The impact of AI on the labor market: new assessment methods and initial data
This material presents the results of a study on the risks of worker displacement by artificial intelligence. If you are interested in the topic of AI agents and the…
Introduction to Large Language Models (LLM)
Applications and Benefits of LLM in IT
Local AI in Obsidian without Subscriptions: A Working Combination with Ollama, Gemma 4, and Infio Copilot
I wanted to create a local AI assistant for Obsidian that can work with my notes offline and without subscriptions. As a result, I tested several approaches and settled…
Semantic terminal: AI generator of one-liners based on natural language description
This note is inspired by the article I replaced Google with 50 lines of Python. A month later, I forgot how to write tar -xzf. At some point, after hitting Alt+Tab to…
I got banned on Claude Code for my $200 account
Yesterday, my account on Claude Code was blocked.
Security Week 2614: Supply Chain Attack on LiteLLM Library
Last week's major news was the hack of the LiteLLM library, used as an intermediary for communication with a large number of language models. Through another malicious…
One timestamp, one round-robin, one floating tools list: 7 anti-patterns that kill the LLM prefix cache
Caching is enabled, but cached_tokens still don't grow? Often the issue is neither with the model nor the provider. The hit rate is usually reduced by other factors:…
Algebra of Justice: How Engineers Digitized Courts 50 Years Before AI
Currently, Legal AI is dominated by a rather naive idea: if a large language model can already write decent legal text, then all that's left is to give it a corpus of…
How I Wanted One AI Agent but Got a Whole Village
It all started with a simple desire: for the AI agent to quietly develop my projects while I was busy with other tasks. I set the task, left, and returned to a completed…
Stop the Copy-Paste: Manifesto
You have received a notification about an unread message
How to Evaluate RAG Systems: Metrics, Methods, and What to Measure First
When a RAG system fails, it is almost impossible to tell from the final answer exactly where the error occurred. Generation with retrieval is a basic pattern in LLM applications:…
My experience with project monetization: correspondence chess with Paddle
I have been working on the Language Dove project for over a year, and it's finally time to monetize it.
How we did AI code review through Ollama without the cloud?
How to run AI code review via git diff on your machine through Ollama - without the cloud and API keys? Installation, config, and sample report.
Grep-AST or How We Replaced Vector Search with Just One Library
Hello everyone! This is Sofia from the large language model application team at ecom.tech. Today, I want to share a little-known library that, by fate, we found on GitHub,…
Distillation of LLM as a way to create your own line of AI
Recently, Anthropic was outraged by the distillation of Claude by Chinese LLM developers, with evidence including the monitoring of Claude accounts, tracking their history,…
ML Digest: Autonomous Agents, New Security Standard, and the Inference Race
While the industry debates the "bubble," discussing why the valuation of another startup with a wrapper over GPT-5 is so high, the real sector is building AI infrastructure.…
Why LLMs Continue to Generate Errors in Code - and Scaling Won't Fix It
Code from neural networks looks perfect until it hits production. That's when subtle bugs emerge: logic slightly diverges from requirements, edge cases are inadequately…
AGI: Why It Won't Happen and What Model We Can Create Instead
AGI is the next stage in AI development. Such a model would be able to learn independently, solve problems, and make new scientific discoveries, essentially serving as…
I Infected 200 Neural Networks with a Virus
Lyokha is the only biologist among my friends. We are sitting in a bar, and he is shoving his phone in my face. On the screen is a Petri dish. Bacteriophages are poured…