LLMs are a much better experience than what web search has become nowadays. And they probably are a better experience than what standard search could ever be.
But yeah, that doesn't mean they are doing much more than search.
Yesterday my 8yo prompted (using voice recognition) Claude to make a game where it would play her a song (say Twinkle) and she had to play it back and get scored. The UI was a nice piano with visual feedback. The laptop key served as the keys until I prompted for midi support so she could use the electric piano. The whole thing took about 15 minutes.
Meanwhile, one peak at the code and you can already see the state starting to become a bit of a spaghetti mess.
It’s a funny time to live through. A lot of code is being written and a lot of it is going to be a real future burden.
Reminds me of Niven/Pournelle's "The Mote in God’s Eye" where the aliens have a very pragmatic/Jury-rigged approach to technology and everything is more or less improvised.
I have noticed this with co-workers also, when you have the ability to read/write/understand things very quickly, you tend to spend much less time on making things tidy, clear and maintainable.
> This also assumes that the same application build by humans wouldn't become a "spaghetti mess".
nah, it reflects on how applications built by humans usually do become spaghetti messes with all the resulting brittleness and unintended negative side effects of changes that result
but it probably doesn't matter for a little toy piano app
And the models of the future will surely be better than they are today. This stuff is still a very long way from maturity.
Sometimes it seems like they're moving very slowly. That makes sense: It's easy to get used to how they work today and it is also easy to forget how much worse they were last year.
When we look back and realize that just 4 years ago these tools didn't really exist at all, it becomes clear that the rate of progress is rather amazing.
In 4 years, we've gone from "hah, good luck with that crap!" to "little kids writing music-learning games on their own in a few minutes"
That's pretty friggin' awesome, and it's not finished yet. :)
AI writes faster, so the rate it incurs tech debt is proportionally higher. However it's ability to have large context kept on memory compared to humans is also a key component fighting against it. These are occasionally forgotten when code quality of AI for large code bases are discussed. So Yes I think humans also build spaghetti, but as they write slower they get to the same place a lot later. However humans can't correct it, or can't correct it fast enough. AI can.
Suppose she decides from the experience that she likes making games, and wants to expand on it. She wants to support more songs, different types of song sources, colorful animated backgrounds, flashy graphics, change how it scores, a hundred other things. But by the time she gets halfway through it claude just starts getting things wrong and making them worse, and it turns into a nightmare where she doesn't even know how to go back and going back doesn't fix the problem. Or going back undoes some things she did want along with all the stuff it broke, and now she has to do it all over again. She makes a change on one screen and it changes the behavior on a dozen others. Claude starts telling her that things are impossible, or that it did this because there was a comment that said she wanted it, or coming up with other weird complicated reasons, citing random lines of code, why this or that can't work. And then she decides that yeah, making things is an awful experience and she never wants to do it again.
I think the fear is that a lot of this stuff will end up being load bearing. A lot more folks now know enough to be dangerous but not enough to know what to throw away.
I wonder. I remember before emissions, what's underneath a car hood was relatively organized and simple. Then with emissions it became a maze of vacuum hoses and so much other nonsense.
then ... in some places (maybe cars that people care about working on) it became cleaner again. In the other places, they added a second hood to hide the mess.
There is some clean code out there, like maybe the seL4 kernel:
The purpose of a car itself didn't change. But the massive inflated demand, as our city planners decided every adult must be put in a rolling metal cage to participate in society, changed the environment it was originally designed for.
Now it's a matter of geopolitical stability, or even basic human habitability of these spaces, that a car converts as much of that chemical energy into movement, and releases as little toxic byproducts in the process. Whereas before, that cost, at scale, was small enough to neglect.
Just like a modern CPU evolved into an incomprehensible mess, even though the basic consumer needs hadn't changed much, because the politics of computing forced them to run expanding institutional cruft at reasonable speeds, on battery-powered always-on addiction machines.
> then ... in some places (maybe cars that people care about working on) it became cleaner again
Those cleaner looking engine bays are usually worse to work on. Not better.
