Habr AI
AI news source. Articles are auto-selected and adapted by Hamidun News editors.
Latest publications
ContentAI Taught AI to Write Production Code by the Rules: A Success Story
ContentAI demonstrated how vibe coding—AI generation without strict rules—can work in production applications by adding discipline to an intuitive approach.
Pelicans on Bicycles: Simon Willison's Amusing Test for LLMs
Simon Willison, Django creator, invented a fun test for LLMs: asking an AI to draw an SVG pelican on a bicycle. The results reveal unexpected insights about neural network capabilities.
Sam Altman Is Wrong: Parkinson's Law Will Protect Jobs from AI
Despite Altman's predictions that AI will steal jobs, Parkinson's organizational law will protect positions: people in the system always find ways to expand work and stay needed.
Developer Created an AI-Integrator Profession: How to Build a Workflow on n8n
Integration developer Daniil created his first AI automation on n8n: events are distributed via LLM across Google Sheets, Telegram, and CRM without a traditional release.
NVIDIA Tesla V100 for local AI models: testing on real-world tasks
The 2017 Tesla V100 16GB handles modern LLM models like Qwen35B and GPT-OSS-20B. In tests, it generated websites, games, and document summaries at speeds of 38–109 tokens per second.
Sharding in LLMs: How to Distribute Computations Across GPUs
Sharding is a technique for distributing large matrices across multiple GPUs. The quality of sharding directly affects training speed and computational efficiency in large-scale clusters.
Why RAG Chatbots Work Great in Demos but Produce Nonsense in Production
RAG bots shine on internal documentation demos but confidently generate nonsense in real work. A story about the gap between five prepared questions and production reality.
Google combines Street View and AI to create virtual training worlds
Google presented Project Genie in a new version — a generative model that creates 3D worlds tied to real Google Street View coordinates, allowing robots to train without entering the real world.
How a Developer Created a Music Generation Skill for Yandex Alice
A Melbourne-based developer spent two evenings adding a song creation skill to Yandex Alice controlled by voice commands—the first live example of generative models in use within Russia's smart speaker ecosystem.
Personal AI Agents Are Already Here. You Just Haven't Noticed Them
Experts talk about personal AI agents of the future, but we're already using them today — in search engines, recommendation systems, and virtual assistants, we just don't call them that.
The Architectural Mistake of Corporate AI: Why You Need an AI Gateway
Embedding AI in every system creates a zoo of copilots with security and audit problems. You need a separate corporate layer where models, management, and accountability live.
How Masha Leshchinskaya Saved Her Pet Project from Death: A Device Booking System for AI
Masha Leshchinskaya from Surf created a device booking system that's been alive for three months. Her secret: not cutting-edge technology, but minimal maintenance costs and practicality as a testing ground for AI experim
How to Use ChatGPT and Claude Subscriptions in Cursor Without Token Payment Fees
ChatGPT and Claude subscriptions provide $1000 and $2000 in tokens respectively, but Cursor doesn't support them directly due to different API formats. The solution is request proxying to route queries through paid subsc
From Idea to MVP in One Day: AI Built a PostgreSQL Profiler
A developer created a working PostgreSQL profiler in a single day using only an AI assistant—without writing a single line of code manually.
The Deficit Spiral: Why There Won't Be Senior Developers by 2031
Junior developer hiring has fallen 40%, and AI has pushed companies to cut training investments even further. A shortage of juniors today will turn into an acute crisis of experienced talent by 2031.
Video Analytics in Cities: Why Classic Video Processing Is Ineffective
Smart video analytics developers explained why traditional video stream analysis algorithms fail in urban environments and how modular AI architecture solves this problem.
Augustine and GAN: Why the Battle of Good and Evil is an Adversarial Network
The author converted Bosch's triptych into a mathematical model of entropy and drew an analogy between Augustine's philosophy and generative adversarial networks (GANs).
How a Manager Learned to Code in n8n and Claude Code — A Story of Breaking Free from Expensive Contractors
A manager spent months mastering n8n and Claude Code to build automations herself instead of overpaying contractors 300,000 rubles for simple features.
MWS AI Meta-agent: AI Started Assembling AI Agents for Users
The MWS AI team created a meta-agent that assembles other AI agents from text descriptions, bypassing the need for programming — the project won a competition and was implemented on the platform.
From Prompts to Agents: How a Doubletapp Engineer Reformatted the Approach to AI
An iOS developer from Doubletapp shared how he evolved from asking ChatGPT questions to building real engineering infrastructure with agents and context engineering.
Yandex Music Updated Its Recommendation Algorithm to Prevent Musical Tastes from Stagnating
In 2026, the 'My Wave' algorithm received an update. Already half of users' playlists consist of recommendations — music no longer stagnates.
Not a Model, but a System: How Svoi Built a 7-Layer Fintech Bot Architecture
Svoi.ru explains why voice bots for banks can't rely on just a good model — they need a seven-layer architecture where each layer manages its own part of the risk.
An AI Agent Created a Ticket, Took It to Work, and Closed It—The Manager Noticed Nothing
Autonomous AI agents are embedded in real CI/CD pipelines and close live tickets in production. Metrics look perfect, hiding the actual project state and growing technical debt.
SciGraph: how a graph of scientific connections outperforms text search
SciGraph applies a graph approach to scientific papers: instead of just text, AI sees connections between authors, methods, and citations — and answers research questions more accurately.
Local Qwen3.6-27B vs. cloud models: why privacy is not the main point
Local Qwen3.6-27B shows it can compete with cloud-based Claude and GPT. The advantage of local models extends far beyond privacy — it is control, speed, and independence.
llms.txt: how to help ChatGPT, Claude, and Perplexity cite your site correctly
llms.txt is a file in your site's root directory that tells ChatGPT, Claude, and Perplexity how to cite you and what to treat as canonical information.
Programming Languages Will Be Redesigned for AI Agents, and Here's Why
When AI agents start writing code, existing languages won't be enough. The industry faces a wave of new languages optimized for machine understanding, while old ones will remain.
Zabbix and Local LLM Integration: How to Design an Architecture for Smart Alerts
Part three of the Zabbix and LLM integration series explores smart alert architecture design. Discover which HLD components humans handle and where neural networks help.
A conductor instead of an assembly line: how AI rethinks the classic pipeline
The traditional development pipeline—from product to DevOps—now encounters AI agents at every stage. The system must either change or stop working.
RAG systems break on real data: the culprit is retrieval, not the model
A RAG system looks perfect in a demo, but once testing starts on real questions, half of its answers turn out to be wrong. The problem is not GPT-4 or Claude — it is how the system retrieves relevant chunks from document