Python *

START is an open-source LLM designed for precise calculations and code verification. It addresses two major issues that most standard models face: hallucinations and errors in multi-step calculations. This article explains why these problems arise and how START solves them.

DocLing in Working with Texts, Languages, and Knowledge — an in-depth overview of the open-source DocLingtoolkit for extracting, structuring, and analyzing data from documents. The article covers approaches to processing multilingual texts, building language- and domain-specific knowledge models, and integrating DocLing into AI and NLP projects. Includes practical examples and recommendations for developers working with large volumes of unstructured data.

Would you like to know which indexes are used frequently or rarely? Which ones aren't used at all? Which tables and indexes are the largest? It's very easy to create visualizations for this. They're both visually appealing and practically useful.

We continue our journey through the world of CAPTCHAs (Fantastic CAPTCHAs and Where to Find Them, as well as Methods to Combat Them), and today we encounter yet another “tough nut” in the CAPTCHA universe – FunCaptcha (Arkose Labs).

FunCaptcha is a type of CAPTCHA developed by Arkose Labs that offers users small puzzles instead of the usual tasks like recognizing distorted text or selecting images containing buses. In traditional CAPTCHAs (e.g., reCAPTCHA), verification often relies on recognizing distorted characters or simple images. Arkose Labs took a different route: their “entertaining” CAPTCHAs feature interactive challenges with 3D objects, logic puzzles, and audio questions. This approach is intended to be user-friendly for humans while complicating life for bots.

Typical FunCaptcha challenges include:

In this tutorial, I’ll explain in simple terms what AI, AI agents, and workflows are, and then I’ll walk you through building your very first AI agent in Python using Google’s Agent Development Kit (ADK). By the end, you’ll understand the differences between these concepts and have a working content-assistant agent you can run from your terminal or a web interface.

One could write, “Experienced developers working on parsing and automation often face the need to bypass modern CAPTCHAs.” But that’s too boring… I’d rather start like this—continuing to explore the amusing world of CAPTCHAs, I finally stumbled upon the Chinese variant of protection: the GeeTest CAPTCHA. Let’s break down what this beast is, where it dwells, and why you should (or shouldn’t) fear it… You’ll understand as we go!

GeeTest is one of the advanced anti-bot systems combining user puzzles with behavioral analysis. I decided to take a close look at recognizing the GeeTest CAPTCHA under real-world conditions and figure out how to bypass GeeTest with various methods. But first—a classic introduction (which, by the way, may be more interesting than the practical part of the article, since bypassing the GeeTest CAPTCHA is already a non-trivial task for many readers).

Modern services (websites) ubiquitously deploy CAPTCHAs for protection, complicating life for developers, SEO specialists, and automation engineers (although the original aim was not to frustrate automators but rather to minimize excessive load on resources from such actors). Among the multitude of CAPTCHAs, Google reCAPTCHA Enterprise stands out as one of the most advanced bot protection systems in this line-up. It is capable of analyzing user behavior and blocking suspicious activity, thereby significantly hindering the recognition of reCAPTCHA Enterprise and its automated bypass. Let’s break down what reCAPTCHA Enterprise is, how it differs from versions v2/v3, and why it is so difficult to bypass (or is it?).

Automatic data scraping (parsing) has become an essential practice for developers, analysts, and automation specialists. It is used to extract massive amounts of information from websites—from competitors’ prices and reviews to social media content. To achieve this, numerous “scrapers” have been developed—libraries, frameworks, and cloud services that enable programmatic extraction of web data. Some solutions are designed for rapid parsing of static pages, others for bypassing complex JavaScript navigation, and yet others for retrieving data via APIs.

In this article, I will review the top scraping tools—both open source libraries and commercial SaaS/API services—and compare them according to key metrics: • Speed and scalability; • Ability to bypass anti-bot protections; • Proxy support and CAPTCHA recognition; • Quality of documentation; • Availability of APIs and other important features.

As part of my scientific and research interests, I decided to experiment with bypassing complex types of CAPTCHAs. Well, by “experiment” I mean testing the functionality and verifying that my electronic colleague can write code on my behalf. Yes, there was a lot of extra stuff—follow ethical norms, blah blah blah… But the simple fact remains: dude, I’m doing this solely as part of research, and everyone agreed.

