Hacker News new | past | comments | ask | show | jobs | submit login

> used the various LLM tools, I've had coworkers use the various LLM tools. When you start to tally up the time spent micromanaging, sifting through jank code and tuning it to get what you want the end result is in my opinion more effort for less.

I just can’t believe that to be the case form anyone who has actually spent significant time with real code in real projects. I get that this is someone would say to dismiss it if they hadn’t used it. Like weavers claiming the mechanical loom will introduce too many errors to be with it or typesetters complaining about the kerning of automated processes. Even if you hate all the functions, classes and procedures generated by copilot, it’s a huge productivity gain for the context aware auto completion alone.

> If you want to just shit out some code that barely works for a basic SPA then you can maybe work faster. But you're paying the cost in a larger amount of technical debt that will hit you later [1]

I never once complained that my typechecker couldn’t write a SPA on its own, I never once complained that my linter couldn’t write and SPA on its own. Why should I complain that copilot isn’t capable of doing literally everything unsupervised. It’s cutting away at least 40% of the time I used to spend coding, that’s amazing, why would I not use it just because some set in their ways programmers stubbornly refuse to see the value without even using it?

> The idea that people should just 'adopt it and move on' seems like you can't actually convince people of its value proposition.

For unit testing I needed to explain the value propitiation. For automated build pipelines I needed to explain the value proposition. For copilot I don’t have to. It’s literally just a case of “turn it on, work as usual and you’ll see what all the fuzz is about and it’ll save you tons of time” I have yet to have a single developer who actually started using it need to have the value explained to them, it’s obvious, it’s right there, it’s not a matter of pro’s vs cons.

The only people who want to be sold on it are the ones who won’t turn it on to begin with and instead want some meaningless argument about “the value of real inteligenge” no one ever wanted that argument for unit tests or for ci/cd. I refuse to have those discussions now, it’s just a tool, flip it on, continue to work. If you’re worried AGI will replace you, then not wanting to use copilot makes absolutely no difference either ways. The only thing anyone accomplishes by stubbornly refusing to pick up the major advances we’re getting these days is that they get fired because a fresh out of uni dev is outperforming their 20y or experience and management is never going to buy the “using AI is cheating” mentality they these neckbeards are pushing.

But I see your worry, tech debt is a very real phenomenon which didn’t exist before ChatGPT and therefore people need to decide do they want the old way of always flawless software or the new reality of bugs, tech debt and bad developers which didn’t exist before. /s Of cause you’ll see bugs in code written with copilot just like you’ll see it in code without copilot. But if you have competent people the difference will only be how quickly you get through it not if it’s perfect or not.




Consider applying for YC's Fall 2025 batch! Applications are open till Aug 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: