Gordian: for licensing, and apparently everything else too.

A project manager uploaded his P6 schedule to Gordian. It surfaced a float error traced to an activity that should have completed three years prior, in minutes.

An engineering manager evaluated a change across an entire design document set — two weeks of work done in three days, approved with one bring-back.

Over a weekend, in about an hour, a plant engineer used Gordian to troubleshoot a live plant issue. He researched operating experience, built a fault tree, and planned out what actions to take. Then he could get back to watching the game.

A manufacturer wanted to change quality systems, but didn't want to pay a consultant to start the gap analysis. Gordian did it in minutes. The consultant might not be happy with us.

None of that is what those customers came to us for.

We set out to solve one hard problem: nuclear research takes too long, and nuclear documents are too complex and too interconnected for general-purpose AI. That led to the results we've shared before — a 208-page document completed in one day instead of 200 person-days, an environmental assessment compressed from eight weeks to one.

We built a platform that works for licensing. Then our customers started using it for everything else.

I'm not complaining. But I am paying attention.

What's actually happening

All of these people can get value out of Gordian because of the care we've taken in building scalable technology. Here are just a few concrete examples of capabilities now in production.

Operating plants run hundreds of 10 CFR 50.59 screenings per plant per year, each one taking hours to weeks. The 50.59 workflow we built cuts that research time significantly, produces consistent citations back to source documents, and outputs in the format teams already use. Fast and traceable is what you'd expect these days. Accurate and capable of finding connections the experts wouldn't have thought of — that's where Gordian really shines.

Procedure management is a different kind of hard — not research, but consistency, human factors, and configuration control across thousands of documents. We built a rubric-based review system with over a hundred weighted criteria, validated against the plant's own quality framework. One site scanned thousands of procedures, surfacing suggestions that the writer accepts or rejects. Changes are tracked to accelerate approvals. A template-compliant procedure comes out the other end. Total effort? About 3 hours for an end-to-end revamp, not 25.

Companies developing new designs all want to build on nuclear's rich history and lessons learned. Gordian makes it possible to replicate nuclear reasoning at scale. One overseas reactor company submitted their Regulatory Engagement Plan to the NRC in just three months after their first formal NRC engagement. They used Gordian to benchmark other plans, work through current guidance and reduce the time spent on research and drafting tasks. Another nuclear startup developed initial drafts of a Quality Assurance Program Description and a site-specific Quality Project Plan in three days — work that typically takes four to six weeks. In both cases, the humans stayed in charge. What changed was how much of the slog they could skip.

These uses work for the same reason everything else in Gordian does: the documents describe nuclear work, and Gordian was built to replicate nuclear reasoning.

What we built in response

When we saw customers going further than we'd shown them, we built to meet them there.

The clearest example is skills. A skill is a repeatable workflow — the way your team does a particular kind of work, saved and available to everyone on the platform. Our customers are building their own. One team built a custom procedure-writing skill that produced high-quality drafts in hours, reducing consultant costs and freeing their team for the work that actually required their judgment. The platform makes it possible; the teams are doing it themselves.

We've also made it easier to work together inside Gordian. Build a document (or a blog post?) in Gordian? Go till you're confident, then share the thread and context with your coworkers. They can iterate alongside you, or fork the chat to make their own deliverables. Your knowledge is seamlessly shared, even when you take a much-needed vacation. Our "Projects" feature lets you sync multiple chat threads to the same set of reference documents and context — and bring your coworkers in to review, contribute, or grab the baton. The work and value accumulates in one version-controlled, auditable place.

We built these features because the usage patterns told us to. That's how we operate: listen to what customers say explicitly, listen to what the data shows, and act on it.

Where this is going

People are asking for seats. Colleagues are being pulled in. Departments that weren't in the original scope are finding their own applications and asking to get on the platform. We're seeing inbound demand from teams and organizations that heard about the work from someone else — not from us.

Customers who started with one use case are expanding to others as their users find new applications. I don't think we've seen the ceiling yet.

The foundation scales because the underlying requirement is the same across all of it. We built AI that actually understands nuclear. The breadth is what follows.

Don't just take my word for it — I'm happy to demo Gordian live, share case studies, and connect you with references. Reach out at theresa@everstar.ai.