If you push it to the limits, it's clearer which limits are a temporary problem and which are much harder to fix.
Structures, architectures, context, these can be improved with more training, more data, more hardware.
If you mean something that mocks things up quickly, Figma has always done that and it hasn't destroyed any jobs.
But a huge part of coding has always been understanding the product and building a good feedback-product loop. The product loops will be a lot more efficient. But the glue between feedback-product has always been humans. If you somehow make a robot QA and a robot PM, these will probably be maintained by engineers. In the last few years, we've seen QA becoming very technical, and we'll probably see PMs be more technical or engineers being more product oriented. Even in my current job at a newer startup, there's no PM, only a funnel from sales to product.
I think there will be a time when we see no code. No Java, no JS, and machines will just write machine language. This might still be... 30 years ahead, especially if you look at patterns in the last 30 years.
But until then, we still need people who understands how one screen links to another, how APIs are called, how to debug them from logs. Claude Code is this future - it's amazing at this abstraction, but you can feel where it starts to fall apart.
Structures, architectures, context, these can be improved with more training, more data, more hardware.
If you mean something that mocks things up quickly, Figma has always done that and it hasn't destroyed any jobs.
But a huge part of coding has always been understanding the product and building a good feedback-product loop. The product loops will be a lot more efficient. But the glue between feedback-product has always been humans. If you somehow make a robot QA and a robot PM, these will probably be maintained by engineers. In the last few years, we've seen QA becoming very technical, and we'll probably see PMs be more technical or engineers being more product oriented. Even in my current job at a newer startup, there's no PM, only a funnel from sales to product.
I think there will be a time when we see no code. No Java, no JS, and machines will just write machine language. This might still be... 30 years ahead, especially if you look at patterns in the last 30 years.
But until then, we still need people who understands how one screen links to another, how APIs are called, how to debug them from logs. Claude Code is this future - it's amazing at this abstraction, but you can feel where it starts to fall apart.