Zhemal Khamidun

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…

Research·Apple ML Research

Apple ML Research Reveals Inefficiency in Mixture-of-Experts Routing

Apple researchers found that tokens in MoE architectures utilize a negligible fraction of theoretically possible routes, and proposed an architecture to fix this.

Hardware·Habr AI

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…

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.

01South Korea to Direct Chip Boom Taxes into State Development FundJul 8, 2026·2 min 02Apple confirms: speech diffusion models scale like autoregressive onesJul 8, 2026·2 min 03Perfectly Imperfect: A Newsletter About Celebrities' Unusual Tastes Against 'AI-Slop'Jul 8, 2026·2 min 04AI in Medicine: How Neural Networks Diagnose in the NHS AppJul 8, 2026·2 min 05Broadcom Extends Partnership with Apple for Custom Chip Supply Until 2031Jul 8, 2026·3 min

Top themes

Creative·TNW

NCA Warns Parents: AI-Generated Images of Child Abuse Are Rapidly Multiplying

Medicine·TNW

Microsoft to cut 3,200 Xbox jobs and shed five game studios

LLM·TNW

Mistral CEO warns business: closed AI providers gain control over your data

LLM·TNW

Reddit Launches AI to Combat Hidden Ads in GEO-Marketing Era

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…

Creative·TNW

Worldmodeldata from Cambridge raised £7 million to train AI on video games

Cambridge startup Worldmodeldata closed a seed round at £7 million—the company converts gameplay into training datasets for 'world models', a new class of AI…

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.

Ethics·TNW

Denmark backed Belgium in EU court: platforms must pay publishers

LLM·TNW

Sysdig Documents First Ransomware Attack Fully Executed by AI Agent

Hardware·MIT Technology Review

Sam Altman negotiates OpenAI stake for every American family

Creative·3DNews AI

Reddit deployed language models to fight AI-generated spam

Research·Apple ML Research

Apple ML Research Solves ASR Model Degradation on Long Audio Recordings

Research

Apple Published Fortress — A Framework for Stabilizing Recommendation Systems Through Feature Pruning

Medicine

Apple ML Research investigates why AI safety data labelers disagree

Medicine

Can AI write prescriptions? A case from Utah 2026

Ethics·Guardian

AI Surveillance Intensifies and Could Hamper Social Progress — Schneier and Penny

Bruce Schneier and John Penny warn: AI surveillance will soon encompass public and private life — and could suppress civic activism if society fails to adopt…

Business·TNW

ITV sells TV channels and streaming service ITVX to Sky for £1.6 billion

ITV has agreed to transfer its broadcasting division — ITV channels and ITVX platform — to Sky (Comcast) for up to £1.6 billion, becoming a pure studio.