New Brunswick, Canada
We study how AI agents build software, then put that practice to work on yours. Agents do the heavy lifting. Two humans keep the vetoes.
01What we do
Half of the lab explores what AI agents can do. The other half takes on contracts and builds real software with them. Each side keeps the other honest.
Web apps, internal tools, and platforms built to your spec. AI agents draft, test, and refactor. We review every change, and nothing ships without a human saying yes.
We test how agents and models behave on real engineering work, and turn the promising ideas into prototypes you can click.
The process that lives in texts, inboxes, and a spreadsheet that became the company. We turn it into software and agents your team can actually share.
Old or home-grown systems only one person understands. We map them, then move them forward. It is the idea behind CobolGraph, our tool for COBOL.
We ship and operate our own apps, so the methods we sell are tested on our own work first. See the projects.
Describe the messy process. If software is not the right fix, we will tell you before you spend anything.
Describe your process02How the lab works
We are heavy AI users. Agents write, test, and revise around the clock. Two humans stay on staff for strategy, engineering, and vetoes.
You bring the problem. We pin down what done looks like, together.
AI agents write the code, the tests, and the docs, then revise them fast.
A person reviews every change. Anything that would confuse you goes back to the agents.
Approved work goes live, and we keep watching how it behaves.
An example run: an idea is scoped, agents draft code and tests, a human reviews and either vetoes or approves, and approved work ships.
AI is part of how every project gets built, not a feature bolted on at the end. That is what lets a small lab move quickly.
Agents propose. People decide. Strategy, engineering judgment, and the final yes or no stay with the two humans who run the lab.
What we learn in the research lab changes how we build for clients, and client work tells us what to research next.
03Meet the lab
The lab is run by two people who go by their codenames. They set the direction, review the work, and answer your email.
Codename
Brad Logan
Builds and engineering
“Ships the products. Reviews the diffs. Still allowed to refactor after hours.”
Codename
Joseph MacCallum
Strategy and thinking
“Keeps the bets honest. Taste, numbers, and the veto when something would confuse a customer.”
04Our projects
Live apps from the lab. Each one opens in a new tab. They are also how we test our own methods.
Run live vehicle auctions online with anti-snipe protection still on. Built for auction houses that cannot afford a last-second mess.
For auction houses and remarketing teams
Map legacy COBOL into living graphs so the people who own the system can finally see what it does.
For enterprises still running COBOL
Compile what is true in your work so ChatGPT, Claude, and Cursor can continue without being told twice.
For people who already pay for more than one AI
Coordinate farm-to-table pickup in New Brunswick with live pickup codes instead of a trail of texts and sticky notes.
For farms and buyers in New Brunswick
Leases, maintenance, and compliance in one place — so property managers are not running the building from a spreadsheet.
For property managers
Puck for the cell, Clip for the room, Line for the landline — and a ping to someone you trust, so a call that claims to be the bank or the CRA does not become an Interac e-transfer.
For anyone with a phone
Medications, tasks, vitals, and updates for the person you love — one private app instead of a family group chat.
For families and caregivers
Pay two dollars CAD, drop a resume, LinkedIn, or bio, and get a verdict and what to change. Kairen does not keep the file.
For anyone dropping a resume, LinkedIn, or bio
05Work with us
Tell us about a project, a research question, or a process that should be software. A human reads every message and answers.