ZAYA1-8B is a mixture-of-experts language model that matches or exceeds substantially larger open-weight models on complex reasoning, mathematics, and coding tasks while using fewer than one billion active parameters. Trained on full-stack AMD infrastructure, it introduces Markovian RSA for test-time compute and is open-sourced under Apache 2.0.

Generative artificial intelligence is being used by 17.8% of the world's working-age population, but the gap between wealthy and developing nations continues to widen, according to a report published Tuesday by Microsoft.

OpenAI is launching an optional safety feature for ChatGPT that allows adult users to assign an emergency contact for mental health and safety concerns. Friends, family members, or caregivers designated as a 'Trusted Contact' will be notified if OpenAI detects that a person may have discussed topics like self-harm or suicide with the chatbot.

The article explores the real-world impact of AI adoption in businesses, noting that while AI has significantly increased output, revenue has not followed suit. It analyzes challenges in utilizing saved time, customer acquisition difficulties, risks of altering core workflows, and warns against brand dilution. It concludes by urging strategic allocation of productivity surplus for long-term competitiveness.

Aramco and IBM announced plans to collaborate on industrial AI, automation, material science, and other domains, leveraging their strengths to drive innovation and operational excellence.

If you've lost your Roku remote, don't worry. Here are four methods to control your TV without it, including using the Roku app, casting, screen mirroring, or AirPlay. You can also purchase a replacement remote at an affordable price.

Renesas Electronics Corporation has completed the acquisition of Irida Labs, a Greece-based embedded software company specializing in AI-powered visual perception. The acquisition strengthens Renesas' edge AI offerings, enabling integrated system solutions for industrial, robotics, smart city, and other markets. Irida Labs' technology will be integrated into Renesas' new Renesas 365 platform.

Mozilla leveraged the Claude Mythos preview to identify and fix hundreds of vulnerabilities in Firefox, with a dramatic increase from 20-30 monthly fixes to 423 in April 2026, showcasing the improved efficacy of AI-generated security reports.

agent-data is a CLI tool that provides AI agents with real-time, structured data from various endpoints like flight information, restaurant availability, AI blog aggregation, and social media. It uses fewer tokens than browser automation, enabling more efficient data retrieval. The tool works with popular AI agents such as Claude Code and Hermes, and includes a demo contrasting agent performance with and without agent-data.

This article explores why AI's impact is not yet reflected in productivity statistics, drawing historical parallels to the computer and automobile revolutions. It introduces the concept of the 'kit stage' in technology adoption, where amateur tinkering and user creativity drive innovation before being formalized by firms. The article argues that AI is currently in such a lag period, with falling transaction costs and rising internal integration costs leading to a bifurcation of firms into large enterprises and small operators.

Nutrepedia is a multilingual nutrition reference site built by a solo founder using Clojure, HTMX, and Postgres. It offers 47,415 localized food pages across 29 locales, with search supporting multiple scripts. The content pipeline is AI-assisted with an evaluation layer. The goal is a free, global food encyclopedia.

This essay argues that AI will not eliminate the need for coordination, management, and judgment. While AI reduces costs of production, it makes filtering, attention, and trust more important. The adversarial nature of business and scale of organizations create irreducible coordination work.

Bookstoread.ai is an AI-powered book recommendation engine for non-fiction. It offers features like Books Like..., Master a Topic, and Match My Vibe. Currently exploring titles such as The Terror, The Pillars of the Earth, and more.

The EU has agreed on simplified AI rules. The "Digital Omnibus on AI" pushes back deadlines for high-risk AI to late 2027 or 2028 and eases requirements for small and medium-sized businesses. "Nudification" apps are now explicitly banned. The labeling requirement for deepfakes and AI-generated text still takes effect in August 2026.

This Guardian editorial criticizes the use of facial recognition technology, arguing that civil liberties are being sacrificed for policing efficiency. It highlights lack of oversight, racial bias, and risks of misidentification, calling for better redress and independent regulation.

Steerable is a command-line tool that uses AI to convert code diffs into narrated explainer videos. Install via npm, run 'steerable review', and get a shareable link. It integrates with VS Code, Cursor, Claude Code, and more. Free tier available with 5 renders/month, paid plans from $49/year. Early access with 800+ users.

An analysis of AI progress in Q2 2026, focusing on Claude Mythos, revenue growth, and whether the explosive AI 2027 scenario is materializing. The author argues AI is advancing rapidly but remains a normal technology with adoption hurdles.

NVIDIA and Corning are partnering to massively expand US manufacturing of advanced optical connectivity for AI infrastructure. Corning will increase optical connectivity capacity 10x and fiber production over 50%, building three new facilities and creating over 3,000 jobs.

Memory Arena is a new approach that enables AI coding agents to improve by analyzing past programming tournaments.

AWS introduces the first payment capability designed specifically for autonomous AI agents, enabling seamless transactions within agentic workflows.

OpenAI launched three new speech-focused models: GPT-Realtime-2, its first voice model with GPT-5-class reasoning; GPT-Realtime-Translate for live translations; and GPT-Realtime-Whisper for fast transcriptions. GPT-Realtime-2 offers 11% performance improvement, extended context window to 128k tokens, parallel tool calls, and adjustable reasoning effort. Pricing remains the same as GPT-Realtime-1.5. The translation model supports 70+ input languages and 13 output languages, priced at $0.034/minute. The transcription model costs $0.017/minute. OpenAI also highlighted three patterns for voice AI: voice-to-action, system-to-voice, and voice-to-voice.

Airlock is a self-hosted platform for running 'cyborg agents' that are half code, half AI. Each agent runs as a long-lived Docker container with its own Postgres schema, S3 storage, web dashboard, custom HTTP routes, webhook ingress, cron scheduling, chat platform bridges, and proxied access to LLMs and MCP tools. It features RBAC, real-time event streaming, and full audit trails. The project is in alpha, licensed under AGPL-3.0 with commercial options available.

Anthropic's deal to use all of xAI's Colossus data center capacity raises environmental concerns, as the facility has a poor record. Meanwhile, xAI's sudden deprecation of several Grok models the night before the announcement highlights risks of vendor dependency.

DARA is an open-source compiled memory system for AI agents that provides persistent context across multiple AI platforms. It uses markdown files and Python, requires no cloud, and features a 10-step compilation pipeline for validation, deduplication, and integrity. It supports governance by consensus, self-healing, and two reading modes. It has been tested across 4 models with 92.6% average score.

A software developer explains why he cannot fully trust AI language models, sharing personal experiences with AI-generated coffee brewing recommendations and a hallucinated coffee shop, and arguing that AI output requires expert oversight.

ZDNET readers love this E Ink tablet, thanks to its paper-to-pen writing experience and versatile uses.

Reefy is a new tool that transforms spare PCs, mini PCs, laptops, and GPU boxes into private AI servers with a simple USB flash drive setup. Built on Buildroot, it offers fast boot, NVIDIA GPU support, encrypted backups, remote access, and one-click deployment of popular AI workloads.

A study published in Neurology found that both short-term and cumulative exposure to air pollution, as well as heat and humidity, are associated with increased migraine activity. The results may help predict high-risk periods for attacks.

At the Milken Institute Global Conference, five key figures from across the AI supply chain discussed critical bottlenecks: chip shortages, energy constraints, novel AI architectures, agent trust, geopolitical tensions, and impacts on critical thinking.

memoirs is a fully local long-term memory engine for AI agents, using SQLite, sqlite-vec, and FTS5 for hybrid retrieval. It supports MCP, HTTP, and CLI interfaces, has a built-in local LLM for memory curation, and prioritizes privacy and performance.