The Bit of Robotics AI Doesn't Touch
The thing AI is eating and the thing your child is actually learning doing coding and robotics are not the same thing, even though from the outside they look like they might be.
You've been reading the news: another piece on AI taking entry-level coding jobs, another article about computer science graduates struggling to find work, another think-piece on the bottom of the career ladder being kicked out from under the next generation.
AI does do the coding bit, and that this is much less of a problem for your child than the news cycle is making it sound. The thing AI is eating and the thing your child is actually learning doing coding and building robots are not really the same thing, even though they look like they might be.
The bit AI is good at
When people say AI does the coding now, they are, broadly, right. The current generation of language models will write a sorting algorithm in the time it takes you to ask for one, fix syntax errors before they've finished appearing on the screen, and produce a serviceable version of almost anything a junior developer would have been asked to do in their first six months on the job. The remembering-the-comma part of programming, the where-does-the-bracket-go part, is being absorbed into the tools, and it is reasonable to assume that your child will not, in 2040, be earning a living by typing for-loops faster than the next person.
If that was the only thing teaching children about code was supposed to achieve, you'd be right to wonder whether the whole project has quietly passed its expiry date.
The bit AI is nowhere near solving
The physical world is much harder for machines than the digital one, and the reasons aren't going away: lighting changes through the day, surfaces are slippery in ways the model didn't predict, things move when nobody expected them to. AI is racing ahead at writing code. It is barely moving on any of this.
The hard part of robotics is not the code but the gap between the code and the world.
What the kit is teaching, even when nobody says so
When a seven-year-old programs their robot to drive forwards three feet and turn left, what happens often is that the robot drives forwards two feet and stops, or four feet and crashes into the cat's water bowl, or covers the right distance but turns the wrong way and ends up under the sofa. The child notices the difference between what they told the robot to do and what it actually did, goes back to the program, changes something, runs it again, finds it still doesn't quite work, changes something else, and eventually arrives at a robot doing exactly what it is supposed to do.
That loop, the noticing-the-gap-and-fixing-it loop, is the lesson, and it is the same thing we pointed at when she wrote that block-based programming isn't really programming so much as debugging dressed up in friendly colours. Logic and debugging, in the sense of looking at a system that isn't behaving the way you expected and working out why, is by some distance the skill AI is currently least good at replacing. The current crop of models is increasingly capable of producing a confident-looking first draft of almost any solution; they remain strikingly bad at noticing when that first draft has produced a robot wedged under the sofa with its wheels spinning.
What this means for you
None of this means you need to work out whether Scratch or block-based coding is the right starting point, or fret about whether the kit you bought in 2024 has the most current features. What matters, when you're working out what your child should be spending their time on, is whether the activity puts them in regular contact with physical things that don't quite behave the way the instructions said they would.
Robotics does this almost by accident, but it isn't the only thing that does. Cooking does it. So does woodwork, growing vegetables in a small patch of garden, fixing a bicycle that won't shift into third. The underlying skill is the same across all of them: ideas don't become reality cleanly, and most of being useful in the world is knowing how to close that gap. Robotics happens to package this in a form that also looks like the future, which makes the gap-closing feel current rather than nostalgic, and in the present climate that isn't nothing.
The skills they're picking up transfers to most other things if they wander off and become obsessed with watercolours or birdwatching instead.
We won't be able to predict, with any confidence, which jobs will exist in 2040, because nobody currently can, and any advice that takes the form of "learn this specific thing because the market will want it" has a long track record of ageing badly. Advice that takes the form of "spend time on activities that build judgement about the messy physical world" tends to age rather better, for the simple reason that the messy physical world isn't going anywhere.
