Habr AI
AI news source. Articles are auto-selected and adapted by Hamidun News editors.
Latest publications
Auchan Tech showed how AI saves up to 70% of a systems analyst’s time on routine tasks
Auchan Tech used real examples to show how GPT, Claude, and Qwen help a systems analyst gather requirements, build diagrams, and draft technical specifications, saving up to 70% of time on routine tasks.
ChatGPT and Anthropic amplified the "AI veganism" trend: why fatigue with neural networks is growing
Against the backdrop of ChatGPT's growing user base and the adoption of Claude, more and more users are thinking about "AI veganism": neural networks help at work, but also heighten anxiety, distrust, and fatigue.
Kling, Veo and Sora: 10 leading AI models for video generation in 2026
The AI video market is seeing another shift in leaders: a new review compares 10 models — from Veo 3.1 and Kling 3.0 to Sora 2 and open-source solutions — by quality, use cases, and cost per second.
Huntley's take on Ralph loop: why Anthropic and Vercel approaches shouldn't be conflated
The term Ralph loop is already being used to lump together at least five agent architectures, and the analysis shows where the line between execution loop and evolution loop lies.
Habr AI explained why the em dash became a marker of AI-generated text
Habr AI published an analysis of how the em dash went from a sign of good editing to a suspicious trace of AI, and why authors now deliberately simplify typography.
PSB outlined its approach to RAG in fintech: architecture, metrics, and the testing cycle
PSB showed how RAG systems are evaluated in fintech: from chunking and vector search to Hit Rate and MRR metrics, manual verification, and a continuous testing cycle.
How a product manager without a technical background built two pilots in a week with Claude and Kiro
A product manager without hands-on development experience showed the stack used to launch a personal site and an agile service in a week: Claude handled discovery, Kiro built the product, and mcp covered the database, co
Wiren Board connected machine vision, WMS, and the conveyor system in a 41-destination sorting line
Wiren Board helped unify machine vision, WMS, and conveyor control in a 41-destination sorting line with throughput of up to 5,000 boxes per hour.
Manus and AI agents are changing development: MVP code now appears in 20 minutes
A Habr AI author tested how the Manus agent builds an MVP from an idea description and concluded: a draft product can now be made in minutes, but architecture and responsibility remain with humans.
How AI helped identify latent vertical phoria after a year of fruitless doctor visits
A personal case showed how AI led a person to a diagnosis of latent vertical phoria in 20 minutes after a year of fruitless visits to ophthalmologists and helped them get the right tests.
Habr AI showed that reinforcement learning still trails classical optimization in logistics
Habr AI tested whether RL can replace classical solvers in refueling-stop planning: the agent learned to make decisions, but across 86 routes it still lost to mathematical optimization on cost.
Habr AI compiled the best text rewriting services — from BotHub to QuillBot and Alice AI
Habr AI published a review of rewriting services: the roundup includes BotHub, Writesonic, ReTextAI, QuillBot, and other tools with varying quality for Russian-language text, pricing, and trial limits.
How Norges Bank Investment Management integrated AI into 171 processes across the fund
Over two years, NBIM, the world's largest sovereign wealth fund, deployed AI in 171 processes, required employees to train, and reorganized teams to cut costs and speed up work.
KOMPAS-3D gets an AI agent that builds parts, drawings, and exports DXF on its own
MCP for KOMPAS-3D has grown from a demo into a working AI agent: it now builds 3D parts, manages the feature tree, prepares drawings, and exports DXF with a single request.
Generative AI puts copyright and the very idea of intellectual property into question
The latest turn in the dispute around generative AI is no longer limited to the content market: the technology blurs the role of the author and forces a rethink of the very logic of copyright.
BHV adapted a book on Python and ChatGPT: from architecture to CI/CD and hardware integration
BHV released a Russian adaptation of a book on Python development with AI: instead of a set of prompts, it offers a methodology, an end-to-end Raspberry Pi project, and an honest look at ChatGPT's limitations.
Habr: according to Yandex Wordstat data, demand for "AI agents" trails bots and services
Habr published a column on the gap between the hype around AI agents and real demand: users search for bots, stores, and payments, not architectural terms.
Why artificial intelligence will not replace doctors and scientists, but become a tool in their work
An author with 15 years of experience in pharma explains why neural networks will not replace doctors and scientists, but are already useful for diagnostics, data analysis, and hypothesis generation.
Piter Publishing released a book on GraphRAG and advanced RAG on knowledge graphs
Piter Publishing introduced a book on GraphRAG — an approach that combines knowledge graphs and vector search to make RAG systems more accurate and useful in real-world tasks.
Apple showed how the @Generable and @Guide macros in FoundationModels simplify Swift development
In Apple’s FoundationModels framework, the @Generable and @Guide macros make it possible to describe data structures in Swift and receive pre-filled typed objects from a language model.
Apple unveils Foundation Models: how to embed local AI into SwiftUI apps
Apple has released Foundation Models, a framework for local AI in SwiftUI apps that runs directly on the device, preserves privacy, and removes cloud costs.
Why artificial intelligence triggers old fears — from witches to algorithms
The analysis shows that anxiety around artificial intelligence follows the same pattern as fear of witches, the plague, and machines: society first demonizes what it does not understand, then learns to live with it.
Habr AI: how a working no-code AI prototype raised questions about developers' future
In a Habr AI column, the author describes the moment when a product manager with no programming skills built a working prototype using AI and explains why this is already changing the developer's role.
Contextual amnesia: why AI agents in 2026 forget everything they knew yesterday
In 2026, AI agents work like senior developers, but each new session starts from a blank slate — with no memory of decisions, mistakes, or the project's architecture.
Cursor sped up development 10x and left a financial company with a million lines to review
After adopting Cursor, one financial company increased its code output tenfold, but along with the speed it also accumulated a million lines awaiting review, a rise in vulnerabilities, and stress across the team.
Claude Opus helped a business analyst launch two apps in two weeks without a team
The author of the experiment, with no programming experience, built a time tracker and a game for RuStore in Flutter in two weeks, and called Claude Opus the best assistant in development.
Yandex SpeechKit and CosyVoice compared on voice bot and audio podcast tasks
In a new review, Raft’s author examined how the open-source CosyVoice and the proprietary Yandex SpeechKit perform in two applied scenarios: a realtime bot handling abbreviations and narration of long texts.
Generative AI in software development: not a replacement for junior developers, but a new source of burnout
For two years, researchers at the AI Research Institute have been observing how developers work with generative models — and documenting not efficiency gains, but a new type of professional burnout.
Habr AI showed how to build your own RAG retriever in LangChain for names and terms
Habr AI published a practical walkthrough on how to build a custom RAG retriever with TF-IDF and LangChain for cases where vector search performs worse on names, titles, and terms.
ChatGPT instead of search: how to track brand presence in AI responses
77–85% of informational queries do not lead to clicks to websites — AI assistants provide the answers. New GEO and AEO metrics help determine whether an AI model mentions your brand.