When you open up the hood and immediately see lines everywhere, that also means they’re within reach. This is great.
The engine bays that look nice and clean for the showroom still have those same lines. They’re just buried in there. If you need to work on them you’re going to be reaching underneath things, climbing under the car, or even removing other parts to access something simple.
Also, it’s not all about emissions. A lot of those lines are for modern comforts like cruise control and improvements like features that make cold starts easier or make the engine behave better at extreme temperatures. Some of those have been superseded by electronically controlled versions which is why some of those lines are disappearing on modern cars, but the overall complexity has increased further.
I think the vacuum hoses are actually for the brake booster and why you don't have any regular brakes if your engine dies. But yes, smaller engine + turbo or twin turbo is definitely more complex than a simple big block. There's also an absolute shit ton more sensors on everything now.
The job of code like this is going to serve as a makeshift spec for future coding agents, so they extract the intended use and redo it on command. Better models will be able to improve the actual code until you hit some diminishing returns for the problem you've solved.
> Meanwhile, one peak at the code and you can already see the state starting to become a bit of a spaghetti mess.
That is how compiler-generated assembly looks to humans, as well. Human-produced is typically much more readable. Yet, here we are. Most programmers only know the very basics of assembly programming, but the world keeps spinning just fine.
Comparing LLM output to compiler output is such a stale meme by now that it's surprising to see people still saying it. Obviously a deterministic translation of a higher level programming language to machine code is different than the slop cannon.
Picking over the theoretical maintainability of one-off tools and toys that were generated in minutes by what will soon be an outdated model, feels very...missing the forest for the trees, when it comes to speculating about the future impact of this stuff.
The fact that this is possible and works at all is mind blowing - even more mind blowing is that my 8 yo is growing up in a world where they can talk to a machine to produce a custom application in seconds and they don’t realise how mind blowing it is!
In terms of the code, it would take even less time than that to tidy it up. For this application you wouldn’t bother. That’s almost a form of “premature optimisation” unless you’re actually planning on doing more work on it.
My hunch is that what the world is about to see a lot of is much bigger bits of work, or changes to other bigger existing systems done by people without the skills to know how to contain the complexity. That’s going to come with a burden.
So GIMP kept crashing on me when importing 48 bit TIFFs. I threw a trace at Codex (without so much as a local GIMP repo) from which it figured out that the culprit was thumb preview handling code. It suggested turning off two non-obvious to me options that would circumvent that codepath it indeed worked.
It's not like this is an unknown gimp problem - conventional search turns up plenty of results describing this class of bugs and advising to disable the previews.
Cancelled by Strength App, Zwift for indoor bike trainer, and strava for running apps and built my own app that brings all 3 together and works better.
https://coachsamsyn.com/
Among about 100 little smaller tools.
All of these are released and work. Some of them have other customers. One of them has 1000s.
There is absolutely no way I could have done all of this without AI. I don't really consider most of it too be slop, ive been an engineer for 20 years. I care a lot about the details and the underlying properties of the system. I had shipped my side projects years before AI that had real users so I have a good idea of what it takes to ship something end to end that people can actually use. It's mostly about the details and ease of use and the last 20% is the hardest. Some of us know that from our day jobs but it is different when you have to it all yourself and be responsible for every choice.
I use rssi to find my stuff regularly. There are a ton of android apps to do it. I also use it for room presence detection with homeassistant (using the BLE in my ESP32s and Shelly switches).
Have done exactly this 15 years ago to find my lost phone (:
Also wrote a cheap 30 line shell script to allow playing videos and music to “follow me” from one room to another with well placed raspberry pi’s, 11 years ago
I used bluez and bash to lock xscreensaver using some very minimal bash. It wasn't my idea I believe people on the Gentoo forums were doing it.
But you can just loop over something like `hcitool rssi "$MAC"` and project it somewhere, there's a variety of ways.
I like using dunstify with the -p option to persist on screen.
I think it's really impressive what these agents can do, but you should also consider whether you're asking it to burn tokens reinventing the wheel for you, or making a pretty wrapper around a wrapper.
