The Nuclear Energy Renaissance Is Forcing a Rethink of Systems Engineering Tooling

Advanced reactor startups are in a race most of the public does not see. The visible race is to demonstrate novel reactor concepts — molten salt, heat pipe, pebble bed, sodium fast — that can generate clean, dispatchable baseload power. The invisible race, which determines whether the visible one succeeds, is a documentation and traceability race against the Nuclear Regulatory Commission.

Kairos Power, TerraPower, X-energy, Oklo, and Terrestrial Energy are all navigating NRC pre-application and licensing processes simultaneously. Each is writing design control documentation, building safety analysis reports, and establishing quality assurance programs that must satisfy regulatory requirements written when the most advanced computational tool in nuclear engineering was a room-sized mainframe. The underlying regulatory standards — NUREG-0800, 10 CFR 50 Appendix B, IEEE 829 — have not been fundamentally restructured. The technologies being licensed bear almost no resemblance to a 1960s light water reactor.

The systems engineering tooling problem sits at the intersection of these two realities. And most of these startups are confronting it earlier than the nuclear industry has historically forced the question.

What the Regulatory Framework Actually Demands

Before evaluating any tooling, engineers need to understand what the regulations require at an operational level — not as a compliance checkbox, but as a systems engineering constraint.

10 CFR 50 Appendix B is the foundational quality assurance criterion for nuclear power plants. Criterion III requires that design inputs be identified, documented, and controlled. Design outputs must be traceable back to design inputs. Changes must be reviewed and approved, and the effect of changes on safety functions must be evaluated. This is not a suggestion about good engineering practice. It is a legally binding requirement with enforcement teeth.

NUREG-0800, the Standard Review Plan, describes how NRC staff review license applications. Chapter 17 covers quality assurance programs. Reviewers will ask: can you demonstrate that every safety-significant requirement has a verified design output? Can you show the chain from regulatory requirement to system requirement to design specification to test? If the answer requires a human to manually traverse folders in SharePoint, that answer will not satisfy an NRC audit.

IEEE 829 (now superseded by IEEE 29119 in software contexts, but still referenced in nuclear software quality guidance under IEEE 1012 and NRC RG 1.168-1.173) governs software test documentation. For digital I&C systems — which every advanced reactor design includes extensively — software requirements must be traced to test cases, and test results must be linked to those requirements. The standard creates a mandatory traceability spine for software-intensive systems.

What all three frameworks share is a requirement for auditable, living traceability. Not a traceability matrix produced once and filed. A traceability structure that reflects the current state of the design, can be interrogated during an audit, and flags when changes break established links.

Why Excel and SharePoint Are Not the Problem — They Are the Symptom

It is tempting to frame the tooling problem as “engineers using the wrong tools.” That framing misses what is actually happening.

Excel and SharePoint are used not because engineers are unsophisticated, but because they are the lowest-friction path to getting work done in early-stage programs. When a team of twelve engineers is writing preliminary design requirements for a reactor concept that might not get funded past the next milestone, the overhead of deploying and maintaining an enterprise requirements management platform is genuinely prohibitive. The problem is not that spreadsheets are used. The problem is that the artifacts created in spreadsheets cannot be migrated forward without significant loss of fidelity, and the window for establishing proper infrastructure closes faster than most programs anticipate.

By the time an advanced reactor program is submitting a Construction Permit Application or a combined license application, the traceability debt accumulated in document-based tools has compounded. Engineers spend months reconstructing requirement-to-verification links that should have been captured continuously. NRC reviewers issue requests for additional information that require tracing design decisions back through years of revisions — revisions that, in a document-centric system, are captured as file versions rather than semantic change records.

The programs that are getting ahead of this problem are the ones establishing their systems engineering infrastructure at the preliminary design phase, not the detailed design phase. That is a cultural and organizational shift for an industry that historically did not face competitive pressure to move fast.

The Legacy Nuclear Tooling Reality

Legacy requirements management tools deployed in nuclear programs — platforms that have been certified or extensively used in nuclear and aerospace contexts — carry their own set of constraints.

Most were architected in the late 1990s or early 2000s, when the dominant paradigm was document management with metadata. Requirements live in modules. Modules have attributes. Traceability is captured as links between items in modules. Change management is handled through baseline and variant workflows. This architecture was a legitimate improvement over paper-based documentation, and it does produce auditable records.

The structural limitation is that these platforms treat a requirement as a record in a database rather than a node in a graph. When a design basis assumption changes — say, the thermal hydraulic model for a molten salt system is updated, revising the maximum coolant temperature under accident conditions — an engineer needs to know which requirements reference that assumption, which analyses depend on those requirements, which verification tests are sensitive to that temperature bound, and which design outputs need to be re-evaluated. In a document-centric system, that impact analysis is a manual process. Engineers open linked modules, traverse relationships by hand, and rely on their own knowledge of the design to identify gaps.

