If you want to trick an AI system your need access to it or at leat to some kind of oracle that tells you the score of your text. Then you can train your own model to achieve maximum scores. If you have access to the actual model parameters it becomes even easier.

In my opinion Julienne or Bâtonnets are the best cuts for carrots in a salad.

@awesomeest looks like devRant did crazy things to that picture. I eat quite a lot so I often overestimate what others would eat. I can hammer down on quite a large zucchini using the halve and fill technique with some rice on the side :D
The other advice would be to harvest the zucchini early. It standard to harvest them before they ripen, usually around 100-300g. If you let them fully ripen they can reach 5kg and have a wood like interior. They will last longer and can be kept for winter.

You can split the zucchini into two halves and carve the insides, then use them together with garlic, onions, sieved tomatoes and some feta cheese to make a nice filling. Add some salt, pepper and spices to your liking (I like to add some balsamic vinegar for some extra sourness and sweetness). Put the filling back into the carved zucchini halves and bake in the oven for a nice meal.
For me this has been the most efficient way to decimate zucchini supplies.

@gitstashio the notes are just normal files (.xopp). You can back them up like you would any other file.

@anux If your looking for a good note taking app for Linux I can recommend Xournal++. It has been working for me quite well and annotating PDFs is also possible.

@bosslogic Sounds very German. Probably a direct translation ;)

Your German prompts are quite amusing :)

Have been using NixOS for quite some time I like the fact that I can use my configuration on a new device and have everything configured promptly.
But beware: A lot of software relies on the classic Linux-FHS to work, so you may need to patch some stuff.

I have never heard of good winds so I can only guess.

What is a Patricia trie?

@AlmondSauce Depending on how the model works, fixing the seed might be enough to get reliable output.

It's kind of funny how bad the code on the cup is written. I am pretty sure there is no namespace keyword in Java, there is also no struct keyword. I wonder how much more issues we will find.

@awesomeest Maybe just use the same wheat farmer metaphor you used here to show him what he is potentially doing. I know nothing about his personality though, so it is up to you to gauge how best to package the message you are trying to teach.

@BlueNutterfly Is a sponsor of the university Amazon by any chance? 🤔

@keigezellig She is actually using Linux because I installed it for her!

It's garbage all the way down!

@electrineer I was thinking the same but it has a male name in it making an accident like this quite unlikely.

@electrineer That would be quite an interesting way of subtle hinting!

@Demolishun Possibly yes since they used my mail, I can do a password reset and check out everything they put in their profile. Would be pretty funny if they try and stay anonymous by using someone's mail address but then just store their own phone number with the account.

Try debugging a machine learning training script that runs for 8h and then crashes. That right there is a long time. I learned the value of checkpointing the hard way :D

@electrineer Maybe you should ask a LLM about it? ¯\_(ツ)_/¯

It's highly likely that the volume just contained something that caused the error. Have you ever tried to delete it between runs?

KDEs okular is quite good in my opinion, but I don't know about the exact functionality you want.

@Parzi You are thinking of polymorphic code I believe which is an entirely different concept.

@iSwimInTheC It largely depends on what you see as a controller.
Lets assume conventional model-view-controller.
In my general interpretation, model is where I would suspect the core logic to reside. The controller receives user input, be it via a text field in the context of a ui or as some json payload for a REST API and converts it to a structure/command that is forwarded to model which applies it.

Do be honest the most important thing in architecture generally is to draw clear boundaries between logic that is integral to the application and any external technologies used (storage (databases, disk, ...), transport (REST, gRPC, ...), ui technology (QT, GTK, ...), etc.). This way you can exchange the technologies used without having to rewrite the entire application.

@kiki What makes you believe I am defending brave? The point I wanted to make is that when it comes to deciding which browser to use, it is a lot about whom you trust or not, which in turn is based on your personal experience or the small bits of drama your heard about/stumbled over on the internet.
I don't think I can wholeheartedly recommend any browser currently on the market because neither did I study the source code nor am I a security researcher specialized in browsers.
I am a proponed of hearing both sides of the argument and forming your own opinion.

@kiki I don't know. I think it's difficult to pick a browser nowadays, especially if you are concerned with what it did do in the past. Firefox had the pocket debacle and stored (might have to check if it is still that way) a unique user identification string in the settings (pocket_id/impressionId). Brave did some things that people called shady. Chrome and edge attempt to associate your browser with an account as soon as possible. Vivaldi isn't fully open source. The list goes on. There is not perfect browser in my opinion.

This post has been cross posted on a lot of forums/subreddits already. I would recommend reading the response from brave regarding the points brought up and then make your own decision on whom you believe.
https://reddit.com/r/brave_browser/...
(this was the response on the brave subreddit where the poster deleted his own post later)
If you don't want to believe anyone than you can read the the source code and check the requests made yourself. Please feel free to report back on your findings I am really interested in any results you may have.

To be honest I think the biggest problem is the data silos. Products create a walled garden by holding your data hostage. If you want to move away from Spotify e.g. you lose all your playlists etc. This goes for a lot of big monopolies on the market. I think an important step to make here is to give users control of their data back, or rather for the consumes to demand control over their own data.

I think it would be more efficient to enforce the grammar during training via an appropriate loss. If the models output generally isn't in the ballpark of the grammar you might have to backtrack a lot. So you can not just pick any grammar it must be one that is adjacent to, or rather a subset of the one the model was trained with.