AI makes expertise more valuable.
As AI accelerates generation, experience, evaluation and technical judgement become the decisive bottleneck.
Print production taught me early that an error does not remain abstract. A wrong file becomes waste, delay and real cost. AI now changes the speed at which I can build solutions. It does not remove that responsibility.
AI can generate code, build interfaces and turn an idea into a convincing prototype in a short time. That changes software development. It does not make experience obsolete. On the contrary: the cheaper generation becomes, the more important the ability to assess the result correctly becomes.
The demo shows only the visible part
A prototype answers whether an idea can work in principle. Production asks different questions. What happens with invalid input, expired tokens or an interrupted deployment? Where are secrets stored? How is data migrated? Which events need to be observable so that users are not the first to report a failure? And how do I safely return to the last working version?
This work is rarely the spectacular part of a demo. Yet it determines whether a system impresses once or remains dependable. Experience appears not only in writing code, but in recognising risks, boundaries and missing requirements.
AI amplifies judgement
I like using AI to explore alternatives, understand unfamiliar areas faster or shorten routine work. Its output is material to me, not authority. I still have to decide what fits the architecture, which assumptions need verification and when a simple solution is better.
With my Typst tools and spotpdf, knowledge from print, publishing and PDF production guides those decisions. AI can accelerate implementation. Knowing which problems matter and what a robust solution looks like comes from the domain.
That makes expertise more valuable, not less. The bottleneck shifts from generating to evaluating. Speed is useful. Judgement makes it dependable.