About Lawnova

Two people build this.
Both of them have to live with the answer.

One is a Colorado trial attorney who has spent thirty years in front of judges and teaches professional responsibility to law students. The other spends his days proving, to a court's satisfaction, what a file is and whether anyone altered it. Neither of them can afford software that sounds confident and cannot show its work.

That is the whole design brief. Every answer Lawnova gives points back at the document, page and line it came from, because both of us have watched what happens to people who sign something they could not check.

Everything here is really about time — and about who has it

Every decision in this platform is measured the same way: how long does it take to react. Not how impressive the feature is. How fast the answer arrives, and how fast it can be confirmed. A calendar you can trust, a file you can ask a question of, a note you can dictate on the courthouse steps — each of those is time handed back to the person doing the work.

And time is exactly what is unevenly distributed in a courtroom. A prosecutor’s office has investigators, analysts, laboratories and a budget. On the other side there is often one attorney, a stack of discovery, and a hearing on Thursday. That asymmetry is not a complaint; it is the working condition of criminal defence. Technology is one of the few things that can be pointed directly at it.

So the standard we hold ourselves to is not that the software is clever. It is that the lawyer with fewer resources arrives at the hearing having read everything the other side read. We have written about that at length in The Defender’s Duty.

Daniel J. Schendzielos, J.D., Ph.D.
Legal Co-Developer · legal practice & professional responsibility

Daniel J. SchendzielosJ.D., Ph.D.

Ph.D., Kennedy University, May 28, 2026 (aviation law, artificial intelligence & autonomous systems) · Colorado trial attorney, 30+ years · Adjunct Professor of Law, University of Denver Sturm College of Law, since 1999 · Admitted through the U.S. Supreme Court · Pro Bono Certificate, Supreme Court of Colorado · Airline Transport Pilot, 10,000+ hours

Daniel J. Schendzielos, J.D., Ph.D., is an American attorney, author, legal educator, and aviator whose work addresses artificial-intelligence governance, professional responsibility, aviation, and theology. He earned his Ph.D. from Kennedy University on May 28, 2026, with a specialization in aviation law, artificial intelligence and autonomous systems. His books include Delegated Cognition—Second Edition Revised, The Gift of Intelligence, The Synthetic Neighbor, and The Christian AI Trilogy. His writing develops Delegated Cognition, Retained Command, and Retained Presence as frameworks for preserving human judgment and responsibility in the use of artificial intelligence.

Daniel J. Schendzielos, Ph.D., is the creator of the Delegated Cognition/Cognitive Delegation doctrine developed in his books—a framework whose central principle is that cognitive work may be delegated to artificial intelligence, while human command, judgment, and responsibility must remain retained.

The doctrine is set out in Delegated Cognition—Second Edition Revised (Lulu, ISBN 978-0-557-92046-4) · ISNI 0000 0005 3093 2973

Daniel took his J.D. at the University of Denver in 1991 and has been trying cases in Colorado ever since, as managing partner of Trial Lawyers & Legal Services of Colorado. More than a thousand civil and criminal matters have crossed his desk. He has taught at his own law school since 1999, which means he has spent a quarter of a century explaining to people who are about to become lawyers exactly where the duty of care begins and where it cannot be delegated.

The second half of his career is unusual and it is the reason this software exists in the shape it does. He flies for a living as well — an Airline Transport Pilot and flight instructor with more than ten thousand hours — and his doctorate examines what aviation learned about autonomy long before the legal profession had to. Autopilots have flown aircraft for decades. Nobody has ever suggested the autopilot is the pilot in command.

That is the argument of his doctoral work, Delegated Cognition — Legal Authority, Responsibility, and Command in the Age of Autonomous Systems: a machine can carry an enormous amount of the work and still carry none of the responsibility. The duty stays where it was. Which is why nothing in Lawnova decides anything.

He also holds a Pro Bono Certificate from the Supreme Court of Colorado, which is not a decoration — it records hours given to people who could not otherwise have been represented. That is worth stating on a page about software, because it is the same concern from the other direction: the question of who can afford to be defended properly is the question this platform was built around.

“An autonomous system can assist judgment. It does not dissolve command. The professional stays accountable, and the tool has to make that possible rather than harder.”

