Habr AI
AI news source. Articles are auto-selected and adapted by Hamidun News editors.
Latest publications
"AI Slop": Telegram Bot Developer Figured Out Why the AI Label Irritates the Audience
Developer built a Telegram bot for three months, and the first top comment called it "AI slop" — the author figured out why these two words irritate technical readers.
How a developer cut an AI agent's token consumption by 4.6M per day: analyzing the leak
Cron tasks, diagnostics, and massive tool lists silently drain millions of tokens — a developer found the leak and reduced daily consumption by 4.6M tokens.
Anthropic Prompts vs grill-me Skill: A Two-Hour Test by a Non-Developer in Claude Code
A Habr user with no development experience compared five ready-made prompts from Anthropic's library and the grill-me skill — and wrote a Habr parser during the session.
Yttri 0.86.0-beta.1: First Obsidian Plugin, MLX Engine and Unified Agent Interface
Yttri released version 0.86.0-beta.1: agent moved to unified interface section, plugin platform launched with Obsidian integration, and local MLX engine for Mac now available.
Hallucinations in AI coding tools: why agents are slower than search-and-replace
A developer spent two hours on a task for a CLI agent that could be solved manually in 20 minutes — and investigated why detailed prompts amplify LLM hallucinations.
Fable vs 10 LLMs: Who Better Handles the God Node from a Real Agent
The author tested 11 language models—including Fable and six Chinese ones—on the task of decomposing a real god node from a LangGraph agent, using cross-evaluations and three trust verification methods.
AI in development: vibe coding instead of developers
Katya Obraztsova of Bitrix24 on how a five-person team built a corporate service from scratch in three months using vibe coding and evolved into an AI-driven team.
Inside Chinese AI: ByteDance, Z.ai, humanoid robots requiring passport ID and the server GPU card market
Eight days of negotiations with ByteDance, Z.ai, SiliconFlow and MiniMax in China: what launches July 10, why Z.ai is switching from NVIDIA to Huawei and how humanoids are sold in Shanghai with passport verification.
MCP and AI Agent Security: How the Protocol Created a New Attack Surface
Model Context Protocol standardized AI agent connections to tools — but each MCP server is now a trust point, vulnerable to tool poisoning and prompt injection.
VK Created Discovery AI Neural Search for Multi-Billion Content Base of Zen and VK Video
A leading VK developer unveiled the architecture of Discovery AI — an LLM-based system for search and recommendations across Zen, VKontakte, VK Video, and Mail.
How to build an AI harness for SaaS development: combining Orca, Pi Agent, and GitHub
A developer on Habr explains why a well-built harness around a model matters more than the choice between GPT or Claude itself — and shows his working setup.
A Reinforcement Learning tutorial without the math: just code for programmers
A developer from Cinimex publishes an introduction to reinforcement learning on Habr — no math formulas, just understandable code and explanations of why each step is needed.
CrewAI Launches Conversational Flows: Building Agentic AI for Banks and Telecom
CrewAI has introduced an experimental Conversational Flows mechanism to balance deterministic scenarios and AI reasoning in enterprise-class agentic systems.
AI Engineer World's Fair 2026: Agentic Pipeline Analyzed 560 Conference Sessions
A developer built an agentic pipeline and processed 560 sessions from AI Engineer World's Fair 2026 — the first Russian-language breakdown of the conference with summaries, timecodes, and navigation.
Saint HighLoad++ 2026: AI Gets Its Own Track and a New Term "harness"
At Saint HighLoad++ 2026, AI occupied its own track for the first time: the terms "harness" and "ai-native" were heard, and speakers did not notice contradictions in their own theses.
SimpleOne explained the architecture of AI agents in its enterprise GenAI platform
The SimpleOne team broke down how AI agents are built in their GenAI platform: components, use cases in corporate processes, and the approach to security when working with real data.
Sberbank Converted GigaChat3 to Diffusion Mode: How the GFusion Project Works
The GigaChat Pretrain team (Sberbank) converted the 10-billion parameter GigaChat3 model to diffusion mode without training from scratch and obtained two experimental checkpoints: GFusion.
MTS Web Services Switched from RAG to AI Agent for Initial Support Ticket Triage
The MTS Web Services team explained how their RAG system for searching Jira and Confluence struggled with incident investigation — and how it was replaced by an autonomous AI agent.
Vibecode in Bitrix24: Building a Customer Profile with AI Agents — A Step-by-Step Experience
The team tested the Vibecode platform within Bitrix24: built a "customer profile" application with contacts, deals, and leads, and honestly reported where the AI failed.
Machine Vision: How AI Technologies Are Patented in Russia and Globally — OnlinePatent Overview
OnlinePatent patent agent analyzed on Habr how companies protect developments in machine vision — from recognition algorithms to neural network architectures — in Russia and abroad.
60% of AI Projects Fail to Reach Production — How ESM Creates a Foundation for Success
According to Gartner, 60% of enterprise AI projects fail to reach production due to unready data — not budget. ESM platforms offer a structural solution to this problem.
Production Agent on Claude Without LangChain and RAG: 712-Line Architecture for an Interior Design Studio
A developer built a production AI agent for an interior design studio on Claude Haiku 4.5 and Sonnet 4.6 without LangChain and RAG — and honestly analyzed all architectural tradeoffs.
Rented Intelligence: Why Russia Is Losing the Battle for AI Sovereignty
A Habr analyst breaks down why renting Western AI models leads to a dead end, how labor replacement pumps money abroad, and what lawmakers will respond with.
Anthropic Blocks Foreigners from Fable 5: Global AI Splits Into Sovereign Models
On June 12, 2026, Anthropic unexpectedly revoked access to Fable 5 and Mythos 5 for all foreign users — an incident that accelerated the race for sovereign AI models.
Prompt injection via Claude Code skills: how to check a plugin before installing it
OSS skills for AI assistants can contain executable code with user privileges. We explain why prompt injection via plugins is more dangerous than ordinary vulnerabilities.
Marina Lomadze, Cloud.ru: what to become in three years if your colleague is an AI agent
Cloud.ru head of hiring explains which roles are emerging around AI right now — and how familiar professions are transforming when working alongside multi-agent systems.
Oracle, Microsoft and Meta laid off 165,000 employees to invest in AI infrastructure
The biggest tech companies — Oracle, Microsoft, Meta and Amazon — laid off 165,000 employees, redirecting the freed-up funds toward building AI infrastructure. *Meta is designated an extremist organization and banned in
Sber AI Lab adapts object detection methods for event forecasting
Researchers at Sber AI Lab found a formal analogy between object detection and event forecasting and presented the results at the AAAI 2026 conference.
Fuel Crisis Exposed a New AI-Era Phenomenon: Dozens of Identical Services in One Week
Over seven days, more than ten identical gas station monitoring services emerged in Russia — independently created with AI assistance during the gasoline shortage.
Subagents in Claude Code: how to delegate reviews and tests to separate agents
Claude Code runs into context limits, not model capacity — subagents solve this by moving reviews, tests, and audits to separate agents with their own context and rules.