Building the AI that defines how people live — and how cities take shape.

Global construction is a ~$16-trillion-a-year market: roughly $2.2 trillion in the US alone, about $4 trillion in China, and ¥100 trillion (~$700 billion) in Japan. One of the largest industries on earth — and one of the least digitized.

Harvey in legal, Sierra in customer service — vertical AI agents that take on entire professional workflows are growing explosively worldwide. Building review is the blank space that wave hasn't reached. That's where we chose to begin.

General-purpose AI "knows" building codes — but it can't review. Extracting attributes from drawings with sources, cross-checking them against statutes, and flagging issues with confidence scores and citations takes deeply domain-specific engineering. mign already automates hundreds of review items at over 90% accuracy.

And the system compounds: every customer's data and every piece of reviewer feedback makes the AI more accurate, which drives more usage, which builds more data. That flywheel is the core of our moat.

mign is in its customer-validation phase. Multiple leading companies — global enterprises among them — use our product on paid contracts while we scientifically validate product–market fit. To accelerate, we have raised several hundred million yen from leading venture capital firms.

Join now and you don't ride a finished business — you decide its shape. That discretion, and that responsibility, belong to the core members who join at this stage.

We chose one of the world's strictest and most complex review regimes — Japan — as our first proving ground. Accuracy that holds up against more than 100,000 statutes and local ordinances holds up in any market.

Automating permit review is only the entry point. On the accuracy and trust built here, we expand: to the applicant side — developers and design firms — to review and inspection work beyond permits, and to markets including the US and Asia.

Beyond that, we aim to generalize the very process of creating vertical AI agents and bringing them to the world. We will keep turning the work adjacent to review into specialized AI agents — and then, across industries, transform all kinds of professional work into agents. Our goal is to let anyone easily build 'AI employees' that inherit their company's culture, know-how and expertise, and share them inside and beyond the company — a factory for vertical AI agents. It makes a company's history something that can be stably inherited, accumulated, and passed on to the next generation.

We treat strategy as a testable hypothesis about what to abandon and what to bet on. Decisions run on hypothesis-and-validation loops, not intuition — randomized controlled trials show founders trained in the scientific approach outperform those who aren't. We run management itself as science.

A team beyond borders, built around graduates of top universities and research institutions in each country and holders of doctoral degrees, implementing frontier technology directly into the product. 35% of our technical members hold PhDs, channeling insights honed at the research frontier of AI, machine learning and architecture straight into product development.

We don't just follow the latest research. We publish our own work, primarily at the world's top conferences, and share what we learn with the global research community. Because research and implementation sit this close together, we deliver a level of accuracy and reliability that general-purpose AI cannot reach.