You are browsing as a guest. Sign up (or log in) to start making projects!

skue

@skue

Joined August 17th, 2026

  • 19Devlogs
  • 4Projects
  • 3Ships
  • 54Votes
Random guy that tries random things.

Currently making parole , a tui tool that show the lyrics of what you're listening to
ALso making Noomo , a game about guessing what an LLM could say
Ship

Hiii !
Thanks for reviewing my shiiip, hope you have a great time !

I present to you…

Noomo !

Noomo is a game where you need to guess words that a LLM generates

In game, you need to guess the next words to the sentence you inputted. I let you figure yourself how to do that, because I hope my UX is good enough to guide you. If not, i guess you can put a bad rating.

In this ship (my second one), I added:

  • Daily challenges: I wanted to add them in my first ship, but end up not adding them. Now they’re here, and the current file should last until 2027 !
  • Better support for other scripts: First version forced the model to use latin script. Now, it automatically chooses the right script from the text inputted. Which means it should support things like Japaneses, but I can’t check since I don’t speak Japanese (sadly).
  • Some cleanup on the codebase
  • Game settings: now you can change the settings of the game (the number of words and clues) !
  • Fixed some bugs like the loading screen not showing because of dumb mistakes

Here you go, hope you like it !

  • 3 devlogs
  • 11h
  • 17.05x multiplier
  • 179 Stardust
Try project → See source code →
Open comments for this post

3h 26m 2s logged

Devlog 8 - And probably the last one !

Stardance is ending soon, so I need to ship :yay:
In this devlog, I added a better support for other languages (it should work in Japanese, but idk I don’t speak Japanese). It still sucks in French and idk why but it’s okay.
I also did a lot of debug (basically I ran an AI thing on my codebase then I fixed every bug found, nothing much). It should feel better tho !
I also populated the challenges, there is enough challenges to last until 2027 :)

0
0
8
Open comments for this post

4h 2m 11s logged

Devlog 7

It’s been a long time since the last one…
I added settings, so you can decide of the number of words and clues ! I also linted and made some cool helper scripts

0
0
8
Ship Changes requested

3… 2… 1… Introducing…

Parole !

When I was a bit younger, I installed Linux and decided to rice it. I loved Cava, and I had the wonderful idea to make an app that could display the current lyrics in my terminal !
Unfortunately, I did not have the competences to make this back then, so I foolishly delegated it to AI… And I never had the courage to rewrite it myself, because it looked sooo scary and hard… And the ai spaghetti code wasn’t here to help me.
But I discovered Stardance, and I thought “Hell yeah, why not trying this ?”. Turns out it was easier than I thought !

Parole is a TUI app to display the lyrics of the current song, compatible with Linux (using mpris) and Windows (using winsdk) !
It fetches the lyrics data from LRCLib and displays them with Textual.

It may not work perfectly on Windows, because Microsoft decided that yes, it’s a good idea to not return the actual position in the song but rather a position that updates not every time + the moment it was updated. So in most apps, if you play/pause, you break it. That kind of choices is a good reason to switch to Linux, if you want my opinion.

Anyway, have fun playing around with it !

  • 5 devlogs
  • 11h
Try project → See source code →
Open comments for this post

2h 16m 51s logged

Devlog 5 - Yayyy i’m going to ship !

This took way much more time than I thought, but everything is set up !
I made a readme, a contributing guide, and even issue templates !
Also I made a release workflow, and I’m soooo tired of the fact that Github Actions has no way of testing them, because it keeeps failing for whatever reason aaaaaah

Well now it’s published so I just need to test !

0
0
15
Open comments for this post

1h 54m 20s logged

Devlog 4 - I haaaate windows omg :microslop:

So I basically rewrote the logic to get song info. It’s not using playerctl anymore, which was not made to work like that and was not multi platform.
Now, I have a little thingy that can provide two “providers”, one for Linux using mpris (dbus-next) and one for Windows using winsdk.
The implementation on Linux was pretty easy, but omg why do people still use Windows ?
Basically, the Windows API do not return the current position on the song, but a position and a time since it was last updated. So in 90% of the players, if you pause, all the timing will get wrong.
And to complement all of this, the experience of using windows is sooo bad, like, wdym the middle mouse button stops working randomly ?
Anyway I think i just need to write docs and I’m ready to ship !

