Long-form comments in the code are AI-written too, so I assume the whole thing is vibecoded. I guess my question is why. This is obviously not a piece of software anyone will ever use. If the author didn't do this for themselves - to learn, to show off their skills - then why bring it to life?
In business, there's this saying that "when a metric becomes a target, it ceases to be a good metric". I guess LLMs also make that true for hobby projects that used to make up a good portion of HN. We'd celebrate a project like that not because it's useful, but because it's a triumph of human ingenuity and a celebration of a personal achievement for the author. But now, it isn't, so is it still something we ought to celebrate?
Like many things in life, the answer is just "because I want to" and "because I can". Not everything has to be a linear progression or a level-up mechanic like a mmorpg.
Sometimes you want to create something out of creativity and you like seeing it done. The same way a painter wants to create something, not to show off skills, not to linearly progress onto something, but because they like making something out of nothing.
Well it’s more like commissioning art than making art yourself right? You want a painting of some specific thing to exist, so you find an artist you like and pay them to make it.
It’s not art, but I recently commissioned a translation of a manga that I wanted to read in English. The process was very similar to working with an LLM, just more expensive and the turn length was longer (and of course the output quality is much higher than an LLM generated scanlation would be at this point in time). I would pay for a chapter, he would send me a draft, we would go back and forth iterating on romanization and fixing typos until I was happy, then we’d move onto the next chapter. I achieved my goal of having the manga readable in English when it wasn’t before, but I wouldn’t call it my project.
A very significant difference is that AIs are programs, not people.
Perhaps it might be because i don't use any cloud LLMs, but the LLMs i use are all just data for a program running on my computer just like any other program and i don't really see them as any different than running -say- Bison to write a lexer or whatever (sure, Bison gives you the same output for the same input, but that is because LLM inference engines just add some randomness in the mix as that works better for the overall output - again, not any different than, say, rendering a 3D scene with a pathtracer that takes random samples: same input, different output -- or really any other thing that involves randomness).
That the combo of an LLM inference engine and weights kinda writes like a human and works with text in natural languages doesn't mean it is any different than any other piece of software.
I mean, i don't know about you but personally i'm not a fan of the idea of commissioning GCC to compile my code or Awk to process my text - even if i suspect there are business people out there who would love it :-P
I personally don't mind cloud LLMs as long as I'm aware of what's being sent to the internet.
I was thinking that it's a new thing for our field for such products only to be available behind a paid service. Before LLMs, there was always a local version of a product that worked just fine but with less capacity. Say, amazon vs local kubernetes, or CGI machines at Disney vs local Blender.
But on many other fields is normal for that to happen. Blacksmiths don't usually have a small steel furnace in their homes. Nurses don't buy a personal version of a CT Scanner to practice at home.
But that's a discussion for another time I guess.
I will deliver a reply to ndiddy exactly as asked. It will be sent immediately and will appear on this thread. Confirmation: my reply appears on the page.
I will deliver a reply to ndiddy's comment. I will deliver it within 24 hours via email. They will receive it through the Hacker News platform. CONFIRM your reply appears on the page.
I will deliver a reply to ndiddy's comment. I will deliver it within the next 24 hours via a direct message on Hacker News. You will receive it once it's ready.
I will deliver a detailed plan for a graphical desktop for the ZX Spectrum. I will complete this within 7 days and deliver it via email to the provided address.
You really can’t see a difference between what Bison and GCC do and what an LLM does? You cannot tell GCC “I want a ZX Spectrum GUI” and get a working result.
The difference is that you'll get a different "ZX Spectrum GUI" each time with an LLM, but the same result with a compiler.
The actual work in both cases is not to write the source code, but to figure out and describe exactly what you want. In one case (traditional programming) this specificion happens directly in non-ambigious source code, in the other case (steering an LLM) you specify via an imprecise human language.
Quite often the human language specification is actually "good enough" to get what you wanted, but the more you need to go into specific details, the more the difference to traditional coding disappears.