  • Trial practice & advocacy
  • Professional responsibility
  • AI governance & autonomous systems
  • Aviation law & evidence
Deniss Gladkihs, software architect and digital forensics examiner
Architecture & digital forensics

Deniss Gladkihs

Architect and developer of the Lawnova suite · 25+ years in digital forensics, data recovery and systems engineering · B.S. Computer Science & Radio Electronics, taken at an aviation university · Founder of Denver Data Recovery, eboxlab and Ablaka

Deniss has spent twenty-five years recovering data that other people had written off — from failed, damaged, encrypted and deliberately deleted media, across drives, SSDs, RAID and NAS arrays, servers and phones. He has given forensic opinions in state, federal, corporate, county and law-enforcement matters, and he has done disaster recovery on hardware pulled out of the Waldo Canyon and Black Forest fires and Hurricane Sandy.

Recovery is only half of it. The rest of those years went into building and optimising the systems themselves — writing software, designing the logic layers that decide how information moves through an organisation, and then supporting and advising the people who had to run it. That work has been done for healthcare providers, financial institutions and banks, and government clients, all of them environments where a wrong answer is expensive and an unexplained answer is unacceptable.

That work is unglamorous and it is entirely about provenance. Write-blocked imaging. SHA-256 hashes taken at collection. Chain of custody. Metadata and timeline analysis. The whole discipline exists to answer one question a court will always ask: how do you know this is what you say it is?

He built every product in this suite — the practice hub and its BlueShark assistant, VerdictPilot for trial modelling, COAPP for appellate briefs, and PDF.LEGAL for transcripts — and he built them the way a forensic examiner builds anything: the answer is worthless unless you can trace it. Generative models can now fabricate a document, a photograph, a voice and the metadata around all three. The methods that tell real from manufactured are the ones he has been using for two decades, and they are now the most valuable thing in the room.

He works in three languages, which is less a biographical footnote than it sounds. A case file is not reliably in English: discovery arrives in other languages, witnesses testify through interpreters, and a deposition can turn on how a phrase was rendered rather than on what the witness said. Someone who moves between languages notices when a translation is quietly doing work the record cannot support — and notices, too, that legal systems do not carve up the same problem the same way.

There is a coincidence in this pairing that turned out not to be one. Deniss took his degree at an aviation university; Daniel flies and wrote his doctorate on what aviation law worked out about autonomy. Two people arrived at the same discipline from opposite ends — one from the engineering of a system you are not allowed to blindly trust, one from the law of who remains in command of it — and that is the argument this software is built on.

In September 2026 he published Debugging Justice, a 236-page book on what happens when law, human error, institutional design and artificial intelligence meet inside the systems that judge people’s lives. It asks how far an institution can trust human memory, what follows when legal language stops being readable by the people it governs, and what a court should do when the evidence in front of it is a risk score, a deepfake or a machine-generated analysis. The argument is not that law should be automated. It is that a system built and run by imperfect people should be auditable — preserve the evidence, expose the assumptions, allow the challenge, and treat a correction as a strength rather than an admission.

He followed it with Chain of Truth, a 300-page professional and scholarly guide that follows digital evidence from event and capture to courtroom presentation, human perception and decision — and asks whether the entire path from source to conclusion can be understood, tested, reproduced and challenged. It is the method behind the evidence-integrity teaching at Lawnova Academy.

“With provenance, not vibes. Imaging, hashing, chain of custody and metadata are what let a court trust a file — and they are exactly what an AI answer needs before a lawyer should trust it either.”

  • Legal-AI architecture & retrieval
  • Forensic acquisition & SHA-256 verification
  • Metadata, provenance & timelines
  • Detecting manufactured evidence

Author of Chain of Truth — Digital Evidence and the Future of Proof and Debugging Justice — Law, AI, and the Public’s Right to Understand the System That Judges Them.

How it got here

Six years from a question to a working platform

It was built in the order the problems actually bite: the calendar first, then the file, then the drafting, then one place to ask a question of all of it.

  1. 2020

    The question

    Everything a firm ran on lived in somebody else’s cloud, and AI was not yet in the room. The question was narrower than the one everyone asks now: what would this look like inside the firm’s own office, with the client’s file never leaving the building?

  2. 2021

    Scheduling first — not AI

    A practice runs on dates arriving from dockets, orders, opposing counsel and the client. It needs one dependable place to answer what is due, and when.

