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

151–160 of 432 posts

Re: Nobody knows how to build with AI yet

#151

Post by fragmede »

prinny_ wrote: post #61
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…

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 pro…

Ah yes, that "is that 6 spaces or 8" in a yaml file itch that just has to be scratched. Programming has a lot of doldrums. LLMs still get stuck at places, and that's just where the new itch to scratch is. Yeah, it's not the same as code golfing an algorithm really neatly into a few lines of really expressive C++, but things change and life goes on. Programming isn't the same as when it was on punch cards either.

Re: Nobody knows how to build with AI yet

#152

Post by fragmede »

beefnugs wrote: post #118
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.

This doesn't make aaaaany sense: IF this actually worked, then why would all the biggest companies in the world be firing people? They would be forcing them all to DO THE TIDE and multiple their 10 billion dollar dominance to 100 billion dollar or more dominance. The truth is something like: for this to work, there is huge requirements in tooling/infrastructure/security/simulation/refinement/optimization/cost-saving…

> why would all the biggest companies in the world be firing people

Because of section 174, now hopefully repealed. Money makes the world go round, and the money people talk to the people with firing authority.

Re: Nobody knows how to build with AI yet

#153

Post by fragmede »

tempodox wrote: post #135
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…

It does sound horrible. No more getting in the flow, no more thinking about anything, no more understanding anything. Just touch it with a ten-foot pole every few hours, then get distracted again. I guess if all you do is write React To-Do apps all day, it might even work for a bit.

Unfortunately, I think the evolution of LLMs is going to put more areas of programming within this "React Todo app" envelope of capability that you suggest, and to have it work for longer, rather than going away.

Re: Nobody knows how to build with AI yet

#154

Post by exographicskip »

bloppe wrote: post #62
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.

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

Duplicate logic is definitely a thing. That and littering comments all over the place.

Worth it to me as I can fix all the above after the fact.

Just annoying haha

Re: Nobody knows how to build with AI yet

#155

Post by exographicskip »

esafak wrote: post #124
bloppe wrote: post #62

Earlier quoted context omitted.

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

Create and point them to an agent.md file http://agent.md/

[dead]

Re: Nobody knows how to build with AI yet

#156

Post by vishvananda »

bloppe wrote: post #62
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.

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

I'm really baffled why the coding interfaces have not implemented a locking feature for some code. It seems like an obvious feature to be able to select a section of your code and tell the agent not to modify it. This could remove a whole class of problems where the agent tries to change tests to match the code or removes key functionality.

One could even imagine going a step further and having a confidence level associated with different parts of the code, that would help the LLM concentrate changes on the areas that you're less sure about.

Re: Nobody knows how to build with AI yet

#157

Post by fragmede »

stillsut wrote: post #134
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…

> Just keep it away from me I'm reminded of teaching bootcamp software engineering, when every day #1 we go through simple git workflows and it seems very intimidating to students and they don't understand the value. Which fair enough because git has a steep learning curve and you need to use it practically to start picking it up. I think this might be analogous to the shift going on with ai-generated and agent-gener…

To the article's point, I built my own version of your agro tool that I use to manage my own git worktrees. Even if I had known about your project, I still would have built my own, because if I build it (with LLM assistance, obvs) then I get to design it for myself.

Looking at other industries, music production is probably the one to look at. What was once the purview of record labels with recording studios that cost a million dollars to outfit, is now a used MacBook and, like, $1,000 of hardware/software. The music industry has changed, dramatically, as a result of the march of technology, and thus so will software. So writing software will go the way of the musician. What used to be a middle class job as a trumpet player in NYC before the advent of records, is now only a hobby except for the truely elite level practicioners.

Re: Nobody knows how to build with AI yet

#158

Post by ModernMech »

johannes1234321 wrote: post #103
prinny_ wrote: post #61

Earlier quoted context omitted.

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 pro…

There is code which is interesting to write, even if it isn't the area with clever algorithms or big architecture decisions or something. But there is also the area of boilerplate, where non-LLM-AI-based IDEs for a few decades already help a lot with templates and "smart" completion. Current AI systems widen that area. The trouble with AI is when you are reaching the boundary of its capabilities. The trivial stuff it…

> But there is also the area of boilerplate, where non-LLM-AI-based IDEs for a few decades already help a lot with templates and "smart" completion.

The thing for me is that AI writing the boilerplate feels like the brute force solution, compared to investing in better language and tooling design that may obviate the need for such boilerplate in the first place.

Re: Nobody knows how to build with AI yet

#159

Post by aDyslecticCrow »

JSteph22 wrote: post #100
aDyslecticCrow wrote: post #50

Earlier quoted context omitted.

As search gives the answer rather than the path to it, the job of finding things out properly and writing it down for others is lost. If we let that be lost, then we will all be lost. If we cannot find a way to redirect income from AI back to the creators of the information they rehash (such as good and honest journalism), a critical load-bearing pillar of democratic society will collapse. The news industry has been…

I am looking forward to the "news industry" breathing its last breath. They're the ones primarily responsible for the distrust and division.

No, i fully disagree.

The economic viability to do proper journalism was already destroyed by the ad supported click and attention based internet. (and particular the way people consume news through algorithmic social media)

I believe most independent news sites have been economically forced into sensationalism and extremism to survive. Its not what they wilfully created.

Personally, i find that any news organisations that is still somewhat reputable have source of income beyond page visits and ads; Be it a senior demorgaphic that still subscribe to the paper, loyal reader base that pay for the paywall, or government sponsoring its existence as public service.

Now what if you cut out the last piece of income journalists rely on to stay afloat? We simply fire the humans and tell an AI to summarise the other articles instead, and phrase it how people want to hear it.

And thats a frightening world.

Re: Nobody knows how to build with AI yet

#160

Post by fragmede »

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…

> and maybe most importantly I don't want to work with somebody who works like this.

Which, of course, is your perogative, but in what other ways do we, as fellow programmers, judge software libraries and dependencies so harshly? As a Vim user, do I care that Django was written with a lot of emacs? Or that Linus used emacs to write git? Or maybe being judgemental about programming languages; ugh, that's "just" a scripting language, it's not "real" programming unless you use a magnet up against a hard drive to program in ones and zeros. As a user, do I care that Calibre is written in Python, and not something "better"? Or that curl is written in good ole C. Or how about being opinionated as to whether or not the programmer used GDB or printf debugging to make the library?

Post reply on HN