir University, Istanbul.
The team — and the first centaur — already exist.Trialbook, by Sonsteng, Haydock & Riehl — a text and AI platform (Trialbook GPT · Opal)
Trialbook and Lawyering Practice and Planning already bind Roger Haydock's advocacy and transactional canon to Sonsteng's pedagogy — and Trialbook ships working AI advocacy companions. The Magnum Opus doesn't start from zero; it consolidates a life's scholarship and systematizes what they began into a full practicum.
Sonsteng's cost model is the disarming part: convert half the tenure-track lines to long-term teaching faculty and a school can hire ~two teachers for every line replaced — tripling teaching time at the same or lower cost. AI first-pass grading (Act V) then pushes capacity well past even that.
Not one more article. A layered whole — the ideas, the course, and the machine that delivers it, bound together and given away.
A canonical treatise consolidating Sonsteng's scholarship — including restoration of the crown-jewel articles now surviving only as degraded scans. The intellectual core, made permanent.
The Skills Practicum, structured into teachable modules — skills, exercises, rubrics, and the Midstate case files — so any instructor, anywhere, can run it.
AI-assisted delivery that finally scales the 1:1 apprenticeship — unlimited reps, instant critique, human judgment where it counts. The living opus.
The Book, structured as six
The treatise is an assembly of proven, peer-reviewed pieces — not a new draft from zero. Each book maps to existing Sonsteng scholarship, carrying the recurring "THE PROOF" device from front matter to back.
The argument & the empirical proof — the Minnesota studies, the 17+9 skills, the preparedness cliff.
The pedagogy — four pillars, six learning elements, the three-module curriculum, the cost model.
The course, running live — the working-firm model, credit tiers, rubrics, reflective learning.
The simulations — the six matters plus the extensible bank, and the Practicum Series engine.
The platform's human layer — the open-resource model, the business-of-law thesis, C-LAB.
The platform's AI layer — first-pass critique, infinite feedback, the economics of scale.
Students don't study lawyering — they do it. The practicum runs as a simulated firm handling real matters on the Midstate case files: they draft, argue, question, counsel, and negotiate, then get critiqued and do it again.
The exercise library already spans the advocate's full range:
Assessed against Sonsteng's skills framework
Every rep maps to a defined competency and a rubric — so progress is measured, not impressionistic. Mastery is visible: the student, the coach, and the record all see the same growth curve, skill by skill.
The result is a graduate who has already been a lawyer for a semester before the bar exam — with the reps, the feedback, and the receipts to prove it.
Every course runs a three-step cycle
Independent pre-work — texts, casebooks, prepared materials. Competence is checked with short papers, essays, or an oral exercise before the room convenes.
The core: face-to-face, small-group, faculty-guided real-world simulation. The reps that only happen live — argument, negotiation, trial.
A comprehensive applied assessment integrating the prep and the practicum — memos, problem discussion, the redlined re-write against the rubric.
The matters — six complete case files
Eight representation agreements, eight client billing statements fed by weekly time sheets, briefs, pleadings, and thirteen time sheets all carry points. The seam the Centaur plugs into: any piece scored below 6 may be re-written within a week against the same rubric — the profession's rubric-and-repeat loop, waiting only for feedback to become instant and infinite.
The apprenticeship works because of expert critique. AI doesn't replace the expert — it multiplies the reps and handles the first pass, so the master's scarce attention lands where only judgment will do.
The student performs — a cross, a memo, a negotiation, an argument — as many times as they want, on demand.
StudentAI grades against the rubric in seconds: what worked, what didn't, what to try next. Unlimited, 24/7, judgment-free reps.
AI · scalesHuman faculty spend their scarce hours on what machines can't teach: judgment, ethics, strategy, presence — the "why."
Faculty · anchorsIllustrative — a single rep
This is the differentiator. Peers build AI judges or AI witnesses; we bind a named scholar's life work, an open curriculum, and a faculty-anchored centaur into one system — in the NITA tradition of learning by doing, finally at scale. And we've already shipped generation one: Trialbook's AI companions prove the model works.
The survey named the whole job — 17 practice skills, 9 management skills — and showed school leading on only seven. The Practicum is designed backward from that list: every one of the 26 skills has a home in an exercise, a stage in the teaching cycle, and a human-anchored AI amplification. Nothing is assumed; each skill is graded work.
Sonsteng's method below is drawn from his articles and the live Skills Practicum. The improvements and the AI amplification are our proposed layer over that proven method — never a replacement for it.
The 17 legal-practice skills — where the curriculum puts each one
The 9 management skills — the blackout school leads on none of — live in cluster F below: the working-firm overlay of agreements, time sheets, and billing.
