Software development has absorbed wave after wave of abstraction.
Machine code became programming languages. Infrastructure became cloud services. Reusable libraries became enormous ecosystems. Deployment became automated.
Then AI began changing the act of writing software itself.
AI didn't make developers irrelevant in 2025. It started changing what one developer could accomplish.
The blank editor became less intimidating.
AI coding tools could generate boilerplate, explain unfamiliar code, propose tests, help debug problems and accelerate documentation.
The immediate benefit was obvious: less time spent on repetitive work.
But the more important effect may be what happens to the developer's role as these systems improve.
Typing code is not software engineering.
Software engineering includes understanding requirements, designing architecture, making trade-offs, protecting data, considering failure modes, reviewing behaviour and maintaining systems over time.
Generating code faster does not eliminate those responsibilities.
In many cases, it makes judgment more important because developers can now produce more software more quickly.
The bottleneck moves.
When producing code becomes cheaper, understanding what should be built becomes relatively more valuable.
Architecture matters. Product judgment matters. Security matters. Testing matters.
Small teams become interesting.
If AI meaningfully increases engineering leverage, products that once required large development teams may become feasible for much smaller ones.
That can change startup economics, enterprise development and the amount of bespoke software organisations are willing to build.
The developer isn't disappearing.
The developer is moving upward through another layer of abstraction.
And just as higher-level programming languages did not eliminate software engineering, AI-assisted development is unlikely to eliminate the need for people who understand how complex systems should work.
It may simply give those people considerably more leverage.