I think it's easy to forget that modern tooling (+AWS/GCP/Azure) let's one dev match or exceed the productivity of 15 developers and a couple of PMs in 2005.
It's a laughable claim to say one developer has replaced 15, and obviously not true. The vast majority of code was business logic and still is.
Today's 'process' is no more efficient than, say, deploying rails to heroku in 2007. And even before that, you'd spend half a day writing an automatic deployment script, and then deployments would take a couple of clicks and you'd never think about it again.
That's a very specific meaning, like in terms of scaling maybe? But in terms of actually meaningful problems solved for end users... But there very nature large scale systems aren't very common, but everyone is chasing that unicorn startup which can serve 10 million users; so scalable APIs are more "practical" then simple workflows
Scalability seems overhyped. If you write small efficient systems they can handle a lot of work. If you use big clunky frameworks that convert simple things into map-reduce style problems of course you're going to care about scalability and how much your AWS bill is going to be.
In many technical interviews, they want to talk about "scalability", using fancy big data software for horizontal scalability etc
But I also know from experience that many, many of these problems would be more elegantly solved by more traditional tools like Postgres, especially since servers have gotten more powerful, the cloud service options more plentiful and reliable, and the software more optimized. The "scalable" approach can lead to massive amounts of wasted person hours unless you're sure you really need it. But if you say, "just use RDS or CloudSQL, or maybe BigQuery", you get perceived as a newb by the 24 year old who just got his MS doing Spark work on toy problems.
I spent all day today and Friday just trying to get a Google Cloud Composer project to run locally. I'm still waiting for that increased productivity that modern tooling supposedly grants me.
Today's 'process' is no more efficient than, say, deploying rails to heroku in 2007. And even before that, you'd spend half a day writing an automatic deployment script, and then deployments would take a couple of clicks and you'd never think about it again.