People who don't know things are usually very impressed by solutions that LLMs come up with. I saw one recently that was very impressed that their chosen LLM took screenshots of their vibe coded game to check results, when it's clear from what they were saying the LLM could've literally just read the framebuffer instead, and that's trivial to set up.
Had this song in my head, wrote some lyrics, hummed it into Suno, generated the song. I love listening to it. What a world.
I make little tools like that all the time. My GitHub is filled with these little things I write or gen once never to be used again. Some of them I use daily (GitHub.com/dexterlagan). Since LLMs became decent I make even more of these. The more we move forward, the more people make things for their own use. I think it's cool. I see a lot of naysayers, but man, I have been waiting for this kind of tech since I was 10. Let's enjoy it I say.
You can't attack other users like this, no matter how right you are or feel you are, nor how justified or strongly you feel. Preserving a container for respectful interaction here is more important, and we need commenters to contribute to that, not destroy it. If you would please review https://news.ycombinator.com/newsguidelines.html and not post like this again, we would appreciate it.
---
Edit: as your most recent comments have all been breaking the site guidelines in a similar way, I've banned this account. It's not ok to do that here.
If you don't want to be banned, you're welcome to email hn@ycombinator.com and give us reason to believe that you'll follow the rules in the future.
First, starting off with “as a <insert self aggrandising title>…” already makes you ineligible to emit an opinion on any given subject.
Second, rude.
Third, you might want to check with a professional, but it seems to me you are projecting your own (fairly obvious, and loud) insecurities onto others; I’m sorry that you feel you’re not good enough at music.
I did that a few years ago when my wife lost her fitbit in the woods. She knew the approximate area, +/- 50ft. Thick brush. Using a BT meter on the phone led me right to it like a metal detector
If this happens often for you (it does for me) I can recommend a garmin watch. It also can find the phone based on signal strength.
If the app garmin app is running in the background it can also play a sound on your phone, but the signal strength part saved me many times.
I routinely use the RSSI indications in my UniFi WiFi (thoroughly recommended) to find various kids devices around the house. You go by AP first (each section has its own), then each room tends to give its own RSSI, give or take.
Somehow, on HN, suggesting to use an LLM to solve some problem gives you negative responses.
But then ... posting to HN that you've used an LLM to solve a problem is OK?
Seems like the divide between the voters who vote in the Primaries and the voters who vote in the actual Election. Vocal minority vs. Silent majority.
OP works at a robotics company, hence the inability to use Google search and "git clone" to find dozens of ready made apps, some of them years old.
For most people talking to Claude is a much better experience than doing a Google search.
Google search as become nothing more than a shitty AI by design. Alternative search engines are the only means to actually navigate.
I can still find core material using an alternative.
Don't worry, Google is now embedding that AI into Android's core. Rinse and repeat, expect the same results.
LLMs are a much better experience than what web search has become nowadays. And they probably are a better experience than what standard search could ever be.
But yeah, that doesn't mean they are doing much more than search.
Google already has AI overviews which I use instead of polluting my ChatGPT, Claude, or Gemini chats.
Yesterday my 8yo prompted (using voice recognition) Claude to make a game where it would play her a song (say Twinkle) and she had to play it back and get scored. The UI was a nice piano with visual feedback. The laptop key served as the keys until I prompted for midi support so she could use the electric piano. The whole thing took about 15 minutes.
Meanwhile, one peak at the code and you can already see the state starting to become a bit of a spaghetti mess.
It’s a funny time to live through. A lot of code is being written and a lot of it is going to be a real future burden.
> A lot of code is being written and a lot of it is going to be a real future burden.
This assumes that the models of the future won't find it easier to just throw the code away and rebuild it
This also assumes that the same application build by humans wouldn't become a "spaghetti mess".
Interestingly, I have personally felt this way about projects in which I was involved in the past(written both by me and by others)…
sometimes starting from scratch just faster and/or easier.
Reminds me of Niven/Pournelle's "The Mote in God’s Eye" where the aliens have a very pragmatic/Jury-rigged approach to technology and everything is more or less improvised.
