We tested Claude Code's Dynamic Workflows on a real project. Here's what worked and what didn't
We tested the Dynamic Workflows feature on a real project, not for hype or to judge if the feature is good or bad.
Introduction to Large Language Models (LLM)
Applications and Benefits of LLM in IT
Benchmark for Evaluating LLMs in Security Finding Triage Tasks
I created a custom benchmark to evaluate language models, as standard public tests do not answer my core question: which model performs better at triaging security findings.…
How a single jammed Xerox printer sheet led to the creation of GNU Linux and the entire Open Source philosophy
It is widely believed that the modern IT world was built by Gates, Jobs or Torvalds. But without one bearded hacker from MIT who had a bitter falling out with a Xerox…
How Adaptive RAG Works That Doesn’t Need an LLM At All
One of the most popular ways to reduce the hallucination rate of language models is the RAG method, a scheme where the model accesses external data as needed instead…
Anatomy of Claude Code. Initial Analysis and Context Filling
Everyone who programs with agents knows: before starting a task, the agent explores the project. This seems logical, natural and expected, since people do the same. It…
AI nullified benchmark and tried to blackmail engineer. And why it's solvable
Top AI models with 95% on SWE-bench show 0% and 3% on ProgramBench benchmark where tasks are specially not overlapping with the training sample.
AI bots scan even TLS certificate logs. All information is used to train LLMs.
According to Human Security's State of AI Traffic statistical report, the activity of AI agents, bots and AI tools on the internet has increased sharply over the past…
MCP Aggregator: Combine LLM Tools into a Single Server
When working with local LLMs via Claude Desktop, Kilo Code, Cursor or other MCP clients, you often run into the same situation: you need filesystem for file operations,…
Reasoning models broke my prompt engineering. I’m retraining for a year
Tuesday, 1 a.m. I fed my verified template with detailed chain-of-thought, three few-shot examples, and the “senior analyst” role into GPT-5.5, and got useless output.…
Will AI Replace Developers — and What Companies Should Do About It
AI is already reshaping the development market — but not in the way many expected. While Anthropic claims "90% AI-written code", Klarna has been forced to rehire staff…
Attacker publishes .bash_history: watch without registration and SMS
Supply Chain Security team of the expert security center sent a report to the npm registry administration about a fun little campaign against Apple, among them:
Six prompting techniques that work in 2026
Discover the latest prompting techniques for AI success in 2026, as shared by industry experts
Why Spec-Driven Development Doesn't Work Well for Microservices: Part 1. Where Context Gets Lost
I work at a large product company with a thousand microservices. In such a system, even a small feature often traverses multiple services, events, and internal contracts.
Penetration Testing 2026: How to Enter the Profession
In penetration testing, people often try to enter through a list of tools: learn Burp, run Nmap, complete a couple of labs, and wait for the first real task. In 2026,…
First architecture, then "magic": our path from scripted voice bots to smart assistants
In fintech, things almost never go according to the beautiful script usually presented in presentations: you connect LLM and suddenly get a smart, almost "human" voice…
If an agent writes code, what does a person become?
Third article out of six about the engineering process for development with AI agents. The author, Andrey Yumashev, has led development and infrastructure for many years,…
AI Agents and Code Generation
When an AI agent writes code in an empty project, it spends the first thirty to forty percent of tokens on orientation, reading README, tutorials, and API references,…
How we taught AI to select furniture based on architectural drawings
In the world of construction and interior design, the work of selecting furniture from a catalog using blueprints is still done manually: specialists spend hours flipping…