When Complexity Tips: Tim Weilkiens on the Digital Storm in Engineering

The first keynote of ReConf 2026 was, once again, about AI, as it could have been otherwise. But that's not what this is about. The second keynote on Wednesday morning by Tim Weilkiens offered a holistic view of the developments of our time. Yes, of course, it was also about Artificial Intelligence, but that was just one wave in the perfect storm we are currently in.

A storm from multiple directions

Tim chose the image of the Perfect Storm, a 2000 film. Several developments are converging simultaneously, creating a situation that can no longer be controlled by experience alone. Artificial intelligence is part of this. The presentation makes it clear that it's not about a single technology. Digital twins, end-to-end data chains, and new modeling languages reinforce each other.

This interplay leads to a situation where familiar approaches lose their effectiveness. Previous projects can no longer be used as a reference. Decisions must be made under different circumstances than just a few years ago. In doing so, he also refers a lot to Langdon Morris, whose work I reported on here six years ago.

Three curves diverge

A central argument of the keynote is the simultaneity of multiple exponential developments. The complexity of systems has been increasing for decades. Works by Olivier de Weck These developments show over long periods. In the daily lives of many projects, this increase is perceived as steady pressure. Requirements are increasing, interfaces are becoming more numerous, and dependencies are growing.

The development achieves exponential growth in three areas: complexity, tools, and people and organizations.

Tim Weilkiens

In parallel, tools are developing at high speed. Computing power, software platforms, and most recently, AI-based systems are expanding the scope of action in engineering. This development also follows an exponential curve and has significantly gained momentum in recent years.

The third curve concerns people and organizations. Adaptation occurs, but it is significantly slower. Transformations take years. Agile methods were introduced a long time ago and are still not consistently established today. This asymmetry creates a bottleneck. Systems and tools are developing faster than organizations' ability to use them effectively.

Complexity doesn't disappear

Tim picks up on the fact that complexity cannot be reduced. It can be shifted and structured, but not eliminated. Modularization, clean interfaces, and architectures change the form of complexity. Olivier de Weck calls this 1st law of systems engineering.

To provide context, the presentation refers to the Cynefin Framework, which distinguishes between complicated and complex problems. Complicated problems can be analyzed and solved. Complex problems require experimentation. Approaches emerge through trial, observation, and adaptation.

Systems engineering is increasingly moving into this complex domain. Traditional planning is reaching its limits there. Agility is an attempt to deal with this situation. The talk suggests that this answer also reaches its limits if the underlying dynamics continue to increase.

Transformation meets human limits

A recurring motif is the question of adaptability. How quickly can people and organizations change? Everyday examples show that this ability is limited. A portion of the population does not use digital technologies. Legal initiatives attempt to maintain analog access. Companies consciously decide against certain developments and accept disadvantages.

These observations stand in contrast to technical development. While systems and tools are growing exponentially, the capacity for adaptation remains limited. This creates a structural tension that cannot be resolved through methods alone.

The turning point has been reached

Langdon Morris described years ago this turning point that we are now encountering. Developments reach a phase where their effects can no longer be extrapolated linearly. Artificial intelligence naturally plays an important role in this context and drastically accelerates these developments.

The presentation avoids a dystopian portrayal. We cannot change the development, and attempts to stop it are ineffective. Preparation for this situation is crucial. Organizations must learn to work with the dynamics instead of against them.

Processes remain, implementation changes

An important point is the stability of processes. In particular, inputs and outputs remain constant. Systems and requirements must be formulated, more carefully than ever. And people still have to make decisions.

However, the implementation of these processes is fundamentally changing. Artificial intelligence can take over large parts of the operational work that takes place between input and output. The lecture talks about scenarios in which systems are largely generated automatically, which is done using approaches such as Vibe Coding gets easier and easier.

This makes requirements and testing critical points, at least in the medium term. This is exactly where Tim currently sees weaknesses. Requirements engineering and testing are not keeping pace with the development of tools.

Models become the carriers of semantics

And with that, we come to modeling. Documents and tables are not enough to convey the semantics of data. A value in a table contains no information about its context. Meaning only arises through structure and relationships.

Models offer this structure because they can represent relationships explicitly and make them machine-readable. This makes them the basis for end-to-end data chains. The concept of Digital Thread describes this connection over the entire lifecycle of a system.

In this perspective, models become the crown jewels of engineering. They contain the semantics on which further automation is built.

New Roles in Engineering

Based on these models, a different tool landscape is emerging. The presentation describes a shift away from monolithic engineering tools. Instead, many specialized applications are being created for specific tasks. These applications access common models.

In such a setting, a small group of experts develops and maintains these models. They ensure consistency and define the structure. Most users work with tools built on top of them. These tools are tailored to specific tasks and are significantly easier to use.

This development is closely linked to advances in software development. Applications can be created faster. AI supports their implementation. This makes it economical to operate many small tools instead of a few large systems.

SysML v2 as a Technical Enabler

In this context, SysML v2 enormous potential. The language is not presented as a finished solution, but as a necessary building block. It creates the foundation for precise and machine-readable models.

An essential aspect is Formal Semantics. SysML v2 is based on a formal logic (First-order predicate logicwhich allow for unambiguous interpretation. This is a prerequisite for collaboration with machines. At the same time, through a textual notation facilitates access.

Another point is the integration of temporal processes. SysML v2 models can represent states at different points in time. This allows digital twins to be directly mapped in the model. Variant management is also supported and receives a formal basis.

SysML has come a long way. Many people have developed it, partly on a voluntary basis and with great passion, to bring it to this point. Through its anchoring with the OMG, an international standardization organization, it is designed for the long term and focused on stability. Despite the Acquisition of OMG by EDM This won't change.

Integration instead of documentation

A recurring motif is the demarcation between superficial and in-depth digitalization. Digitization vs. Digitalization. Superficial digitization, as in PDFs, does not make the data in documents truly accessible. They can be stored and exchanged, but not meaningfully processed further. Machines can only handle this form of information to a limited extent.

Value is only created when data is semantically structured. Models enable this structure, making information accessible for analysis, simulation, and automation. The Digital Thread thus becomes practically implementable.

This is becoming more urgent due to current developments, and SysML v2 has the potential to play an important role here. Without semantic models, the potential of AI in engineering remains limited.

A look at the practice

The presentation concludes with a concrete example. The startup Planetary Utilities develops satellites with a small team. The tools used connect modeling, simulation, and implementation in one environment.

The way of working differs significantly from classical projects. In a work environment, the engineer interacts with an agent and sees the effects directly in various model views, whether it's a CAD outline, textual SysML, or an animated visualization in space. Switching to other tools is also possible, such as OnShape for 3D CAD. Feedback loops are short. Changes are implemented and reviewed immediately. Data remains consistent and is available at all times. The focus is on the engineering itself, not on operating tools.

Conclusion

The keynote assesses the developments that indicate a „perfect storm.“ The described developments are real and interconnected. Complexity continues to grow. Tools are evolving rapidly. Organizations must find ways to deal with this dynamic.

The crucial point lies in semantics. Models that convey meaning become the foundation for further engineering. Without this basis, the use of modern tools remains fragmented.

The future of Systems Engineering doesn't come from new processes. It comes from a different use of models and consistent integration of data. SysML v2 could play an important role in this.