I have noticed this with co-workers also, when you have the ability to read/write/understand things very quickly, you tend to spend much less time on making things tidy, clear and maintainable.
> This also assumes that the same application build by humans wouldn't become a "spaghetti mess".
nah, it reflects on how applications built by humans usually do become spaghetti messes with all the resulting brittleness and unintended negative side effects of changes that result
but it probably doesn't matter for a little toy piano app
>just throw the code away and rebuild it
What about all the undocumented "adjustments" ("bug fixes" in a professional context) that were made to make it actually useful?
errors compound
And the models of the future will surely be better than they are today. This stuff is still a very long way from maturity.
Sometimes it seems like they're moving very slowly. That makes sense: It's easy to get used to how they work today and it is also easy to forget how much worse they were last year.
When we look back and realize that just 4 years ago these tools didn't really exist at all, it becomes clear that the rate of progress is rather amazing.
In 4 years, we've gone from "hah, good luck with that crap!" to "little kids writing music-learning games on their own in a few minutes"
That's pretty friggin' awesome, and it's not finished yet. :)
Eh, not really. It's the same issue with porting. You always rely on features you didn't properly articulate.
Good luck to future models figuring out which weird section of code are bugs and which are features.
AI writes faster, so the rate it incurs tech debt is proportionally higher. However it's ability to have large context kept on memory compared to humans is also a key component fighting against it. These are occasionally forgotten when code quality of AI for large code bases are discussed. So Yes I think humans also build spaghetti, but as they write slower they get to the same place a lot later. However humans can't correct it, or can't correct it fast enough. AI can.
She's 8 years old man cut her a little slack on the code aesthetics ;)
And Claude is even younger.
In what world would this be a future burden? It's just a throwaway fun project lol
Suppose she decides from the experience that she likes making games, and wants to expand on it. She wants to support more songs, different types of song sources, colorful animated backgrounds, flashy graphics, change how it scores, a hundred other things. But by the time she gets halfway through it claude just starts getting things wrong and making them worse, and it turns into a nightmare where she doesn't even know how to go back and going back doesn't fix the problem. Or going back undoes some things she did want along with all the stuff it broke, and now she has to do it all over again. She makes a change on one screen and it changes the behavior on a dozen others. Claude starts telling her that things are impossible, or that it did this because there was a comment that said she wanted it, or coming up with other weird complicated reasons, citing random lines of code, why this or that can't work. And then she decides that yeah, making things is an awful experience and she never wants to do it again.
I think the fear is that a lot of this stuff will end up being load bearing. A lot more folks now know enough to be dangerous but not enough to know what to throw away.
You mean the 8 year old shouldn't be thinking about how she will maintain this code when she's 15? This generation is lost.
I wonder. I remember before emissions, what's underneath a car hood was relatively organized and simple. Then with emissions it became a maze of vacuum hoses and so much other nonsense.
then ... in some places (maybe cars that people care about working on) it became cleaner again. In the other places, they added a second hood to hide the mess.
There is some clean code out there, like maybe the seL4 kernel:
https://github.com/seL4/seL4/
https://en.wikipedia.org/wiki/SeL4
"with an explicit goal of enabling comprehensive formal verification..." (and lots more stuff)
maybe we can still have niches like this.
That's what happens with scale...
The purpose of a car itself didn't change. But the massive inflated demand, as our city planners decided every adult must be put in a rolling metal cage to participate in society, changed the environment it was originally designed for.
Now it's a matter of geopolitical stability, or even basic human habitability of these spaces, that a car converts as much of that chemical energy into movement, and releases as little toxic byproducts in the process. Whereas before, that cost, at scale, was small enough to neglect.
Just like a modern CPU evolved into an incomprehensible mess, even though the basic consumer needs hadn't changed much, because the politics of computing forced them to run expanding institutional cruft at reasonable speeds, on battery-powered always-on addiction machines.
> then ... in some places (maybe cars that people care about working on) it became cleaner again
Those cleaner looking engine bays are usually worse to work on. Not better.
When you open up the hood and immediately see lines everywhere, that also means they’re within reach. This is great.
