In 1968, NATO declared a software crisis.
Most people assume we solved it. We didn’t. We just moved it to a bigger room.
Waterfall, Scrum, DevOps, cloud, microservices — each wave helped, and each wave also raised the ceiling on complexity. A “simple” web app today involves distributed caches, managed identity platforms, automated pipelines and a dozen third-party integrations. Teams drown not because they lack process, but because the surface area of responsibility keeps expanding.
What’s different now is that AI doesn’t just add another layer of process. It can absorb complexity itself: interpret code, surface architecture, bridge roles, generate documentation, spot gaps. For the first time, we have a technology that scales the discipline — not just the tooling.
The software crisis didn’t end. We are learning to outgrow it.
I wrote about this on my blog: https://culture-engineers.nl/blog/2025/12/01/the-software-crisis-never-ended-it-just-evolved/
In a small German town called Garmisch, fifty of the brightest minds in computing gathered in 1968 to name a problem that was costing governments and organisations dearly. Projects ran late, went over budget, or failed entirely. Software had no consistent methods, no agreed process, no reliable way to scale.
They called it the software crisis.
What strikes me most is not that it existed, but that it never really ended. Every decade brought a new answer: structured lifecycles, iterative models, DevOps, cloud. Each evolution helped. None eliminated the underlying tension. Software grew faster than our ability to manage it.
Today, AI is the first technology that might genuinely change that equation. Not by adding more ceremony, but by absorbing the noise — so engineers can focus on the decisions that actually matter.
Seen that way, progress in this field is not about solving the crisis. It’s about our ability to meet it.
I wrote about this on my blog: https://culture-engineers.nl/blog/2025/12/01/the-software-crisis-never-ended-it-just-evolved/
What if the software crisis of 1968 never ended — it just changed shape?
From punch cards to waterfall, from agile to DevOps to cloud, every generation inherited the same underlying tension: software grows faster than any team’s capacity to coordinate it. The tools got better. The complexity grew faster.
Here’s what I think is genuinely new about AI: it’s not another methodology. It’s the first technology capable of addressing complexity itself rather than wrapping a new process around it. AI can absorb context, connect dots across roles, surface decisions and handle the parts of the job that expand silently, things like documentation, architecture drift, pipeline maintenance, and integration gaps.
The challenge for leaders is not whether to adopt AI. It’s whether to use it to reduce developer overload or simply to raise the output ceiling and add it back.
That’s the real software crisis of 2025.
I wrote about this on my blog: https://culture-engineers.nl/blog/2025/12/01/the-software-crisis-never-ended-it-just-evolved/
Portrait-format editorial illustration (1024×1536). A vertical timeline runs through the centre of the frame, starting at the bottom with a glowing punch card surrounded by early 1960s mainframe silhouettes, ascending through a waterfall diagram, a circular agile sprint ring, a cloud cluster, and arriving at the top with a warm neural-network node radiating soft light. Each era is rendered as a distinct layer with its own muted colour: amber for mainframes, slate-blue for waterfall, green for agile, sky-blue for cloud, gold for AI. A single orange thread connects all layers. Dark navy background, flat minimalist style, no text, no people, no handshakes.
Wednesday — mid-week is strong for conceptual/historical think-pieces. Avoid Monday (news overload) and Friday (low engagement on long-form).