Habr AI
AI news source. Articles are auto-selected and adapted by Hamidun News editors.
Latest publications
Developer Built a Local Spam Filter with ruBert-base-antispam Model Without VPN
An author on Habr assembled a multi-layer spam filter: after Yandex, Mail, and SpamAssassin, emails are checked by a local BERT model with 177 million parameters.
Neural Networks in Solo Indie Game Development: Help or Hindrance in 2026?
An indie game developer working solo discusses what new AI-powered tools have become available to solo developers and what awaits the indie industry.
Computer Vision Instead of Operators: How PPE Compliance is Checked on Ship Decks Via Cameras
A customer automated monitoring of helmet, life jacket, and flotation suit compliance by crew on a ship's open deck through a surveillance camera system.
AI Agency: How Autonomous Agents Differ From Regular Programs or Scripts
Understanding what agency in AI means and why autonomous agents capable of solving tasks independently fundamentally differ from deterministic scripts.
Who invented the word 'video analytics': Speclab company points to first mention in 2002
Speclab founder Oleg Grishanin said the term 'video analytics' first appeared on the company's website from Ivanovo on October 31, 2002.
AGIMA held a corporate AI hackathon for the RUNIT festival on July 5, 2026
Digital agency AGIMA has been conducting an AI hackathon for over a month in support of the RUNIT sports festival, which will take place on July 5, 2026 in Meshchersky Park.
Habr showed a neural archiver — data compression through inference of any neural network
An author on Habr described the history of archiving and presented a neural archiver — a data compression tool that uses inference of an arbitrary neural network instead of classical algorithms.
Cloud.ru designer told how AI tools outpaced Figma in interface design
Cloud.ru product designer showed on a real project how AI tools surpassed Figma in speed and where manual edits are still needed.
Habr: agent memory removes technical noise from hybrid development team chats
Habr author showed how a team transfers requirements, solutions, and defects from human chats to a separate layer of agent memory for collaboration between humans and AI.
MTS AI opened the AVI code — an external filter that keeps LLM within the law without retraining
MTS AI posted on GitHub a working AVI service — a model-agnostic filter that checks LLM requests and responses for legal violations, toxicity, and ethics.
Stanislav Lvovskiy: AI is the most important cultural event since the invention of writing
Writer and researcher Stanislav Lvovskiy delivered a lecture on AI at London's Idiot Books bookstore and wrote an essay 'They Speak' about its impact on culture.
LLM agents and SOC: developers built a cyber threat database using RAG
Developers described on Habr the architecture of a cyber threat database for LLM agents and SOC systems: PDF documents, CVE records and security articles are combined into a single RAG pipeline with HTTP API search.
Did GitHub become more uniform after ChatGPT? Code study for 2019–2025
A researcher examined whether open-source code on GitHub became more uniform after the mass adoption of ChatGPT and AI development tools.
Python in AI: why the language is 70 times slower than C and losing ground on edge
The AI industry primarily runs on Python, which is 70 times slower than C and ranks 26th out of 27 in energy efficiency among programming languages.
YuMoney Built RAG Platform and AI Assistant Eugene for Corporate Wiki Search
YuMoney explained how it combined corporate knowledge with large language models and created Eugene RAG assistant, which understands query intent rather than matching keywords.
Developer Optimized ML Library lancetnic for Training on 16 GB RAM
The author of open-source library lancetnic found reasons why training on 25 thousand text rows overwhelmed 16 GB of RAM.
Prompt Warrior: Skill Transforms Claude Code Logs Into RPG Character Sheet
Open-source Prompt Warrior skill transforms Claude Code session logs into gamified psychological portrait with achievements and biography.
AlpinaGPT Implemented Multi-Agent System Evolver and Cut Code Review to 15 Minutes
CTO Sergey Andriyanov explained how the multi-agent Evolver system accelerated code review from two days to fifteen minutes and freed up about 100 hours of team work per month.
How Context Engineering Makes Weak Local LLM a Reliable Agent
Author on Habr showed: for local LLM on closed loop, agent quality is determined not by model size but by how its context is structured.
Same LLM Benchmark Yielded 28% and 76% — Difference Is Only in How Questions Are Asked
On an LLM benchmark for corporate documents, the same model showed 28% to 76% — the variation is explained by how the queries are formulated, not by the model choice.
Engineer built a local RAG service for CAD API documentation on Habr
Habr describes a local RAG service for CAD API documentation — a way to automate routine design operations without transmitting data to external cloud services.
Kontur engineer compared D&D wizards and warlocks to the LLM era on Habr
Kontur's senior QA engineer Alexander Ovchinnikov published an essay on Habr where the shift from wizards to warlocks in D&D is a metaphor for how LLMs are changing professionals' work.
Minister of Science Valery Falkov: AI devalues student thesis work
Minister of Science and Higher Education Valery Falkov stated that due to AI development, coursework and thesis work no longer reflect the actual level of student preparation.
AGIMA held a corporate AI hackathon in support of RUNIT sports festival
AGIMA employees spent a whole month developing online games at a corporate AI hackathon timed to the RUNIT festival, which took place on July 5, 2026.
Developer replaced repeated context explanation for neural network with LLM Wiki workflow
Habr author described LLM Wiki workflow: the model not only answers questions about documents but gradually builds a knowledge base in Obsidian through Codex.
How agentic AI is reshaping team structures: analyzing AWS talks and thousands of publications
A Habr author analyzed thousands of publications on how top companies are reorganizing teams and org structure around agentic AI following AWS's keynote.
An AI agent is an LLM in a loop in 20 lines of code: where complexity begins
A Habr author explains how a real AI agent differs from a prompt and chatbot, showing that at the core of an agent is an LLM in a loop just 20 lines of code.
PWN AI expert Artyom Semenov: how AI agents delete files and compromise themselves
PWN AI Telegram channel author Artyom Semenov told Habr how the attack surface on AI agents and MCP protocols has grown and why system prompts are not a security boundary.
AI code generation quietly erases architectural understanding in development teams
A Habr author explains why massive use of AI for code generation leads to no one in the team understanding the system architecture as a whole after a couple of years.
Synthetic Customers: PwC and Bain on a New E-commerce Tool
Synthetic customers are AI models that simulate real customer reactions and help e-commerce companies test prices, ads, and products before market launch, according to Bain and PwC.