The method — Sonsteng's approach, our improvement, the AI at each stage
Keeps the doctrinal rigor school does well — legal reasoning, procedural law, research — but tethers every analysis task to a live matter file. The four-page case-analysis memo forces governing law, both sides' strengths, the elements to prevail, and a recommended path, checked by a short paper or oral exercise before the room convenes (FSS-I).
Make doctrine adaptive — re-drill the elements a student keeps missing — and route every cited authority through a verification gate, so a confident-but-wrong citation can't reach the client's memo.
An always-on reasoning spar that argues the other side of any rule, plus instant rubric critique that flags a mis-stated standard of review in seconds — the FSS-I competence check, available at 2 a.m.
The Midstate files hand students a mess — witness statements, exhibits, discovery — not a tidy hypothetical, and grade diagnosis and fact-gathering as first-class work. Midstate v. Rogers turns on developing one witness per side; Reagan's stipulated diagram is the load-bearing exhibit.
Make the record interrogable — witnesses and documents a student can actually question — so fact-gathering becomes an act, not a reading, and gaps surface as costed risks the way they do in practice.
AI witnesses and clients answer follow-ups in character, and an AI "opposing theory" pressure-tests the case so weak facts are found in the practicum, not at trial.
Roughly 1,438 graded points ride on written product — 7 four-page memos, pleadings, briefs, agreements — under strict page limits that force concision. Anything scored below 6 is re-written within a week against the same rubric, submitted as a redline. Writing is taught as revision, not inspiration.
Keep rubric-and-repeat but collapse its latency: the re-write no longer waits a week for a human read, and feedback lands at the clause level — specific enough to act on immediately.
Instant, rubric-anchored redlines on every draft — unlimited, judgment-free reps — while faculty hours move to the passages where judgment, not form, is the issue.
The IRP core, in trial garb: arbitration hearings, PR argument before a judge and two OLPR panelists, DWI bench trials before Judge Olmstead, and the capstone — a full-day jury trial in Reagan v. Jacobson, pleadings through verdict, with volunteer jurors deliberating.
Give unlimited private reps before the graded live one, so scarce, high-stakes courtroom hours are spent on polish and presence, not first attempts.
AI witnesses to cross, AI opposing counsel to answer, and an AI judge panel that rules on the identical record and returns a distribution — the same adversarial engine already running on real matters.
Peters, Taylor & Thomas is a three-way negotiation with confidential facts distributed per side; State v. James runs plea talks into a hearing; SSHC ends in a settlement conference. The graded strategic settlement & negotiation plan (SSNP) makes the strategy explicit.
Let a student run the same negotiation ten times against ten temperaments, then debrief which moves travel and which were luck — turning one rep into a distribution.
An AI counterpart or client that holds any posture on command — hostile, anxious, sophisticated — so counseling and bargaining are rehearsed against variety no single classroom can stage.
The course runs as a working firm — two-person partnerships with their own letterhead, representation agreements, and 13 weekly time sheets feeding client billing statements. These are the business skills school credits for none of the nine, taught here because a practice is a business.
Surface the economics the way a partner sees them — every task carries a notional cost and a client-facing bill — so students feel the cost of over-lawyering before a real client pays for it.
A simulated intake, pitch, and billing cycle with an AI client whose confidence tracks competence and candor — the market feedback a classroom can't otherwise give.
In re Halbrock argues an attorney-discipline matter before a judge and two OLPR panelists — professional responsibility as advocacy under questioning, not a multiple-choice afterthought — turning a skill school already leads into a live oral crucible.
Wire the ethical check into every exercise, not one dedicated matter — a fabrication or a confidentiality slip is caught wherever it happens, mirroring a fail-closed guardrail.
A dilemma simulator that escalates the hard facts, plus a coach who reserves the human hour for the "why" — the judgment, ethics, and presence machines can flag but can't teach.
School leads on seven. The Practicum — human-anchored, AI-amplified — reaches all twenty-six.Sonsteng's method, faithfully extended · the improvements & AI are the proposed layer
The opus is open source and MIT-licensed, extending the open-access spirit of the Resource Tool. A movement to remake how advocates are trained can't be locked in a paywall — it has to be free to copy, adapt to any jurisdiction, and improve. That's how a life's work becomes a standard.
Open also compounds: every school that adopts it adds exercises, critiques, and data back to the commons — the platform gets smarter as the field uses it.
Confirm direction with John & Roger.
Restore one crown-jewel article + build one centaur exercise end-to-end.
Run it in a live Skills Practicum cohort; measure against the rubric.
Open-release the curriculum + platform for any school to adopt.
Two ways to give feedback as we walk through: tap a ✎ pencil beside any section or card to leave a comment tied to that spot, and/or set an overall reaction per pillar below. Everything gathers in ✎ Comments (bottom-right) → Copy all for Damien to paste back into the chat.