I’m specifically not looking for examples of widely-used projects which already existed before LLMs and which today are accepting LLM-assisted contributions - I know of plenty of examples of those.
I’m looking for examples of new things - both “greenfield” or “cleanroom reimplementation” projects which were (1) created using LLMs from the start and which (2) clearly have a significant number of regular users today.
I know that there are many working pieces of software which satisfy the first requirement, but i’m curious which (if any) satisfy the second yet.
After the recent “someone recreated adobe and microsoft suites with ai, omg” news cycle, many people tried some of those applications. But it isn’t clear to me if many (or any) people are actually using any of them on a regular basis yet.
So is the .ml admin staff just completely and utterly dominated by AI bros? Sheesh.
Basically every AI harness is vibecoded AFAIK. OpenClaw and Hermes for sure. Pi and OpenCode maybe less so.
bunis functionally a vibecoded rewrite of nodejs and npm.And many recent “rewritten in Rust”
Which of those are actually getting any mass adoption though? All I can think of is bun but that’s only because their vibe coded zig codebase was unmaintainable.
Claude code, apparently.
I use PhotoCraft and LightCraft for my photo processing now which replaced two commercial apps for me. I tried using GIMP and DarkTable before, and found the UX was just too clunky for me. Neither app fully replicates the functionality, but they do enough of what I use day to day, and they feel snappy.
Personally, what I’ve found to be more interesting though are the custom tools I can now easily build for myself. For example, I was able to make my own news aggregator that goes through a bunch of different news sources using rss, scraping, reading youtube transcripts. It doesn’t care what language the source is in, so I can get news from many different countries. Then it aggregates them to tell me what news are most widely reported, how stories evolve over time, find connections between them and give me a big picture overview. It took around a week to build, and it goes through hundreds of news sites each day to give me a briefing. Being able to make apps like this quickly that are customized to your exact workflow is something that just wasn’t easily doable before.
did you use something already existing? i installed miniflux and built a custom front end that suits my needs better, but your idea sounds really cool. so if you’re willing to share it i’m interested.
Oh I built my own from scratch https://github.com/yogthos/newsroom
Are there any Android apps that you’ve noticed? Especially on image and video editing?
I haven’t looked too hard to be honest I find I prefer a laptop for any sort of editing.
iOpenPod is 100% slop. That’s why the “coder” updates it every 12 hours because it’s full of bugs and they constantly have to fix the spaghetti code.
I thought it was great, as do hundreds of stars on the Hub. Frankly it’s just shiny. Strawberry and Rhythmbox do the job more reliably.
I’m partial to Quod Libet myself.
there’s a proprietary word processor that’s prevalent where I live called Hangul (*.hwp). their public HWP format documentation is terrible and until very recently there weren’t many methods to read HWP files other than using their proprietary windows/mac program. (this program is literally essential for reading government docs…)
then very recently some dev ‘vibecoded’ a web based equivalent in rust/wasm here and it works surprisingly well. occasionally a few documents are distorted but otherwise it’s great for viewing.
does this count? it has 200 issues on github, a few contributors and 2000 stars. i do use some other tools that were built with ai but many are not widely used i guess, or most are already existing projects or open source version of closed shit.
yeah, i’d say that counts… though (like all of the other answers in this thread so far, except the Hangul word processor replacement) it is also a tool largely for using AI 🤔
it has ai integrations but it’s not like the only thing it does, it’s a multiplexer first and foremost. it’s just that multiplexer started adding integrations for ai to use them too.
i see
T3 Code
The product I work on in work is over 10 years old and has a huge codebase so I don’t know if it really meets the definition of ‘LLM created’ at least not entirely so, but they’ve been pushing us use LLMs since the start of the year so a significant amount of the recent bug fixes and new features are created at least partly by them
Anyway I don’t really use it has such in the way a customer would but I do run it and play around with it when testing my stuff
Also my favourite game(JSRF) recently got a REA PC build, since running it on an emulator as frame rate issues I’ll probably use it next time I fancy playing it
The product I work on in work is over 10 years old and has a huge codebase so I don’t know if it really meets the definition of ‘LLM created’ at least not entirely so
huh? 10 years ago LLMs hadn’t been invented yet.
icymi my post says:
I’m specifically not looking for examples of widely-used projects which already existed before LLMs and which today are accepting LLM-assisted contributions - I know of plenty of examples of those.
I didn’t say they did, read what I wrote:
The product I work on in work is over 10 years old and has a huge codebase so I don’t know if it really meets the definition of ‘LLM created’ at least not entirely so, but they’ve been pushing us use LLMs since the start of the year so a significant amount of the recent bug fixes and new features are created at least partly by them
huh? re-read what i wrote. what you’re describing meets the definition of what i said i was specifically not asking for examples of.
Sure but I didn’t say LLMs were around 10 years ago lol
I would be very surprised if there is any significant piece of software recently developed (or older, but recently modified) that don’t has any AI behind, wether in design coding testing or documentation.
That was not the question. Read OP.
For that question I have an answer: Subwave. https://github.com/perminder-klair/subwave
I have used it and it’s simply genial and great. Fast paced development, super stable, love it. Just stopped using it because doesn’t fit my use case.
wow i just tuned in to their “AI DJ” demo stream and listened to the LLM-to-TTS making radio-DJ-style small talk between tracks and… i don’t know what to say besides 🤮
really disgusting stuff. it does have 1.6k stars on github for whatever that’s worth (people do buy those, and i see this project is trying to sell itself as a hosted service after all…), but i really hope not many people actually use or listen to this slop 😢
Ignore the ai dj, I was using it to generate automatic playlists from my music, with the dj turned off.
But the setup as a radio was overall overkill for my needs.
I would be very surprised if there is any significant piece of software recently developed (or older, but recently modified) that don’t has any AI behind, wether in design coding testing or documentation.
there are still a fair number of very actively-maintained projects which have zero-AI policies, but i do agree that most new software projects started recently are being created with heavy “AI assistance”.
my question is: what are some examples of them which have already gained wide adoption?