0
0
30
Open comments for this post

2h 37m 25s logged

Devlog 3 - It’s working well !

The program can show multiple lines of lyrics, and there is an argument to choose the number of them !
I also improved the lyrics parsing method, and forced the app to use the ANSI colors from the terminal.

I think now I will change the metadata fetching to MPRIS via Dbus instead of playerctl, and will do that in a separate file for better organization when I will port to windows

0
0
24
Open comments for this post

55m 50s logged

Delvlog 2 - Well, it’s easier than I thought

I knew it would be a small project, but this is reaaaally fast !

In this small update, I added a way to actually parse the lyrics, and show them in the TUI ! For now, there is just one line of lyric, but I will add options to add more lines.

I just need to add customization options, fix some errors, maybe find a way to make this run on windows !

0
0
18
Open comments for this post

2h 49m 8s logged

Devlog 1 - Parole, a nice terminal thing for showing lyrics of your favorite song !

Parole is a TUI tool made for your rice ! It will show the lyrics of the song you are currently listening, using Playerctl :)

I use Subprocess to get the playerctl metadata, then I use Textual to show everything.
For now, it can just get the lyrics and format them in a dict. Not very beautiful for now…

0
0
18
Open comments for this post

3h 3m 15s logged

Devlog 6 - Yayyyy thanks reviewers :yay:

Thanks for the reviews ! I’m already starting to try to improve all I can improve !

This is a small update, but here is the changelog !

  • I fixed the loading screen (because it was not appearing)
  • I created a ROADMAP for the v2 (check it out at https://github.com/skueee/noomo/blob/main/V2ROADMAP.md)
  • I changed a bit the layout
  • I added daily challenges ! Because hum I originally wanted them but for the first ship I didn’t implemented it so hum they are here now (kinda)

If you have any suggestion, comment or contact me on Slack !
These changes aren’t deployed yet, but it will be coming in the next few days

0
0
20
Ship

Presenting Noomo !

Noomo is a game about guessing words generated by an LLM
You just input a sentence, then you guess the words that the LLM generated.

And how does it work ?

The backend : The backend is a cool Python FastAPI server running a model through Transformers ! It generates the 100 top probabilities, deletes the ones that are not usable in the game (non-latin things, duplicates…) and sends the number of words requested !

The frontend : The frontend is a nice NextJS app that will kindly ask the backend the words for the sentence you inputed and will force you to guess words until you die (or when you guessed all the words). Then, if you didn’t die before the end, it will store the stats into the sessionStorage and show them nicely to you as a trophy.

But why is it so cooool ?

Well, first, it runs on Hackclub Nest ! Which means “only” 2gb of RAM, and 2 cores of CPU.
But how ? Well, it’s just using a smoll model, Qwen 2.5 0.5B, which makes it take 600M of RAM on idle.
(and no, i did nooot nuked my server 3 times, that was absolutely noooot me…)

Second, I learned everything while working on the project !
I knew how to code in Python, but nothing about ML. So uuum I needed to learn from kinda scratch (ngl, it’s not thaaaat hard for the thing I made).
I also knew a bit of Javascript, but NextJS is… Something else ? This was pretty hard, and why the heck can’t I change a variable in my useEffect ?
All of this means more bugs ! Yayyyyyyyyyyyyy !!!

You hyped me, where can I try ? And how ?

You can just go to noomo.dino.icu !
Enter your sentence, click enter (or the kool button), and then play !
Ah and yes, to play, you just need to input your word and press enter. And hope it was correct.


Alright I’m out, good luck !

  • 5 devlogs
  • 24h
  • 19.73x multiplier
  • 469 Stardust
Try project → See source code →
Open comments for this post

6h 39m 35s logged

Devlog 5 - FINNAAAAL ONE :ultrafastparrot: !!!!

