Comment by tuveson | Hacker News Reader

Comment by tuveson | Hacker News Reader

Parent Full thread

Programming in a structured high-level language like C is much more productive than programming in assembly. Programming in a managed language with GC can be more productive still.

The creation of package managers and the widespread availability of open-source repos means developers don't need to write as much from scratch.

The creation of search engines and Stack Overflow did (and still do) much of the useful things that people use AI for (boilerplate, debugging obscure error messages).

Machines have gotten exponentially faster for the last several decades. This means devs need to spend less time optimizing code. And the time to compile and run speeds up, meaning you can prototype things faster.

Why is it, that somehow none of these inarguable improvements to the speed and efficiency of development haven't lead to a a massive decrease in the number of developers? If we take it as axiomatic that AI significantly improves productivity, why, for the first time in history, does that not result in more programming jobs?

None of the improvements of the past you outline were advertised as capable of replicating agency. They were sold as tools. Note that I am not making any claim about actual capability for agency, there’s other people more qualified than me to discuss this. Regardless factuality of ability to replace people, this is how it’s sold and arguably also how it’s bought. This gives the view that it has impact on jobs more weight.

Also I think it is plausible (if not likely) that LLMs reduce the demand for other kinds of software. They're (relatively) general purpose and can handle many tasks that previously would've needed bespoke software (and a team of people to create/maintain it).

In some ways developing and working with a high level but less efficient language for velocity is actually a vast improvement, imo.

Take node.js for example, devs can just sling code out as fast as they can and shit gets done. Then the node.js core devs can optimise certain paths/features after the fact to negate many of the efficiency problems.

However it does annoy me that what this has meant is that many of my colleagues don't know anything about memory management, debugging, or any other more traditional concepts, so we see bloat & OOMs over time that need to be resolved.

Reply on news.ycombinator.com