Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Nobody knows how to build with AI yet

worksonmymachine.substack.com

Post reply on HN
Print view View original

61–70 of 432 posts

Re: Nobody knows how to build with AI yet

#61

Post by prinny_ »

mnky9800n wrote: post #49
lordnacho wrote: post #33

Earlier quoted context omitted.

But you also can't not swim with the tide. If you drove a horse-buggy 100 years ago, it was probably worth your while to keep your eye on whether motor-cars went anywhere. I was super skeptical about a year ago. Copilot was making nice predictions, that was it. This agent stuff is truly impressive.

I think the agent stuff is impressive because we are giving the AI scaffold and tools and things to do. And that is why it is impressive because it has some directive. But it is obvious if you don't give it good directives it doesn't know what to do. So for me, I think a lot of jobs will be making agents do things, but a lot won't. i think its really strange that people are all so against all this stuff. it's cool ne…

A lot of people join this profession because they like building stuff. They enjoy thinking about a problem and coming up with a solution and then implementing and testing it. Prompting is not the same thing and it doesn't scratch the same itch and at the end of the day it's important to enjoy your job, not only be efficient at it.

I have heard the take that "writing code is not what makes you an engineer, solving problems and providing value is what makes you an engineer" and while that's cool and all and super important for advancing in your career and delivering results, I very much also like writing code. So there's that.

Re: Nobody knows how to build with AI yet

#62

Post by bloppe »

lordnacho wrote: post #33
karel-3d wrote: post #24

Reading articles like this feels like being in a different reality. I don't work like this, I don't want to work like this and maybe most importantly I don't want to work with somebody who works like this. Also I am scared that any library that I am using through the myriad of dependencies is written like this. On the other hand... if I look at this as some alternate universe where I don't need to directly or indirec…

But you also can't not swim with the tide. If you drove a horse-buggy 100 years ago, it was probably worth your while to keep your eye on whether motor-cars went anywhere. I was super skeptical about a year ago. Copilot was making nice predictions, that was it. This agent stuff is truly impressive.

An I the only one who has to constantly tell Claude and Gemini to stop making edits to my codebase because they keep messing things up and breaking the build like ten times in a row, duplicating logic everywhere, etc? I keep hearing about how impressive agents are. I wish they could automate me out of my job faster

Re: Nobody knows how to build with AI yet

#64

Post by nojito »

aprilthird2021 wrote: post #5

Eh, idk. First of all, the article is really wordy to say very few things. That just frustrated me a bit. Second of all, it's easy to fart out some program in a few days vibe coding. How will that fare as more and more features need to be added on? We all used to say "Dropbox that's just FTP wrapped in a nice UI anyone can make that". This protocollie project seems to be a documentation viewer / postman for MCP. Whic…

>Which is cool, but is it something that would have taken a competent dev months to build? Probably not.

Right...but it exists today. The days of wondering "should I spend time building this" are gone.

Re: Nobody knows how to build with AI yet

#65

Post by majormajor »

karel-3d wrote: post #24

Reading articles like this feels like being in a different reality. I don't work like this, I don't want to work like this and maybe most importantly I don't want to work with somebody who works like this. Also I am scared that any library that I am using through the myriad of dependencies is written like this. On the other hand... if I look at this as some alternate universe where I don't need to directly or indirec…

The market for utility software like this predates the internet, we used to pass them around on floppies. It was never subject to particularly high QA or scrutiny. It just has to be "adequate."

But it's never displaced the market for highly-produced, highly-planned, "central" software pieces that the utilities glue together and help you work with, etc.

The growth of that software-as-big-business has only enlarged the need for utilities, really, to integrate everything, but it's a tough space to work in - "it's hard to compete with free." One classic move is selling support, etc.

Might be tough to do non-LLM-driven software development there - the selling support for your LLM-created-products model is still viable, but if there's an increase in velocity in useful utility creation or maintenance, possibly the dev headcount needs are lower.

But does anyone know how to use LLMs to make those giant ones yet? Or to make those central core underlying libraries you mention? Doesn't seem like it. Time will tell if there's a meaningful path that is truly different from "an even higher level programming language." Even on the edges - "we outgrew the library and we have to fork it because of [features/perf/bugs]" is a pretty common pattern when working on those larger projects already, and the more specific the exact changes you need are, the less the LLM might be able to do it for you (e.g. the "it kept assuming this function existed because it exists in a lot of similar things" problem).

What I hope is that we can find good ways to leverage these for quality control and testing and validation. (Though this is the opposite of the sort of greenfield dev demos that get the most press right now.)

Testing/validation is hard and expensive enough that basically nobody does a thorough job of it right now, especially in the consumer space. It would be wonderful if we could find ways to release higher quality software without teams of thousands doing manual validation.

Re: Nobody knows how to build with AI yet

#66

Post by SoMomentary »

brabel wrote: post #58
Quitschquat wrote: post #35

Earlier quoted context omitted.

Google doesn’t have to change search. It already returns AI generated crap before anything useful.

I was a bit wary of trusting the AI summaries Google has been including in search results… but after a few checks it seems like it’s not crap at all, it’s pretty good!

I think their point is that all of the content out there is turning in to AI Slop so it won't matter if search changes because the results themselves have already been changed.

Re: Nobody knows how to build with AI yet

#67

Post by quantiq »

karel-3d wrote: post #24

Reading articles like this feels like being in a different reality. I don't work like this, I don't want to work like this and maybe most importantly I don't want to work with somebody who works like this. Also I am scared that any library that I am using through the myriad of dependencies is written like this. On the other hand... if I look at this as some alternate universe where I don't need to directly or indirec…

This has to be someone working solely on personal projects right? Because I don't know anyone who actually works like this and frequently the code that AI will spit out is actually quite bad.

Re: Nobody knows how to build with AI yet

#68

Post by aDyslecticCrow »

patcon wrote: post #48
Quitschquat wrote: post #35

Earlier quoted context omitted.

Google doesn’t have to change search. It already returns AI generated crap before anything useful.

I have systemic concerns with how Google is changing roles from "knowledge bridging" to "knowledge translating", but in terms of information: I find it very useful. You find it gives you poor information?

Always check the sources. I've personally found it;

- Using a source to claim the opposite of what the source says.

- Point to irrelevant sources.

- Use a very untrustworthy source.

- Give our sources that do not have anything to do with what it says.

- Make up additional things like any other LLM without source or internet search capability, despite reading sources.

I've specifically found Gemeni (the one Google puts at the top of searches) is hallucination-prone, and I've had far better results with other agents with search capability.

So... presenting a false or made-up answer to a person searching the web on a topic they don't understand... I'd really like to see a massive lawsuit cooked up about this when someone inevitably burns their house down or loses their life.

Re: Nobody knows how to build with AI yet

#69

Post by avidphantasm »

oblio wrote: post #56
mnky9800n wrote: post #49

Earlier quoted context omitted.

I think the agent stuff is impressive because we are giving the AI scaffold and tools and things to do. And that is why it is impressive because it has some directive. But it is obvious if you don't give it good directives it doesn't know what to do. So for me, I think a lot of jobs will be making agents do things, but a lot won't. i think its really strange that people are all so against all this stuff. it's cool ne…

People are afraid that instead of skilled craft guild members they will become assembly line workers like Charlie Chaplin in Modern Times. And in 10 years unemployed like people in the Rust Belt.

This, and no one will understand the software that is created. Then you are beholden to AI companies who can charge you whatever they want to maintain the AI code. Will this be cheaper than paying software engineers? Maybe, but I could also see it costing much more.
Post reply on HN