The engine bays that look nice and clean for the showroom still have those same lines. They’re just buried in there. If you need to work on them you’re going to be reaching underneath things, climbing under the car, or even removing other parts to access something simple.
Also, it’s not all about emissions. A lot of those lines are for modern comforts like cruise control and improvements like features that make cold starts easier or make the engine behave better at extreme temperatures. Some of those have been superseded by electronically controlled versions which is why some of those lines are disappearing on modern cars, but the overall complexity has increased further.
It was simple because it was inefficient and archaic. Reducing pollution is not "nonsense"
I think the vacuum hoses are actually for the brake booster and why you don't have any regular brakes if your engine dies. But yes, smaller engine + turbo or twin turbo is definitely more complex than a simple big block. There's also an absolute shit ton more sensors on everything now.
The job of code like this is going to serve as a makeshift spec for future coding agents, so they extract the intended use and redo it on command. Better models will be able to improve the actual code until you hit some diminishing returns for the problem you've solved.
> Meanwhile, one peak at the code and you can already see the state starting to become a bit of a spaghetti mess.
That is how compiler-generated assembly looks to humans, as well. Human-produced is typically much more readable. Yet, here we are. Most programmers only know the very basics of assembly programming, but the world keeps spinning just fine.
Comparing LLM output to compiler output is such a stale meme by now that it's surprising to see people still saying it. Obviously a deterministic translation of a higher level programming language to machine code is different than the slop cannon.
Picking over the theoretical maintainability of one-off tools and toys that were generated in minutes by what will soon be an outdated model, feels very...missing the forest for the trees, when it comes to speculating about the future impact of this stuff.
That's the kind of thing I learned actual programming for, at the same age.
Yes but it’s far better than anything anyone, let alone a small child, could make in 15 minutes
On the other hand an engineer might take a couple hours and build this in a clean way with the right prompting
These “got ‘em” ai criticism comments are getting so old
I wasn’t trying to make a “got em” comment.
The fact that this is possible and works at all is mind blowing - even more mind blowing is that my 8 yo is growing up in a world where they can talk to a machine to produce a custom application in seconds and they don’t realise how mind blowing it is!
In terms of the code, it would take even less time than that to tidy it up. For this application you wouldn’t bother. That’s almost a form of “premature optimisation” unless you’re actually planning on doing more work on it.
My hunch is that what the world is about to see a lot of is much bigger bits of work, or changes to other bigger existing systems done by people without the skills to know how to contain the complexity. That’s going to come with a burden.
The only reason it's a spaghetti mess is because it's not being prompted correctly by an expert.
So GIMP kept crashing on me when importing 48 bit TIFFs. I threw a trace at Codex (without so much as a local GIMP repo) from which it figured out that the culprit was thumb preview handling code. It suggested turning off two non-obvious to me options that would circumvent that codepath it indeed worked.
It's not like this is an unknown gimp problem - conventional search turns up plenty of results describing this class of bugs and advising to disable the previews.
While I don't think this is novel I am a believer that AI has changed computing forever in the same way the tweet implies.
I now use my own video compositor to edit videos instead of after effects https://lowkeyviewer.com/studio/
use my own ecommerce and landing page builder for my businesses. that I use in production.
https://modelpad.app/
Cancelled by Strength App, Zwift for indoor bike trainer, and strava for running apps and built my own app that brings all 3 together and works better. https://coachsamsyn.com/
Among about 100 little smaller tools.
All of these are released and work. Some of them have other customers. One of them has 1000s.
There is absolutely no way I could have done all of this without AI. I don't really consider most of it too be slop, ive been an engineer for 20 years. I care a lot about the details and the underlying properties of the system. I had shipped my side projects years before AI that had real users so I have a good idea of what it takes to ship something end to end that people can actually use. It's mostly about the details and ease of use and the last 20% is the hardest. Some of us know that from our day jobs but it is different when you have to it all yourself and be responsible for every choice.
https://xcancel.com/un1c0rnioz/status/2084686552299634805
I use rssi to find my stuff regularly. There are a ton of android apps to do it. I also use it for room presence detection with homeassistant (using the BLE in my ESP32s and Shelly switches).
