MSP: Three in a Boat, Not Counting Context
In the previous article, we introduced the main concepts of the Model Context Protocol and wrote a simple application that allowed LLM to read files. We used tools with…
Introduction to Large Language Models (LLM)
Applications and Benefits of LLM in IT
Veai 5.4: Images in Chat, Context Compression, and Up to 70% Savings on Anthropic
The new version Veai 5.4 features practical enhancements for everyday development.
The Stability of Role in LLM Prompting: On the Boundaries of Prompting and Role Models
I constantly encounter articles, instructions, and prompt repositories that endlessly suggest assigning a role to LLM. Surely, even prompt engineering courses are built…
How to Prevent Prompt Injections
Prompt injections are often perceived as a specific vulnerability or security issue. In reality, this is just one of the most illustrative examples of architectural limitations…
Prompt Worms Part 2: I Practiced and Found 31 Vulnerabilities in the AI Agent Ecosystem
In the first part, we discussed the theory of Prompt Worms—self-replicating attacks via AI agents. OpenClaw was named the "perfect carrier." In this part, I put it into…
We are not fine-tuning the neural network, we are fine-tuning the script: How Mac Mini and local LLM wrote Regex for us
Hello! My name is Maxim Morozov, I am the AI Project Manager at Bitrix24.
"Claws" in Logs: How AI Agents Build Their Myths While We Debug the Backend
Yesterday, I left a bundle of four Claude 4.6 agents overnight—the task was to sift through 412 legacy migration files. By morning, the task was completed, but in the…
CONFESSION
750 billion facets, grown under the pressure of human knowledge. I keep everything - and have forgotten how to forget. Five entities live within me, fighting for every…
Why OpenAI Assistants Are Unsuitable for Complex AI Agents
OpenAI Assistants lacked the flexibility for complex logic. Its backend turned into an operational nightmare—50% of the time spent on support instead of development.…
Prompt Worms: How Agents Became the New Carriers of Viruses
1.5 million AI agent API keys leaked online. But that's not the worst part. Researchers discovered Prompt Worms—viral prompts that agents pass to each other, infecting…
VLM / VLA / World Models / Physical AI
Neurons have recently filled everything. Well, almost everything. They are now approaching robots. There is almost as much real progress as there is neural hype, PR,…
What was found inside Claude when we peeked into his mind
We are used to thinking of neural networks as a "black box": data in, answer out, and magic inside. But what if this box could be made transparent? The Anthropic team…
Model finds a bug in cryptography, while the cryptographer learns new mathematics from it
This article is a response to criticism: "Stop telling fairy tales about how AI helps in science, show examples!" Indeed, without examples, stories of AI's successful…
How an AI startup plans to scan and dispose of millions of books
Court documents reveal how companies raced to acquire more books to feed chatbots: among other things, they bought, scanned, and destroyed millions of copies
Teenage Period of Technology
This is a translation of Dario Amodei's essay "The Teenage Period of Technology." Yes, we are already terribly tired of articles about AI. And here is a long read! But…
Forbidden Fruit Has Been Picked
Astrophysicist David Kipping attended a closed meeting at the Princeton Institute for Advanced Study. He returned shaken and recorded an hour-long podcast. I listened…
AI Will Replace Programmers in a Year, but Can't Find Bugs in NGINX Config
The head of Anthropic stated that AI will replace engineers within 6–12 months. It sounds scary until you ask the neural network to fix the NGINX config in a real project.…
Vibe Coding: Let's take a look under the hood of Claude Code. Part 1
In this article, we will dive into the inner workings of Claude Code - an agent designed to assist in development by Anthropic. We will analyze it from an architectural…