Last year, I wrote about Learning Culture and Soft Skills — how we learn, how we collaborate, and how much of our effectiveness as engineers is shaped by things that don’t show up in code.

Yesterday, I saw something that connects directly to both.

Not in theory, but in practice.

A Different Kind of Challenge

We ran a hackathon at NN Group with around 300 people. The challenge was intentionally ambitious: find solutions in 120 seconds for problems that normally take weeks or months. Not slides, not concepts, but working solutions — things that could genuinely help customers, whether that means buying a product through ChatGPT, getting a claim paid instantly, or understanding complex life decisions faster.

Agentic AI was everywhere. Not as a gimmick, but as a core building block that fundamentally changed how teams approached the problem.

Something Changed — But Not Where You Expect

And yet, what stood out most had little to do with the technology itself.

It was the behavior.

The hackathon pattern we all know

Hackathons tend to follow a familiar rhythm. They start with energy and excitement, slowly transition into focus, and somewhere during the night that focus turns into pressure. Around midnight, people start pushing harder. Around 2:00 AM, it becomes grinding. The unwritten rule is simple: you stay, you push, and you deliver — no matter how rough the result might be.

That struggle has always been part of the identity.

If it wasn’t painful, did we even try?

When people started going to sleep

Two years ago, we ran out of beds — not because people wanted to sleep, but because they refused to stop working. This time, something else happened. Around 1:00 AM, people started leaving. Not in frustration, not because they were stuck, but simply because they were done for the day.

Out of roughly 300 participants, only about 60 kept working through the night.

At first glance, that feels like a loss of intensity.

But that explanation doesn’t hold for very long.

The Work Itself Has Shifted

Because the context hasn’t fundamentally changed. The same organization, similar teams, similar ambition — and yet the behavior was noticeably different. That suggests the shift is not in the people, but in the nature of the work itself.

Over the past two years, the rise of generative and agentic AI has started to reshape how work gets done. Tasks that used to take hours now take minutes, and parts of the process that were once manual are increasingly automated or assisted.

That sounds abstract — until you see it play out in a hackathon.

The bottleneck moved

A hackathon compresses reality. It exposes where time is actually spent, and where the true bottlenecks are. In previous editions, most of that time went into execution: writing code, fixing bugs, connecting systems, and trying to get something — anything — to run before the deadline.

This time, that part was still there.

But it was no longer dominant.

From execution to judgment

What became visible is that the work shifted, moving away from execution as the primary constraint and towards something less tangible, but ultimately more important. When building becomes faster and easier, the limiting factor is no longer how much effort you can apply, but how well you understand what you are trying to build.

In other words, the challenge moves from execution to judgment.

Rethinking Effort and Value

For a long time, especially in engineering culture, we have relied on a relatively simple equation: effort equals value, hours equal impact, and visible struggle signals importance. That model made sense in a world where execution was the bottleneck.

What we are starting to see now is that this relationship is beginning to break down.

Faster doesn’t mean better

As AI reduces friction in execution, the amount of time required to get something working decreases significantly. At the same time, it becomes increasingly clear that speed alone is not enough. Faster output does not automatically translate into better outcomes.

If anything, it exposes a different constraint: deciding what is actually worth building in the first place.

That is a much harder problem to solve.

What teams did differently

During the hackathon, this shift was visible in how teams spent their time. Less effort went into making something work under pressure, and more into understanding the problem, refining the idea, and shaping the experience.

Not just how do we build this?
But why are we building this, and does it actually solve the right problem?

Yes — people slept

And perhaps most tellingly: people slept.

Not because they cared less, but because the work no longer required continuous grinding to reach a meaningful result. The output did not suffer — if anything, it improved.

What This Means Going Forward

The final teams on stage didn’t present fragile prototypes or disconnected ideas. They showed working solutions that were coherent, aligned with the NN brand, and clearly connected to real customer value.

What used to be exceptional in a hackathon is starting to become the baseline.

That, in itself, is a significant shift.

A different kind of pressure

It also introduces a different kind of pressure.

Not the pressure to work longer hours or push through the night, but the pressure to think more carefully, to make better decisions, and to be more intentional about what is being built.

Because when execution becomes easier, choosing becomes harder.

There is nothing “soft” about this

If AI accelerates the “hard” part of building, then the differentiators move towards what we have historically called soft skills:

  • understanding context
  • making trade-offs
  • collaborating effectively
  • asking better questions

There is nothing soft about that.

Final thought

In a world where everyone can build faster, the real question is no longer how much you can produce, but whether you are producing the right things.

The hackathon didn’t get easier.

But it did become something fundamentally different.