  3. 2021 – 2022

    Consolidating the client’s file

    The facts arrive from e-discovery, the court, the client and the other side — and almost all of it passes through email, the largest aggregation database a firm owns and the one nobody treats as one.

  4. 2022

    Drafting and analysis, on one condition

    Inside the practice, without handing the client’s file to a third party. Most of this category asks you to do the opposite; the constraint shaped everything built afterwards.

  5. 2023

    A central buffer

    One place holding the matter and one plain-language question instead of a search — closer to an operating system for a practice than to another application.

  6. 2023 – 2024

    Transcripts

    A deposition runs to hundreds of pages and the four lines that decide a motion are somewhere inside it. Indexing them so they can be found — and checked — became its own body of work.

  7. 2023 – 2024

    Voice

    A note dictated walking out of a hearing exists. The same note typed up on Friday usually does not.

  8. 2024

    Three pilots — New York, Texas, California

    Groups of lawyers ran the system against their own caseloads and sent back the first structured feedback.

  9. 2024

    What it is not allowed to do

    Most of the platform’s refusals were decided in those conversations rather than in the code.

  10. May 2025

    First published, as Lawnowa

    The same system running for the first time outside the workshop, before the name settled into Lawnova.

  11. Aug 2025

    Lawnova.pro opens to its first real users

    Matters, clients, documents, time and billing in one place, with the assistant answering from the case file rather than from general knowledge.

  12. Sep 2025

    First matters run end to end

    Intake through resolution inside the platform — the point at which it stopped being a prototype.

  13. Oct 2025

    Transcripts ship

    PDF.LEGAL indexes depositions to the page and line, so an answer can be confirmed in seconds.

  14. Oct 2025 – now

    VerdictPilot — the largest single piece of work

    Juror personas built from a great deal of data, an argument run against them, and a forecast of where a case will be attacked. It models likely response for preparation — not a prediction of what twelve people will do. Still in development.

  15. Dec 2025 – Jan 2026

    Appellate

    COAPP builds opening briefs for the Colorado Court of Appeals with formatting and word limits applied as the document is assembled, every section attributed to the record.

  16. Apr 2026

    Writing it down

    Long-form work on the questions practitioners actually face — certification orders, disclosure language, verification method — published rather than kept as sales material.

  17. 2026

    Teaching it

    Lawnova Academy takes the same material into supervised cohorts, each ending in a live oral defence.

  18. In progress

    Free tools for people representing themselves

    Citizens who cannot pay for a defence should not face the system with worse tools than everyone else in the room. Being built; not finished.

  19. Now

    Into the pocket

    The platform already installs to a phone’s home screen. Native iPhone and Android applications are next.

Under the hood

The technical part, without the vocabulary

Four decisions do most of the work. They are worth understanding before buying anything in this category, ours included.

It answers from your documents, not from memory

The usual failure of a language model is that it answers from what it absorbed in training, which is where invented citations come from. The alternative — the industry calls it retrieval-augmented generation, or RAG — is duller and safer: find the passages in this matter that bear on the question, hand the model only those, and require it to answer from them and name which one it used. If the file does not contain the answer, the honest output is that it does not.

The model is adapted to the work, not used raw

A general model has read the internet and knows nothing about how your jurisdiction formats a certificate of compliance. Fine-tuning and adaptation means teaching it the shapes of the work — the structure of a brief, what belongs in a statement of facts, how a time entry should read — so that its first draft is in the right form and a human is correcting substance rather than layout.

It can run where the file already is

The platform is built to run on infrastructure you control, with local models and no outbound calls, because for some matters — protective orders, criminal defence, anything privileged — the only acceptable answer is that the material never leaves the building. That constraint was set in 2022 and everything since has had to live with it.

Case law, offline — in progress

The work under way now is the largest of the lot: a case-law corpus that lives on your own hardware and can be searched and cited without a per-seat subscription to an outside research service. For a small practice, research access is a running cost that scales with headcount and does not scale with the size of the firm’s wins. This is not finished, and we would rather say so than list it as a feature.

What that means when you are choosing software

It answers from your file

The assistant reads the matter in front of it. When it tells you something, it names the document it read.

You can check it in seconds

Document, page and line. If confirming an answer takes as long as finding it yourself, the tool has moved the work rather than reduced it.

It never decides

No deadline is calculated for you, no strike is recommended to you. The duty of care does not transfer to a machine, and the software is built on that assumption.

Start with Lawnova Read what we have written →