Bluetoothctl in Linux.
Have done exactly this 15 years ago to find my lost phone (:
Also wrote a cheap 30 line shell script to allow playing videos and music to “follow me” from one room to another with well placed raspberry pi’s, 11 years ago
Yes you can do that.
It’s a standard feature on my Garmin watch. I can access it in a few button presses, sometimes two if it’s the last action I used.
That's because people like me were doing that and writing about it 20+ years ago. Everything from Claude comes from human minds.
Yes, technology at step N is built from technology at step N-1. Always has been.
This is the most obvious thing anyone could ever say about anything… ever.
Of course current technology is a step forward from yesterday’s. Of course what we have today is built on what came before it.
Things don’t exist in a vacuum, anyone over the age of 12 should have this figured out by now.
Agreed. I love that. I'm where I am because of billions of humans before me cutting paths. Its why I like building for the next generation.
You're replying with a trite cliché to a comment saying that this particular technology was step N-(20+) being built at step N.
This is just plagiarism. No new step exists at all.
I used bluez and bash to lock xscreensaver using some very minimal bash. It wasn't my idea I believe people on the Gentoo forums were doing it.
But you can just loop over something like `hcitool rssi "$MAC"` and project it somewhere, there's a variety of ways.
I like using dunstify with the -p option to persist on screen.
I think it's really impressive what these agents can do, but you should also consider whether you're asking it to burn tokens reinventing the wheel for you, or making a pretty wrapper around a wrapper.
People who don't know things are usually very impressed by solutions that LLMs come up with. I saw one recently that was very impressed that their chosen LLM took screenshots of their vibe coded game to check results, when it's clear from what they were saying the LLM could've literally just read the framebuffer instead, and that's trivial to set up.
Had this song in my head, wrote some lyrics, hummed it into Suno, generated the song. I love listening to it. What a world.
I make little tools like that all the time. My GitHub is filled with these little things I write or gen once never to be used again. Some of them I use daily (GitHub.com/dexterlagan). Since LLMs became decent I make even more of these. The more we move forward, the more people make things for their own use. I think it's cool. I see a lot of naysayers, but man, I have been waiting for this kind of tech since I was 10. Let's enjoy it I say.
[flagged]
You can't attack other users like this, no matter how right you are or feel you are, nor how justified or strongly you feel. Preserving a container for respectful interaction here is more important, and we need commenters to contribute to that, not destroy it. If you would please review https://news.ycombinator.com/newsguidelines.html and not post like this again, we would appreciate it.
---
Edit: as your most recent comments have all been breaking the site guidelines in a similar way, I've banned this account. It's not ok to do that here.
If you don't want to be banned, you're welcome to email hn@ycombinator.com and give us reason to believe that you'll follow the rules in the future.
I understand your frustration, but I don't agree with the sentiment that AI-training should not be allowed without compensation.
If I learn to draw, make music, build stuff, etc. I look at what others have done before me. I don't pay anyone either. Why should the AI?
First, starting off with “as a <insert self aggrandising title>…” already makes you ineligible to emit an opinion on any given subject.
Second, rude.
Third, you might want to check with a professional, but it seems to me you are projecting your own (fairly obvious, and loud) insecurities onto others; I’m sorry that you feel you’re not good enough at music.
I did that a few years ago when my wife lost her fitbit in the woods. She knew the approximate area, +/- 50ft. Thick brush. Using a BT meter on the phone led me right to it like a metal detector
If this happens often for you (it does for me) I can recommend a garmin watch. It also can find the phone based on signal strength. If the app garmin app is running in the background it can also play a sound on your phone, but the signal strength part saved me many times.
Claude is MacGyvermaxxing
I routinely use the RSSI indications in my UniFi WiFi (thoroughly recommended) to find various kids devices around the house. You go by AP first (each section has its own), then each room tends to give its own RSSI, give or take.
cool, that’s like the Phone Buddy app for Apple Watch
"Come on Rebecca, you did not lose your phone"