Other operators use agents to quickly deploy one-off bash scripts.
In spite of coding agents supposedly replacing software engineers, I feel like by using them I've had more opportunity to tackle big, pure CS problems than I have in years.
I used to know theoretically that something would be faster, but the opportunity cost of the fast implementation just wasn't worth it.
Now I can send this little research assistant off and get a nice little report. Pretty neat.
Positive users report major productivity gains from AI coding agents tackling complex CS problems by reducing implementation effort, while some negative users dismiss the tools as limited to boilerplate or as replacements for engineers.
No Digg Deeper questions have been answered for this story yet.
I agree with this, but also, I use bash scripts a lot more now that I don’t have to teach myself the syntax every time I want to write a script. Tons of one-off functions in my shell now.
In spite of coding agents supposedly replacing software engineers, I feel like by using them I've had more opportunity to tackle big, pure CS problems than I have in years.
I used to know theoretically that something would be faster, but the opportunity cost of the fast implementation just wasn't worth it.
Now I can send this little research assistant off and get a nice little report. Pretty neat.
@code_star Yep exactly. I've probably done more high quality systems work in the last half year than in my whole programming career before, precisely because the opportunity/human time cost for DoingTheRightThing has been obliterated.
In spite of coding agents supposedly replacing software engineers, I feel like by using them I've had more opportunity to tackle big, pure CS problems than I have in years.
I used to know theoretically that something would be faster, but the opportunity cost of the fast implementation just wasn't worth it.
Now I can send this little research assistant off and get a nice little report. Pretty neat.
@code_star we were just having a discussion about this exact thing at work today!
Go ahead little guy. Try some permeant buffers to avoid allocation overhead. It’s no sweat off my back.
In spite of coding agents supposedly replacing software engineers, I feel like by using them I've had more opportunity to tackle big, pure CS problems than I have in years.
I used to know theoretically that something would be faster, but the opportunity cost of the fast implementation just wasn't worth it.
Now I can send this little research assistant off and get a nice little report. Pretty neat.
@code_star wait, but the coding agent's still replaced you as a software engineer. you're just the nontechnical manager approving your code monkey's report
@finbarrtimbers The machine god giveth and the machine god taketh away
I agree with this, but also, I use bash scripts a lot more now that I don’t have to teach myself the syntax every time I want to write a script. Tons of one-off functions in my shell now.
@arjunqie No?
@code_star This has been my experience as well. So many things I know are possible with the right restructuring and optimizations, but not enough time to do them all.
But now I can brain dump my thoughts and move on to something else while I wait.
@code_star yeah same. the real shift is that deciding what to build stops being blocked by how long it'll take. we rewrote mato's audio pipeline state machine this quarter. previously would've kept kicking that can. when execution cost drops, the design choices become the actual work
@code_star Standard devs just use them for boilerplate, which is pretty mid.
@code_star depends. what would you build instead?
@code_star The opportunity cost point is so true. How many optimizations did I skip because implementing them wasn't worth the effort? Turns out that theoretical knowledge was useful all along.