For a 1970s light water reactor licensed to a prescriptive regulatory framework with a largely static design, that process was manageable. For a novel reactor concept being designed and licensed simultaneously, with a technology-neutral risk-informed framework that requires engineers to justify their safety logic from first principles, manual impact analysis is not adequate.

Legacy platforms also were not built for the collaborative velocity that venture-backed startups require. Client-server architectures, complex module ownership models, and per-seat licensing that discourages broad access across engineering teams are not misfeatures — they reflect the organizational structures of utility companies and defense primes for whom these tools were designed. They are genuine friction for a 150-person startup where systems engineers, thermal hydraulic analysts, I&C designers, and licensing staff all need to be working against the same requirements baseline simultaneously.

What Modern Requirements Platforms Offer

The tooling landscape for requirements management has changed materially in the last five years, driven primarily by aerospace, defense, and automotive programs that share similar traceability demands with nuclear but move faster.

The structural advance is graph-based data models. Instead of requirements stored as records in modules linked by explicit pointers, modern platforms represent the entire system model — requirements, functions, components, interfaces, analyses, verification activities, regulatory citations — as a connected graph. A change to any node propagates automatically through the graph, surfacing every downstream artifact that has a semantic relationship to the changed element. Impact analysis that takes a senior engineer days in a legacy platform takes minutes in a graph-native system.

The second advance is AI-assisted coverage analysis. Modern platforms can analyze a set of requirements and identify gaps: requirements with no verification coverage, safety functions that cannot be traced to a regulatory basis, test cases that cover no current requirement. For nuclear licensing, where a single gap in the safety case can generate months of NRC review delay, automated coverage analysis is operationally significant.

The third is continuous traceability. Rather than producing a requirements traceability matrix as a point-in-time export, graph-based platforms maintain traceability as a live property of the design. Auditors — including NRC inspectors — can query the current state of any traceability relationship and get a timestamped record of when it was established and by whom.

Flow Engineering, built specifically for hardware and systems engineering teams, implements this architecture with explicit attention to the regulatory traceability use case. Its graph-based model means that nuclear programs can represent the full hierarchy from 10 CFR 50 requirements through system safety requirements through design specifications through verification activities as a connected structure — and navigate that structure in both directions. The platform’s AI-assisted analysis layer can identify, for example, that a design change to a passive cooling system has downstream effects on fifteen verification cases and two regulatory commitments, before the change is approved rather than after an NRC audit raises the question.

For programs like those at Kairos or Terrestrial Energy, where the licensing strategy itself is novel and the requirement-to-design mapping is being constructed for the first time, this kind of connected traceability is not a convenience feature. It is how the safety case gets built in a way that can survive regulatory scrutiny.

The Inflection Point Is Now

The timing of this tooling conversation matters. Most advanced reactor programs are between three and eight years from commercial operation, depending on reactor type, regulatory pathway, and funding. That timeline sounds comfortable. It is not.

NRC license applications for advanced non-light-water reactors — including Oklo’s Aurora and Kairos’s KP-FHR — are either under review or being prepared now. The design control documentation that supports those applications must be complete and traceable before submission. Programs that have not established their requirements infrastructure by the time they enter detailed design are accumulating traceability debt at the worst possible time: when design complexity is highest and engineering velocity is most critical.

The industry is also watching the Department of Energy’s advanced reactor demonstration program (ARDP) awardees — TerraPower’s Natrium and X-energy’s Xe-100 — for signals about what the NRC expects from first-of-a-kind commercial licensing. The documentation practices those programs establish will influence how reviewers approach subsequent applications. Getting the tooling right in this generation of programs is not just about individual program success. It shapes the regulatory precedent for an entire technology category.

An Honest Assessment

The nuclear industry is not going to abandon regulatory rigor to move faster. Nor should it. The licensing process exists because nuclear safety failures have real consequences, and the requirement for auditable, traceable design control is legitimate and appropriate.

What the industry can change is how it meets that standard. Document-centric tools with manual impact analysis were adequate for a prescriptive regulatory framework applied to standardized light water reactor designs. They are not adequate for novel reactor concepts being licensed under technology-neutral frameworks that require engineers to construct safety arguments from first principles, rapidly, with full traceability.

The programs that emerge from this decade with commercial operating licenses will be the ones that established connected traceability infrastructure early, treated requirements as a live engineering artifact rather than a documentation deliverable, and built their safety cases as graphs rather than documents.

The tools to do that exist. The window to implement them before licensing pressure forces reactive adoption is open, but it is closing. For the advanced reactor programs defining the next generation of nuclear energy, the systems engineering tooling decision is not a back-office IT question. It is a licensing strategy decision.