Post

This is spot on for any discipline: "AI tools help experienced DEVELOPERS more than beginners... AI is like having an eager junior DEVELOPER on your team. They can write CODE quickly, but need constant supervision & correction." I've referred to Claude as my *Recalcitrant Intern* numerous times! 🙄

Tim Kraut

@timkraut.de

Another pattern that works really well is to use AI to explain code/tools. Have a library with some bad docs? AI can probably help to understand how it works - way faster than manually reading through the source code. Same for unfamiliar programming languages

baxate

@baxate.bsky.social

5 Jan 2025

I don't see the value in using an LLM to do something I already know how to do. As the article states, if I have to review and rewrite its output, I could have just done it properly myself. And I'm already fast at doing things I know how to do.

Mark Techson

@marktechson.com

5 Jan 2025

📌

The resistance to AI in coding reminds me of two major transitions in programming history. When assembly language first replaced machine code in the 1940s, programmers insisted 'if you can't write it in binary, you don't understand what you're doing.

Early assemblers were buggy - devs had to verify output bit by bit. Then in the 50s-60s, when FORTRAN arrived, assembly programmers argued 'why waste cycles on a compiler when I can write better code myself?' Early compilers were unreliable, forcing code inspection.

Both transitions faced skepticism about reliability, performance, and loss of control. Both had early versions that required low-level debugging. Yet both ultimately revolutionized programming by raising the abstraction level while still allowing developers to drop down when needed.