Practical legal judgment, and the habit of fixing the process.

I've spent my career on commercial agreements and the disputes that come out of them. I explain what terms mean for the work, negotiate them, and build the documents and processes that make the next deal easier.

Zach Stern smiling at the camera, in glasses and a black winter jacket, in the stands at a football game with the field and the crowd behind him. — Photo Soldier Field, Chicago

In-house: Dovenmuehle, February 2017 to October 2026
Law firm: Tishler & Wald, May 2007 to February 2016
Independent: ZMS Labs projects, built with AI tools

Experience

In-house · Feb 2017 to Oct 2026

Dovenmuehle

Corporate Counsel

Staff Attorney until January 2020

I was the primary lawyer for purchasing-side commercial and technology agreements, from review and negotiation through signature, amendments and termination. I drafted, explained the practical consequences of the terms, and helped colleagues understand the issues and their options.

Master services agreements, statements of work, confidentiality and data-protection terms, and riders.
Created the company's contract rider and a negotiation reference that I used actively.
Developed the NDA documents and the intake, review and signature process.
Made access to vendors' assurance reports a repeatable process.

The agreements work

Law firm · May 2007 to Feb 2016

Tishler & Wald

Attorney

Law clerk from May 2007 to November 2008

I worked on commercial financing and on contested claims. On the financing side: credit and security agreements, notes, guarantees, mortgages, assignments and intercreditor arrangements, plus closing the transactions and advising on rights and enforcement. On the contested side: commercial litigation, arbitration and creditor bankruptcy work.

Substantive trial drafting and factual investigation.
Keeping changing information organized across parallel proceedings.
Co-designed and ran a chancery claims process for a creditor client, and drafted the central trial-court motion against a $17,487,520.30 competing claim.

The claims case

ZMS Labs · How I work with AI tools

Some people let the wall of text fly by. I read it.

AI tools write the code. I decide what each project is for and check what comes back.

ZMS Labs is the name I use for my independent projects, separate from my legal work. I use AI tools for research, design, testing and writing too, including first drafts of these pages. I direct the work and review it with a critical eye.

In SaveBench, an AI model earned the game's milestones by feeding machines by hand and never built a conveyor belt, even after hand transfers started costing game time. At first I called that a flaw in the model. Then I decided the problem was the test: SaveBench measures factory design, not whether an AI chooses to automate. So the test became a production quota the factory has to keep meeting while nobody touches it.

Other people's libraries, tools and research are part of the work too, and each case study credits them. Every project page says what's real, what's sample content and what hasn't been tested yet.

All work Sources and records

How this grew

In January 2026, I was using Claude Code to work on my Home Assistant setup. Soon I was asking whether it could help on my other machines. That expanded into shared work across devices, gaming projects, a constitution for my AI agents, and quite a few revisions to how the whole thing should work.

I enjoy being able to take an idea, work through the possibilities, and get far enough to actually try it. I also want to understand what the assistants are doing and whether their results hold up. Today that means working with cloud and local agents, and building toward Fleet Orchestrator: one place where I can see the team at work, direct what comes next, and enjoy watching it happen.

How my use of AI grew

A few words the projects use.

AI agent: An AI tool that carries out a multi-step task on its own, such as changing code and running tests.
Matter: In Interleaf: one contract with its related documents and open questions.
Hallmark: In Interleaf: the emblem that shows a grade and, separately, what kinds of evidence back it up. It's modeled on the stamp an assayer strikes into metal.
Surface Bridge: In Fleet Orchestrator: the part that turns a decision about an agent's next step into commands for whichever AI tool is running it, and reports back.
Canon: In Neuraxic: the version of a story's facts the author has accepted. Everything else is a proposal.
Tome and loose leaf: In Neuraxic: a tome groups a story world's ideas, prose and claims; a loose leaf is an idea that hasn't found its place yet.
Fingerprint: A short code computed from a file's exact contents (a SHA-256 hash). Change one byte and the fingerprint changes.

Let's talk.

LinkedIn is the best way to reach me.

Message me on LinkedIn