2026

Research

Musk admits there was no written agreement with OpenAI when it was founded

Business

OpenAI breaks free of Azure: Microsoft rewrites its contract with its key partner

LLM

How Sun Finance and AWS sped up document checks and cut fraud risk

LLM·AWS Machine Learning Blog

AWS unveils a system for migrating and upgrading production LLMs with prompt optimization

AWS introduced the Generative AI Model Agility Solution, an approach to migrating and upgrading production LLMs through prompt conversion, optimization, and quality…

Medicine·TechCrunch

BioticsAI on Clearing the FDA, Raising Money and Keeping a Team Motivated in Healthtech

BioticsAI CEO Robhy Bustami explains how a healthtech startup handles FDA clearance, raises capital and keeps its team motivated in a heavily regulated industry.

Business·TechCrunch

X relaunches its AI ad platform to restore revenue growth

X is launching a rebuilt AI-powered ad platform, betting on campaign automation to win back advertisers and restore revenue growth.

Ethics·Wired

Musk admits under oath that xAI distilled OpenAI models to build its own systems

Questioned by lawyers under oath, Elon Musk did not deny that xAI used OpenAI models for distillation, calling it the norm across the AI industry.

01Stripe unveils Link, a wallet for online payments and purchases made through AI agentsApr 30, 2026·2 min 02Salesforce hands its AI roadmap to customers, speeding up Agentforce and new feature releasesApr 30, 2026·3 min 03Google brings Gemini to cars with Google built-in, will update millions of vehiclesApr 30, 2026·3 min 04Meta confirms leak: AI agent's advice exposed sensitive data to employeesApr 30, 2026·2 min 05Nvidia CEO Jensen Huang urges AI companies to stop scaring people with doomsday talkApr 30, 2026·3 min

Top themes

Claude Code Without the Magic: Habr Breaks Down Architecture, Context Noise and Engineering Practices

Ollama and LiteLLM: Turning a Python Script into a Full Console LLM Chat

Sber shows how RAG and an in-IDE LLM turn manual test scripts into automated tests

Autonomous·Habr AI

VS Code 1.111 launches Autopilot: an AI agent that writes, tests, and deploys on its own

Business·MIT News

MIT Conference: Karen Hao and Paola Ricaurte on the Right Path for AI

At an MIT conference, journalist Karen Hao and researcher Paola Ricaurte discussed who controls the trajectory of AI, and how to put the technology back in the…

Habr explains why language models and classic RAG lose track of connections

Habr examines why large language models with classic RAG are good at finding text fragments but struggle to reconstruct links, causality and hidden dependencies.

Startups·TNW

Apollo.io buys Pocus to bolster its AI sales platform and push deeper into enterprise

Apollo.io is acquiring Pocus and adding signal analytics to its GTM platform, aiming to move deeper into enterprise and unite sales, data and AI automation in one system.

Autonomous·MIT Technology Review

Generative AI Put to the Test in a US Supreme Court Case: Can It Replace Lawyers?

Ethics

White House unveils AI regulation plan: Trump wants a single national standard

Research

Elon Musk's xAI Sends Engineers Into Client Offices to Win Contracts From OpenAI

Creative

BotHub lists 10 free AI tools for text, code, video, 3D and documents

Hardware·Bloomberg Tech

Super Micro co-founder charged in alleged scheme to divert Nvidia-powered AI servers to China

US prosecutors have charged a Super Micro co-founder and two others with illegally shipping servers containing Nvidia chips to China in violation of export controls.

Business·Guardian

Hachette pulls horror novel Shy Girl over suspected AI use

Hachette has canceled the US release of Shy Girl and halted UK sales after a review, amid allegations that the novel may have been written with heavy reliance on AI.

LLM·TNW

WordPress.com lets AI agents write and publish content, with human approval

Automattic has added a write mode to WordPress.com's MCP integration: AI agents can now create posts, pages and comments, but every action requires human approval.

LLMs in development: four approaches teams use and how they differ

LLMs are no longer just autocomplete. This piece breaks AI-assisted development into four modes along two axes: how much code a human controls and how the team…

01OpenAI and the UK Government: Eight Months After the Deal, No Pilots Have StartedApr 30, 2026·2 min 02Cursor, Copilot and Claude Code make a roundup of 12 popular AI agents for developersApr 30, 2026·3 min 03OpenAI to nearly double its headcount by end of 2026 as Google and Anthropic close inApr 30, 2026·3 min 04Kled AI and other services are buying people's personal data to train AI modelsApr 30, 2026·3 min 05Michael Smith pleads guilty to fraud scheme built on AI-generated tracks and fake streamsApr 30, 2026·3 min

Top themes

Creative·3DNews AI

OpenAI to nearly double headcount by 2026 to push ChatGPT sales to businesses

LLM·MarkTechPost

MarkTechPost shows how to build an LLM system with self-evaluation, confidence scores and web search

Hardware·Habr AI

OpenAI, Google and Tesla set the week's agenda: GPT-5.4 mini, AI Studio and Terafab

AI Independence Bench compares 49 models and measures how well they resist user pressure

Creative·Habr AI

How the CEO of digital agency bijobs.ru built a personal speaker site over a weekend with Claude Code

Bogdan Nepryakhin of bijobs.ru explains how two weekends with Claude Code replaced a month of designer and front-end work, at a fraction of the cost.

Robotics·Bloomberg Tech

Elon Musk Launches Terafab in Austin: Tesla, xAI and SpaceX Prepare a Chip Fab

Elon Musk has picked Austin for Terafab, a future in-house chip fab for robots, AI and space-based data centers, launched by Tesla, xAI and SpaceX.

Hardware·3DNews AI

Huawei unveils Atlas 350, claims it outperforms Nvidia H20 in AI inference

Huawei has announced the Atlas 350 accelerator, built on the Ascend 950PR chip, and claims it is more powerful than Nvidia's H20 for the Chinese market, doubling down on its in-house AI chip platform.

Python developer reins in Claude Code: an open-source config against regressions

Gemini: How Google Turned the Bard Fiasco Into 750 Million Users and AI Leadership

Autonomous·Habr AI

Perplexity turns AI search into an agent that builds Excel files and checks the data itself

Creative·3DNews AI

OpenAI slows data center buildout to calm investors ahead of IPO

Hardware·MIT Technology Review

Silicon Valley animal advocates want to recruit AI researchers as allies

LLM

GitAgent proposes a single AI agent format for LangChain, AutoGen and Claude Code

Hardware

Upstage in talks with AMD to buy 10,000 AI accelerators for South Korea

Autonomous

DeepMind proposes ten cognitive scales to measure progress toward AGI

AI Agents vs RAG: How ReAct Works and Why Multi-Agent Systems Exist

A single LLM response is no longer enough; real tasks need a chain of actions. We explain what AI agents are, how they differ from RAG, how ReAct works and why…

Leman Tech shows how it combined LLM, RAG and classic ML in tech support

The company describes a hybrid support setup: fast ML classifiers handle bulk tickets, while an LLM with RAG steps in for Wiki-based answers and smart escalation.

Ethics·Habr AI

Raiffeisenbank Ships a Kotlin RAG Assistant With No Python and No New Hires

Raiffeisenbank's team built an internal RAG assistant on a JVM stack, Kotlin and Spring Boot, without Python and without bringing new skill sets into the team.

Garage Eight: How AI Is Changing Analysts' Work and Why Junior Roles Are Disappearing

A Garage Eight analyst outlines six trends already reshaping the profession: AI takes over routine work, raises the bar for juniors and lets managers handle data on their own.