Nvidia RTX 5090 32Gb! Happy as an elephant. Installed ArchLinux and CUDA. Planning to soon get smart about boosting transformer deep neural networks and have a bunch of ideas for digital art based on concepts other than diffusion models.
Performance: Just ran a test, model GPT_OSS_20b_UD_Q4_K_XL generates 350 tokens per second with a context of 131072 tokens. That’s roughly an A4 page in a few seconds. Gemma3 27B – 55 tokens per second. Qwen3_30B_A3B_Q6_K – 259 tokens per second.
Cloudflare broke down. Comrade asked a very valid question
I’ve launched an electronic open source application for my book Recommender Algorithms! It’s a “sandbox” where you can “run” various recommendation algorithms with different settings, and view specific visualizations for each algorithm that help understand how it works. For instance, for algorithms like ItemKNN, SLIM, or EASE, a key visualization is a heatmap of the learned similarity matrix (item-item similarity matrix). This allows you to see which pairs of items the model considers “similar” (or “influencing” each other). For SLIM, for example, a useful “Sparsity Plot” shows that the similarity matrix indeed turned out to be sparse. For associative rule algorithms (Apriori, FP-Growth, Eclat) the visualization is not a graph, but interactive tables with found “Frequent Itemsets” and generated “Association Rules,” which can be filtered and sorted.
Additionally, there is a parametric mechanism for creating a “game dataset” — Dataset Wizard. It works like this – there are template datasets that describe items through characteristics. For example, recipes through flavors. Or movies through genres. The system generates random users with a random set of characteristics from the same set — and there are many sliders to make this distribution more contrasted or complex. Next, a matrix of user ratings of items is created – conditionally, if the characteristics of the user and the item match, then the rating will be higher because “tastes match”; conversely, if they differ, then the rating will be lower. Here too, sliders add noise and scarcity – randomly removing part of the matrix. The characteristics of products and users are not fed into the recommendation algorithm; they are hidden, but they are used to visualize the results.
The third component of the application is the tuning of hyperparameters. Essentially, it’s an auto-configurator for a specific dataset. An iterative approach is used, which is much more efficient than a full search (Grid Search) or random search (Random Search). In short, the system analyzes the history of past runs (trials) and builds a probability “map” (surrogate model) of which parameters will likely yield the best result. Then, it uses this map to smartly choose the next combination to test. This method is called Sequential Model-Based Optimization (SMBO).
The code is open source and will be further supplemented with new algorithms and new visualizations.
Link to the code in the comments.
Link to the site where the code is deployed and where you can check out the application is also in the comments.
“Hello. I am an Albanian virus, but due to the low level of technology in my country, I cannot do anything to your computer. Please kindly delete one file on your computer and then forward me to other users.”
Here’s the 2025 version. The line they ask to insert into the terminal – echo “” | base64 -d | bash
This line contains curl, pointing to 217.119.139.117 whose result is passed to `nohup bash`. And from this address, a script is loaded, of course obfuscated.
Naturally, no available LLM agrees to decrypt it. But Qwen didn’t mind.
Upon execution, the script gathers information from Chrome, Brave, Edge, Firefox, and others, extracting cookie files, autocomplete history, and system login data, collects crypto wallets like Electrum, Coinomi, Exodus, Atomic, Wasabi, Ledger Live, and others, gathers content from the “Notes” macOS app with attached media files, data from the Keychain (passwords), and also scans the desktop and documents for files of certain extensions. The collected data are archived and sent to a remote server with the IP address 217.119.139.117.
To ensure persistent access, the script creates hidden launch services (LaunchDaemons) with random names, making it difficult to detect. It can download and replace the legitimate Ledger Live application with a modified version.
Such is the Albanian virus)
I finally released a book on #RecSys! It’s called Recommender Algorithms, where I’ve compiled over 50 recommendation algorithms with detailed mathematical derivations, thorough explanations, and code examples.
https://www.testmysearch.com/books/recommender-algorithms.html
It all started early this spring in Germany, when I attended an ACM conference and sketched out the first structure of the book while analyzing the talks from the RecSys track. And now, just six months later, it has come to life.
