Learning Rust from scratch to the point where I could write my website software was an 18 month effort which began in early 2025. My starting point was many years of developing in Java and Python. But to slow me down, all coding happened outside of work, squeezed in at coffee shops, during lunch breaks or the train home at night. It could definitely have been quicker, but maybe not that much so once you balance everything out. Suffice to say it was a huge commitment and strained my resolution to the limit. It was certainly not a casual skill acquisition.

I do wonder though, having witnessed the advent of AI assisted coding if future generations will learn programming languages in depth. The sudden evolution of the 'software developer' this year has changed the game. Coding from scratch is now replaced by prompting an AI to write code for you. Think I am over-dramatising? Then see this screenshot from an Anthropic presentation in May this year. This transformation has happened so swiftly and yet here we are. My industry will not return to as it was and we must all live the consequences. So will developers ever spend that time learning the intricacies, idioms and basic syntax of programming in a new language for themselves?

In the short term at least, businesses could save money with less engineers. Money spent on tokens instead of fallible, messy, demanding and emotional human beings. But in doing so, the pipeline of new coding talent has been halted. Riding instead on the experience of older developers whose role has narrowed to be AI-prompters and code reviewers, leveraging their decades of toil in the trenches learning it all for themselves. They know a lot of what the AI knows. But this duplication of effort, where countless developers all learn the same lessons, is over. That learning is now centralised and the dependence on a human developer can only diminish with no new developers coming through.

The immediate consequences for developers are a complete re-evaluation of their role. Perhaps they will find themselves in a sweet spot which they could ride out for months or maybe years. During this period they ask advised/encouraged/forced to let AI do the hard work. They deliver the same with less effort or more with the same effort. They rarely writes code and a fundamental aspect of their job which they enjoyed is now gone.

For the junior developer, AI builds code faster and nobody has time to wait for them to figure it out. The AI assists and guides as they write, or just builds it all from a decent specification. The graduate who delivers first and fast is the one that will prevail. The engineer who realises code is being written that they could not write themselves and interrupts to take control will struggle.

The new status quo only succeeds today due to the flipping of the experienced software engineer's role. There are no noticeable disadvantages for the business to begin with. Productivity could increase, headcount could lower and experienced developers accept their role change in the knowledge that cuts most likely lie ahead. But businesses are left with a slightly worrying conundrum. Who will understand their platforms, their systems and their products in future? For short-term benefits they have embraced AI without considering where this will leave them in years to come. The old developers will be paid off in time and the next generation will be unable to implement improvements or diagnose issues for themselves.

In the past skilled draftsmen gave up their pencils and paper to make way for CAD. The hardware engineer who used to put the disks in a new Sun server, install the Solaris OS and put it on the internet is no longer required, we just click buttons on the cloud console instead. But abstractions are good, they release us to perform more efficiently and do higher level tasks. Efficiency trumps all concerns or else a competitor will edge ahead. So is this really any different?

I think it is next step on an existing trajectory:

1. Optimize roles for efficiency - skills change but no knowledge lost e.g. draftsman 2000
2. Centralize solutions to cloud - skills lost but much knowledge retained e.g. UNIX engineer 2015
3. Obliterate roles into AI - skills and knowledge lost e.g. software developer 2030

Surely only nostalgia makes us pine for the old ways. But retaining these skills enables technological independence and self-sufficiency. If not widely understood they will be further centralised into single powerful entities. We need to cultivated them in present & future workplaces. If not everyone outside the central monoliths will lose even more than we already have.

From this view point, learning Rust and writing my website software all from scratch was not simply a hobbyist project, maybe it was also a very, very small act of defiance against the AI tide. And this may also explain why the software for this website is not open source, but that is a topic for another day...