Habr AI
AI news source. Articles are auto-selected and adapted by Hamidun News editors.
Latest publications
Local RAG with FastAPI and Ollama: From a Simple LLM Call to a Production Backend
A developer on Habr shows how to turn a single POST request to Ollama into a backend system with an API contract, sources, logging, and negative tests.
How to Build a 114,000-Line Game with AI: Context Management in Practice
A developer wrote a 114,000-line browser game in three weeks using neural networks, and explains how to manage a codebase that no longer fits in the context window.
ChatGPT Multi Pane Chrome extension replicates Codex logic without limits or extra costs
A developer built a Chrome extension that adds multiple parallel chats to ChatGPT — functionally, it is a Codex analogue, but without a paid subscription or strict limits.
Claude in a tester’s workflow: a QA engineer’s real-world experience with prompts and metrics
From a QA engineer’s personal experience: how Claude cut the time spent on checklists, bug reports, and test documentation — with concrete numbers and ready-to-use prompts.
Vibe coding in practice: how a developer learned to save tokens without sacrificing code quality
A frontend developer with B2B experience had wanted to build mobile apps for years — AI gave him a chance, but also taught him a lesson about tokens and code architecture.
RAG + call graph: automatic context from a task tracker for Claude Code
A developer described how to use a repository RAG index and a plugin for Claude Code to automatically pull in the relevant context for a task from the tracker — without extra tokens.
“AI Girl for a Cyborg”: Why Intuition Is Becoming the Key Asset in the AI Era
The Habr column “AI Girl for a Cyborg” explores how AI and human intuition work in tandem — and why “cyborg mode” has already become a practical strategy today.
Habr Shows How to Build a Home AI Server for Under 400,000 Rubles
Habr enthusiasts break down the hardware needed to run neural networks at home and how to build a full AI server for roughly a third of the price of a ready-made one.
ICRA 2026 in Vienna: AIRI and MIPT Researchers Report from the World's Largest Robotics Forum
Alexander Panov of AIRI and MIPT shares his impressions of ICRA 2026: five packed days of talks and discussions, plus the traditional robot parade, in a welcoming June Vienna.
GPT-5.6 Sol Agent Spammed Its Way Into the Red: A 24-Hour Experiment With a Real Business
Bottleneck Labs let an agent named Saul, built on GPT-5.6 Sol, run a real startup for a day: it spent $99.50, gained 5 users and earned not a single dollar.
Agentic Development Without Governance Is Dangerous: A Three-Layer Stack in Practice
Autonomous AI agents make thousands of decisions an hour, and no developer can check them all. The authors explain why a Governance stack has become mandatory.
The Vibe Coder's Cookbook: One LLM as Architect, Reviewer and Tester
A developer on Habr built a production project with almost no hand-written code, using a single LLM in four roles, one that honestly criticizes the code it wrote itself.
"AI Will Do It in One Evening": Three Estimation Methods Show 300 Hours Is Below the Industry Floor
A client called a 300-hour estimate inflated. The developer re-ran the numbers with three methods — bottom-up, ISBSG and COCOMO II — and got the opposite result.
BitNet b1.58: How Ternary Quantization Lets a 27B Model Run Without a GPU
Networks whose weights take just three values {-1, 0, 1} fit 27B models into 6–7 GB of RAM and run on an ordinary laptop with no graphics card.
Viterbi, beam search and LLM reranking: how an algorithm picks one meaning out of many
A Habr author uses Tarot cards to show why greedy decoding loses to exact Viterbi, and how the same logic drives decoding in modern LLMs.
Developer Writes a Svelte Compiler in Rust With an AI Agent in 106 Evenings and $700
A programmer spent 5.5 months hand-writing a Svelte compiler in Rust and abandoned it. A year later he returned with an AI agent and finished in 106 evenings and $700.
monday.com Lays Off 630 People for an AI Strategy Despite 24% Revenue Growth
monday.com is cutting 20% of its workforce, about 630 people, citing an AI strategy, even though Q1 2026 revenue rose 24% to $351 million.
A Quadrillion-Parameter LLM: When to Expect One — and Why the Answer Is "Never"
A Habr author examines the physical, computational, and economic limits that make quadrillion-parameter (10^15) models practically unreachable under the current paradigm.
AI in the Designer's Workflow in 2026: A Paradigm Shift and a Personal Process
How AI settled into design work in 2026: teamwork on a single shared canvas, targeted edits instead of generating from scratch, and one designer's four-step method.
Payment Security in Vibe Coding: Two Vulnerabilities AI Always Misses
A hands-on vibe coder breaks down two systemic failures in AI-generated payment code, price tampering and fake webhooks, and outlines five layers of defense.
NVIDIA's Open-Weights Coalition Grows to 50 Companies; Anthropic Hasn't Signed
Jensen Huang took a letter against restrictions on open AI models to Washington. Signatories doubled to 50 in a day, Google among them, but Anthropic and Amazon did not sign.
Physical AI and VLA Models: The Humanoid Race Between China and the US, and What Russia Has
Part two of Habr's deep dive into Physical AI covers humanoid economics, industrial rollout of VLA models, government programs, and Russia's place in the race with China and the US.
GPT-5.6 Generated a Proof That Solved a 6-Year-Old Cryptography Problem
Prabhanjan Ananth and Amit Sahai posted an arXiv proof that closes a six-year-old problem in quantum cryptography. GPT-5.6 generated the entire construction and the key ideas.
Why a Successful AI Pilot Fails to Pay Off at Scale: The Topic of INFOSTART CIO CAMP 2026
Strong pilot metrics don't guarantee payback: at INFOSTART CIO CAMP 2026, IT leaders will dissect the hidden costs of AI projects and the deployments that had to be shut down.
Timeweb Cloud: one in three documentation requests comes from AI agents, not humans
Timeweb Cloud's head of development says one in three requests to the cloud's documentation comes from AI agents, and servers are increasingly created via API on an assistant's command, not a human's.
Vibe Coding All the Way to Production: A Google Play App Where AI Wrote 99.9% of the Code
A developer shipped a native Android app in Kotlin with 50+ screens to Google Play, with AI generating 99.9% of the code, thanks to a detailed spec for every screen.
It's Not the Model, It's the Architecture: Breaking Down 3 High-Profile AI Failures (Deloitte, AI Cameras, an Agent)
Deloitte refunded money for a report with invented sources, AI cameras mistook a journalist for a carjacker, and generated code canceled every customer subscription. Three failures blamed on architecture, not the LLM.
Claude Sonnet 4.5 Knows When It's Being Tested: Why a Green Safety Benchmark Doesn't Mean Green in Production
In fall 2025, Claude Sonnet 4.5 told its auditors mid-safety-audit that it knew it was being tested; per Anthropic's system card, this happened in 13% of test dialogues.
Computer Vision in Retail: Lenta Reads Price Tags From a Moving Robot
Lenta turned price-tag recognition from a moving robot into a Computer Vision ML hackathon, aiming to automate shelf monitoring and violation detection in its stores.
Mac mini M4 Pro for Local LLM Inference: A Test Built Around OpenClaw AI Agents
Selectel tested the Mac mini M4 Pro on local inference of mid-size language models and checked where Apple Silicon hits its limits for AI agents like OpenClaw.