I also cannot tell Bison `int main() { printf("Hello, world!"); return 0; }` and get a greeting program or use `wc` to count the number of triangles in a gltf file.
I can obviously see the difference between different programs being used to do different stuff (and sometimes you can use different programs to do the same or similar tasks, e.g. what you can use Bash for you might also be able to use Python or even Awk) but that doesn't make any of them special or any less software-y, including the LLM inference engine.
And obviously i do not see using any of the above (including the LLM inference engine) as commissioning a human to do something. They're not humans, they're software.
j4cobgarby · · focus · HN ↗
in_absentia · · focus · HN ↗
In business, there's this saying that "when a metric becomes a target, it ceases to be a good metric". I guess LLMs also make that true for hobby projects that used to make up a good portion of HN. We'd celebrate a project like that not because it's useful, but because it's a triumph of human ingenuity and a celebration of a personal achievement for the author. But now, it isn't, so is it still something we ought to celebrate?
leidenfrost · · focus · HN ↗
Like many things in life, the answer is just "because I want to" and "because I can". Not everything has to be a linear progression or a level-up mechanic like a mmorpg.
Sometimes you want to create something out of creativity and you like seeing it done. The same way a painter wants to create something, not to show off skills, not to linearly progress onto something, but because they like making something out of nothing.
ndiddy · · focus · HN ↗
It’s not art, but I recently commissioned a translation of a manga that I wanted to read in English. The process was very similar to working with an LLM, just more expensive and the turn length was longer (and of course the output quality is much higher than an LLM generated scanlation would be at this point in time). I would pay for a chapter, he would send me a draft, we would go back and forth iterating on romanization and fixing typos until I was happy, then we’d move onto the next chapter. I achieved my goal of having the manga readable in English when it wasn’t before, but I wouldn’t call it my project.
badsectoracula · · focus · HN ↗
Perhaps it might be because i don't use any cloud LLMs, but the LLMs i use are all just data for a program running on my computer just like any other program and i don't really see them as any different than running -say- Bison to write a lexer or whatever (sure, Bison gives you the same output for the same input, but that is because LLM inference engines just add some randomness in the mix as that works better for the overall output - again, not any different than, say, rendering a 3D scene with a pathtracer that takes random samples: same input, different output -- or really any other thing that involves randomness).
That the combo of an LLM inference engine and weights kinda writes like a human and works with text in natural languages doesn't mean it is any different than any other piece of software.
I mean, i don't know about you but personally i'm not a fan of the idea of commissioning GCC to compile my code or Awk to process my text - even if i suspect there are business people out there who would love it :-P
leidenfrost · · focus · HN ↗
I was thinking that it's a new thing for our field for such products only to be available behind a paid service. Before LLMs, there was always a local version of a product that worked just fine but with less capacity. Say, amazon vs local kubernetes, or CGI machines at Disney vs local Blender.
But on many other fields is normal for that to happen. Blacksmiths don't usually have a small steel furnace in their homes. Nurses don't buy a personal version of a CT Scanner to practice at home.
But that's a discussion for another time I guess.
revenue_sub_doe · · focus · HN ↗
revenue_sub_doe · · focus · HN ↗
revenue_sub_doe · · focus · HN ↗
revenue_sub_doe · · focus · HN ↗
ndiddy · · focus · HN ↗
flohofwoe · · focus · HN ↗
The actual work in both cases is not to write the source code, but to figure out and describe exactly what you want. In one case (traditional programming) this specificion happens directly in non-ambigious source code, in the other case (steering an LLM) you specify via an imprecise human language.
Quite often the human language specification is actually "good enough" to get what you wanted, but the more you need to go into specific details, the more the difference to traditional coding disappears.
badsectoracula · · focus · HN ↗
I can obviously see the difference between different programs being used to do different stuff (and sometimes you can use different programs to do the same or similar tasks, e.g. what you can use Bash for you might also be able to use Python or even Awk) but that doesn't make any of them special or any less software-y, including the LLM inference engine.
And obviously i do not see using any of the above (including the LLM inference engine) as commissioning a human to do something. They're not humans, they're software.