:stardance_star_purple: Yipeee :stardance_star_purple:

I almost finished everything ! I added clues, refined the generation system, and hosted on Vercel (I also got a lot of bugs, but without bugs you cant make a good project).

I alsooo learned to make responsive things, and it looks okay if you dont try it on a very smol phone (I hope so, at least).

Now I just do some testiiing and it’s time to ship :stardance-streak-10d:

Try the website out at noomo.dino.icu :yay:

0
0
14
Open comments for this post

7h 14m 22s logged

Devlog 4 !!!!

I almost finished everything for the game !

I made most of the website. There is now a Game page (like, the most important thing of the whole project) and a Win page ! I also learned a lot about TS and HTML while making this (because this is kinda my first website).

It’s almost ready, I just need to do some adjustments to the layout of the website, rebalance the game and fix some generation bugs

2
0
58
Open comments for this post

4h 53m 35s logged

DEVLOG 3 !

I did a looot of things.

The backy backend :
I configured my Nest server (and I broke it 3 times) ! Now, it should work, but sorry to the Nest team for this :(. I also got some problems with the dockerimage (like, 19gb for a single image ?).
I improved the AI too ! It will now lean towards longer words, to make it easier to guess, and will exclude non-latin characters (sorry japanese people, but I had to).
I also learned sql (ngl i feel like a dumbass for not understanding this) and made a db ! It will store most popular sentences and maybe daily challenges ?

The frontennnnd :
Huuummm I didn’t do much but there is a header on top now ig

0
0
51
Open comments for this post

3h 18m 31s logged

[DEVLOG 2] I made the main page ! The buttons dont do anything now, but it’s my first time using HTML… I also done some fixes on the API and I got a Nest server ! Already tested it a bit and it should work if I only host the api on it. Anyway there is still a lot of work to do

0
0
12
Open comments for this post

1h 40m 32s logged

[DEVLOG #1] I’m building Noomo, a game about predicting what an LLM could say. You input a simple sentence, then you need to tell the most “logical” words, generated by a really small llm (currently, Qwen2.5-0.5B). For now, I got a small API backend that can be called to generate the predictions, and it’s running on my machine. I still need to finetune the model so that the game is more interesting, then I need to develop the frontend and host all of that !

0
0
13
Open comments for this post

1h 24m 1s logged

DEVLOG 6 - I added OpenAI detection to my AI watermark detector library ! So uhm I was tired with Videoseal, so I just decided to do the openai implementation instead. It was pretty simple, and I tried to do it without using the OpenAI sdk because i didn’t wanted to add yet another dependency. I also added a config to pass to the detector, and a way to ignore a detector.

0
0
23
Open comments for this post

4h 23m 1s logged

DEVLOG 5 - It’s been a looong time, it’s so hard… I tried to implement meta videoseal, and after countless of tries, I managed to export the detector to an ONNX model, that I could use in the library. The only problem is that I don’t know a lot about machine learning, so it doesn’t work at all. The JS script made to use the model just gives random results, often giving more score to non-ai images than to watermarked images. If anyone knows how to help with this, I would love a biiit of help. The Github repo is opened to issues and all the code is Open Source :). (Also I did the Jame Gam it was really fun :3)

0
0
14
Open comments for this post

1h 28m 55s logged

Devlog 4 - Very tiny update :). I don’t have my main computer, so I can’t do heavy things, which is what I need to do right now. But I used Blob objects instead of paths so that the library is compatible with bot Node and web browsers. I also tried some things with the openai detector, but the public api needs an API key and the private one has a turnstile challenge to solve. If anyone has tips to solve that, I would be really happy to hear them ! I also experimented with pytorch export to ONNX, but I don’t have the right pc right now (not committed yet). So yeah very smoll update, thanks for reading till the end.

1
0
41
Open comments for this post

1h 17m 5s logged

DEVLOG 3 - I made a logo for my AI watermark detection tool ! Also, I fixed a bug that made C2PA flag human made c2pa content as ai content.

3
0
15
Loading more…

Followers

Loading…