Why did I write it? Because neither online nor in print is there a single, accessible resource that deeply explores recommendation algorithms of various types and purposes. There are articles focused on small subsets, but collecting and systematizing approaches—from foundational methods to the very latest—seems to have never been done before. I don’t know if I succeeded, but I’d love to hear your feedback.
Please like & share!
P.S. Click at READ SAMPLE to see the first 40 pages. The table of contents is there as well.
https://www.testmysearch.com/books/recommender-algorithms.html
https://www.testmysearch.com/books/recommender-algorithms.html
Finally, I have released a book! It is called Recommender Algorithms — it contains more than 50 recommendation algorithms with mathematical explanations, detailed descriptions, and code examples.
It all started early in the spring in Germany, when I attended the ACM conference and made the first sketches of the book’s structure, analyzing reports on the RecSys stream. And now, six months later, the book has been published.
Why did it appear? Because there is no single, accessible source either online or in print where the recommendation algorithms of various types and purposes are thoroughly examined. There are articles focused on narrow aspects, but to collect and systematize the developments — from fundamental to the most recent — until now, it seems, no one has managed to do it for some reason. Maybe no one needed to. Suddenly, I found I needed to. I don’t know if I succeeded, but I am eager for your feedback.
Available on Amazon and Barnes and Noble. There is a Russian automatic translation (surprisingly, but very decent), but I do not know how to sell it yet.
https://www.testmysearch.com/books/recommender-algorithms.html?FB
(This is not my only book, but today — just about this one.)
Preparing a book for publication on Solr&Lucene. What do you think about publishing such a translation on Amazon? 🙂
The book is about algorithms and under-the-hood engineering. I haven’t seen books from this angle yet, maybe someone will find it interesting.
While working on a book, I realized what kind of product I’m missing. It’s an AI diagram generator based on textual descriptions.
The idea is that the master document for the diagram is text. This textual description can be (and should be) quite detailed, so the generated diagram exactly matches the author’s vision. The diagram itself is not edited. That is, it can be edited – moving circles around, but ideally, after making changes, the system should update the text, generating from which will result in what the user adjusted.
The result — the diagram — should correspond as closely as possible to the description. If it does not match the description because, for example, it’s impossible to make a triangle with three obtuse angles, the system should do its best and provide a verbal response about what didn’t work. The user can then modify the task so that the system complies and produces the diagram correctly.
But then we understand that the author might have randomly achieved something that they liked with their flawed text. And if regenerated, it might turn out differently, and not necessarily better. Therefore —
You could ask the system to generate a diagram description from the diagram, which, if inputted back into the diagram generator, would result exactly in what the description was generated from. Yes, this description would be more verbose and complex, but it would more reliably describe the result.
So, from this point, you are no longer working with the diagram. You are working with text. If a diagram is needed — you simply compile the text into a diagram and it turns out as needed. But you don’t even work directly with the text. You work with this diagram-description text through an LLM, asking it to add some block, and the text changes, but changes in a way that everything doesn’t suddenly shift.
The final diagram should be in an object form, from which raster (PNG) or vector (SVG, EPS) images can be created.
It would also be great if such a system could take existing diagrams or diagram templates so that it could borrow styles and existing conventions on how to display what.
So, these are my fantasies. If anyone has ideas on how to implement this — let’s discuss 🙂
I decided a while ago to write a book on recommendation algorithms. With mathematics, code examples, a repository, etc. English, of course.
Accordingly, I am looking for volunteer reviewers who are knowledgeable in the field. Also those who have experience with print-on-demand on Amazon.
There’s already about 200 pages of content. About three months of work left. Working title Recommender Algorithms in 2026: A Practitioner’s Guide. Roughly half of it is still in draft form, with the first 80 pages about 80% complete.
I’ve built a mechanism to publish in HTML and PDF simultaneously. The HTML version is fully functional, with navigation. The navigation block reflects the current section, and as you scroll, it shifts to the one in front of the reader. Clicking on a section, of course, teleports you to what you clicked on. It’s all completely automatic.
