Two developers, three decades apart — the bottleneck was never typing
AI removes implementation bottlenecks and team boundaries, but makes architecture, testing, and security judgement more important—not less.
Details
In 2024 a Plone core developer quoted a customer for image infrastructure the community had wanted since 2020: a few hundred hours, a five-figure sum. It was not commissioned — the idea was fine, the price was not. It is now open source and in production, carrying the shared web platform of Austria's architecture institutions.
His son, an AI integration engineer, dissolved a handoff between two teams at an Austrian software company. The backend team is retiring legacy. The company is hiring.
The bottleneck was never typing — it was boundaries, and what an organisation does with the space they leave is the whole story. What did not fall: architecture, testing and security judgement.
What attendees take away
- How far up the ladder you actually have to climb — prompting, context, fine-tuning, training — the three conditions that decide it, and a support desk that automated the easy half and kept everyone.
- A Kubernetes operator in Go, a transcoder in Rust — by someone who writes neither — what happens to architecture decisions once your team's skill inventory stops deciding them for you.
- Where the multiplier meets zero — the same tool that multiplied a maintainer of two decades would not have created one — and what it costs, in discipline, at either end of a career.
- The upgrade you kept deferring — twelve institutions stuck on an old release because migrating was never worth the price — now moved, with a web archive back to 2002 intact. The price changed, not the platform.