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.
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.
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
I honestly think people on here are very much overstating 'spaghetti code' when AI does it... first, it works correct? The reason I say that if the AI can continue to upgrade code that might not always look 'the prettiest' (to you..) because it doesn't care about what some of us consider to look like clean code. If it can continue to manage, document, and upgrade the code successfully.. does you calling it spaghetti code actually matter? Maybe you have a hard time reading it but the AI certainly doesn't.. I think that's worth consideration.
Every one of us has our own conventions and ideas of what 'clean' code looks like sometimes. And honestly 95% of the time we only thing it's clean if we wrote it ourselves.
>A lot of code is being written and a lot of it is going to be a real future burden.
I don't think we'll be looking at code at all within the next few years. I know I already don't anymore. The code is not the product for me, and as long as any model, commercial or local, can maintain it (or recreate it), I'm fine. When I make a drawing in a vector drawing program I don't export to PostScript to see if they calculated the bezier curves correctly either.
I'm convinced most code will be ephemeral in a few years.
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.
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.
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
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).
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.
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.
HN isn’t a monolithic bloc. A good part of the HN audience is very pro AI, another part is very skeptical. You have enough of each group to promote both sides. If HN posts could be downvoted things might bend hard one direction but with only upvotes you get a mix
I have a bad Ethernet cables through unused chimney from my apartament to the basement. The workers I asked to pull it just did it brute force and 2 of pairs are broken. When I was chatting with Claude to see if I can use it somehow to send Ethernet packets one of the Claude responses was to use broken cables as a pilot to pull fiber channel!
Good lord, the damage LLM use has done to deep thinking is amazing for such a short time. Maybe if it stayed hard to find your phone, you would have put some time into considering the root cause of the problem: you put an employer MDM on your phone. Except, well. Now it's their phone. You just pay for it.
58 comments
[ 0.60 ms ] story [ 29.6 ms ] threadI 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.
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.
So it's on par with humans, then.
I honestly think people on here are very much overstating 'spaghetti code' when AI does it... first, it works correct? The reason I say that if the AI can continue to upgrade code that might not always look 'the prettiest' (to you..) because it doesn't care about what some of us consider to look like clean code. If it can continue to manage, document, and upgrade the code successfully.. does you calling it spaghetti code actually matter? Maybe you have a hard time reading it but the AI certainly doesn't.. I think that's worth consideration.
Every one of us has our own conventions and ideas of what 'clean' code looks like sometimes. And honestly 95% of the time we only thing it's clean if we wrote it ourselves.
And also the 1980s electronic toy, “Simon”
I don't think we'll be looking at code at all within the next few years. I know I already don't anymore. The code is not the product for me, and as long as any model, commercial or local, can maintain it (or recreate it), I'm fine. When I make a drawing in a vector drawing program I don't export to PostScript to see if they calculated the bezier curves correctly either.
I'm convinced most code will be ephemeral in a few years.
Well in this case, for the idea it was already in the training data. User said was out of ideas...trying to prop up the LLM smartness...
https://www.lifewire.com/how-to-find-a-lost-bluetooth-device...
https://android.stackexchange.com/questions/194584/locating-...
https://forums.macrumors.com/threads/i-lost-my-cellphone-som...
https://discussions.apple.com/thread/3317786?
And about the code...well...
- Complete Android code example specifically explains that Bluetooth RSSI can be used, cautiously, as a distance indicator. - https://mobile.fhstp.ac.at/development/rssi-messungen-von-bl...
- android-Bluetooth-RSSI-Scanner - https://github.com/TimingJL/android-Bluetooth-RSSI-Scanner
- https://stackoverflow.com/questions/22914532/ble-continuous-...
- https://github.com/alt236/Bluetooth-LE-Library---Android
These LLMs are mainly knowledge bases and compression engines.
And we thought it would be _visual programming_ that brought children to programming. Cool. Its logic.
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.
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.
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.
But then ... posting to HN that you've used an LLM to solve a problem is OK?
two sentence horror
sigh