Comment by oblio
2 days ago
What tech stack do you use?
Betting in advance that it's JavaScript or Python, probably with very mainstream libraries or frameworks.
2 days ago
What tech stack do you use?
Betting in advance that it's JavaScript or Python, probably with very mainstream libraries or frameworks.
FWIW. Claude Code does great job for me on complex domain Rust projects, but I just use it one relatively small feature/code chunk at the time, where oftentimes it can pick up existing patterns etc. (I try to point it at similar existing code/feature if I have it). I do not let it write anything creative where it has to come up with own design (either high-level architectural, or low level facilities). Basically I draw the lines manually, and let it color the space between, using existing reference pictures. Works very, very well for me.
Is this meant to detract from their situation? These tech stacks are mainstream because so many use them... it's only natural that AI would be the best at writing code in contexts where it has the most available training data.
> These tech stacks are mainstream because so many use them
That's a tautology. No, those tech stacks are mainstream because it is easy to get something that looks OK up and running quickly. That's it. That's what makes a framework go mainstream: can you download it and get something pretty on the screen quickly? Long-term maintenance and clarity is absolutely not a strong selection force for what goes mainstream, and in fact can be an opposing force, since achieving long-term clarity comes with tradeoffs that hinder the feeling of "going fast and breaking things" within the first hour of hearing about the framework. A framework being popular means it has optimized for inexperienced developers feeling fast early, which is literally a slightly negative signal for its quality.
No, it's a clarification. There is massive difference between domains, and the parent post did not specify.
If the AI can only decently do JS and Python, then it can fully explain the observed disparity in opinion of its usefulness.
You are exactly right in my case - JavaScript and Python dealing with the AWS CDK and SDK. Where there is plenty of documentation and code samples.
Even when it occasionally gets it wrong, it’s just a matter of telling ChatGPT - “verify your code using the official documentation”.
But honestly, even before LLMs when deciding on which technology, service, or frameworks to use I would always go with the most popular ones because they are the easiest to hire for, easiest to find documentation and answers for and when I myself was looking for a job, easiest to be the perfect match for the most jobs.
Yeah, but most devs are working on brownfield projects where they did not choose any part of the tech stack.
They can choose jobs. Starting with my 3rd job in 2008, I always chose my employer based on how it would help me get my n+1 job and that was based on tech stack I would be using.
Once a saw a misalignment between market demands and current tech stack my employer was using, I changed jobs. I’m on job #10 now.
4 replies →