Back to archive

Thread

8 tweets

1
Man if I see one more „the end of programming is nigh“-sayers… programming and creating large scale systems that are resilient, architectured well, maintainable, that‘s so much more than „how do I OAuth in typescript.“
2
Also given the current state of the art, it takes real expertise to spot the mistakes in code that superficially looks right, but fails for subtle reasons.
3
Coding is not just about churning code that works, it is about designing, sometimes discovering, the right abstractions, building structures and tools along the way.
4
Sure, we‘re getting a new bunch of tools that are amazing and are changing programming the way stackoverflow changed it, but that is just a part of all the activity and skill that is required to build software.
5
There is a side here that‘s seldom mentioned which is why we ended up in a situation where we‘re in need of such tools. The real problem is that we‘re often required to memorize dozens of lines of boilerplate code to get something done.
6
We add complexity to our libraries and APIs, we use language that don‘t allow us to properly abstract all that. Patching this up with multi-billion neuron sized smart autosuggest is one way to go at it, but it is covering up this problem.
7
I want to add that people questioned why I wanted to study CS back in 1994 because „these computers will soon program themselves.“ From all I can see, for those who write code, things got increasingly more complicated, not less.
8
I mean let‘s also spin this further… if we have a considerable amount of AI generated code in the future, more code that nobody understands, but you will eventually have to look at when the AI tools stop working... Sounds like you‘ll need a lot of expertise going forward.