If you are into automation and keep up with trends, you’ve probably noticed that, at some point, difficulties in hCaptcha bypass began to emerge. What kind of difficulties? Several major captcha recognition services, such as 2captcha, have removed any mention of how to bypass hCaptcha from their documentation, and the presence of thematic tweets on Twitter (along with official responses from the service) confirms my suspicion that something has happened… Let’s figure out what happened, why 2captcha no longer bypass hCaptcha, and what role solvecaptcha plays in creating new hCaptcha solver.

These days, Chinese products and services have seeped into nearly every niche. Sure, when someone mentions a Chinese development, you might chuckle and be reminded of those 90’s internet gems like “Glasses, do you need ‘em?”—and honestly, not much has changed. Even DeepSeek ended up being neither truly deep nor entirely seek. Yet, there’s something they’ve perfected, which leaves many SEO optimizers weeping salty tears while trying to bypass the Geetest CAPTCHA.

Написал лонгрид на английском о текущем состоянии открытых средств проектирования ASIC-ов. Заодно познакомил англоязычных читателей с практиками шаманов Сибири и фигурой Ивана Сусанина. Упомянул планируемые семинары в Мексике и Армении.

A text on the current state of Open-source ASIC design tools. Includes side discussions of the upcoming hackathons in Mexico and Armenia, Docker and Python, Static Timing Analysis and RISC-V, Siberian shamans and treacherous swamps in Belarus.

Any SEO expert knows the pain of collecting Google keyword data. It’s one thing if you can count all the queries on one hand, but what if they number in the thousands? How do you check the search volume in Google for each keyword? Frankly, once you hit tens of thousands of keywords, it’s enough to make your head spin. You’ll be tempted to reach for outdated, familiar tools, only to find modern reality throwing a curveball: the old formula of Key Collector + Google Ads + a few proxies simply doesn’t cut it anymore. We’re entering a new era, and without direct access to the official API, things get grim and complicated fast.

Automation can be an exhilarating, albeit exhausting, journey‑especially for those just dipping their toes into it. The tasks are often labeled as “interesting” or “non‑standard,” which, let”s be honest, often translates to “challenging” or even “impossible.” Among these challenges, one question halts around 50% of novice automators in their tracks: How to bypass CAPTCHA!

In my previous article, Puppeteer CAPTCHA bypass: Tokens or Clicks? Let’s Break It Down (which I also published on Dev.to), I compared two CAPTCHA bypass methods (clicks and tokens) using Puppeteer. I also announced that in the next article (this one), I would conduct a practical comparison of the same methods using Selenium. This will complete the CAPTCHA bypass picture, so to speak. Well, let’s not waste time and get straight to the point.

The explicit reparameterization trick is often used to train various latent variable models due to the ease of calculating gradients of continuous random variables. However, due to its peculiarities, explicit reparameterization trick is not applicable to several important continuous standard distributions, such as mixture, Gamma, Beta and Dirichlet.

An alternative method for calculating reparameterization gradients relies on implicit differentiation of cumulative distribution functions. The implicit reparameterization trick is much more expressive and applicable to a wider class of distributions

This article provides an overview of various reparameterization tricks and announces a new Python library, irt.distributions, for sampling from various distributions using the implicit reparameterization trick.

Every SEO professional has encountered CAPTCHA. If not, they are either not professionals, misunderstand the term SEO (possibly confusing it with SMM or CEO), or are very new to this challenging field.

One could endlessly deny and argue that CAPTCHA is overrated and does not deserve significant resources. But these arguments end the moment one needs data from a search engine result page, like Yandex, without knowledge of XML requests… Or, say, a client wants to scrape the entire Amazon (just because they do) and offers good pay...

How to bypass reCAPTCHA v2/v3: Automatically solve and pass Google reCAPTCHA using Python and Selenium, Puppeteer, Javascript, PHP

Are you tired of writing messy and unorganized code that leads to frustration and bugs? You can transform your code from a confusing mess into something crystal clear with a few simple changes. In this article, we'll explore key principles from the book "Clean Code" by Robert C. Martin, also known as Uncle Bob, and apply them to Python. Whether you're a web developer, software engineer, data analyst, or data scientist, these principles will help you write clean, readable, and maintainable Python code.

I came across a script on the Internet that allows you to parse product cards from Amazon. And I just needed a solution to a problem like that.

I wracked my brain while looking for a way to parse product cards from Amazon. The problem is that Amazon uses different design options for different outputs, in particular – if you need to parse the cards with the search query "bags" – the cards will be arranged vertically, as I need it, but if you take, for example, "t-shirts" – then the cards will be arranged horizontally, and in such way the script falls into an error, it works out opening the page, but does not want to scroll.