Epoch AI's independent database covers 67 large AI data centers globally, tracking their construction timelines via satellite imagery and public documents. The largest facility is SpaceXAI's Colossus 2 in Memphis, with 946 MW IT power and compute equivalent to 1,112k H100 GPUs. The US dominates, especially in Texas and Ohio. Total IT power capacity reaches 10.8 GW, with facility power around 14 GW. Hardware mainly features NVIDIA GPUs, with Google and Amazon using custom chips.

BenchLocal is a local-first desktop app for running, comparing, and managing installable LLM Bench Packs against local or remote models. It features model registry, Bench Pack installation and updates, run execution history, and an agent API for AI workflow control.

The UN convenes a global dialogue on AI governance in Geneva, aiming to anchor rights such as free expression. However, without enforcement, recommendations are unlikely to be followed. The article critiques major vendors and nations, emphasizes transparency and auditability, and argues that governments have leverage over AI vendors.

Tortma is a Chrome extension that uses drag-and-drop and AI suggestions to quickly organize YouTube videos into playlists, with free and paid tiers.

Mozilla.ai launches Otari, an open-source LLM control plane that provides a unified platform for managing routing, budgets, governance, deployment, and reliability across multiple LLM providers, reducing infrastructure complexity for developers.

A year after cutting around 9,100 employees, Microsoft is making further layoffs today as it begins its new financial year. The software maker is laying off around 4,800 employees today, approximately 2.1 percent of its workforce. Most of the employees affected by today’s cuts are in Microsoft’s commercial sales business or the company’s Xbox division.

An AI coworker that lives in your Slack to get work done.

Moderna uses AI to scale productivity and accelerate mRNA sequence design. The company has deployed ChatGPT Enterprise across its workforce, with employees building over 750 custom GPTs, achieving 100% adoption in legal. Its mRNA Design Studio compresses vaccine design from months to days, exemplified by the COVID-19 vaccine's 42-day timeline from sequence to clinical batch.

A developer shares how they successfully automated building a Kotlin Multiplatform application with an AI pipeline.

Brands spent 2025 racing to prove they're "AI-powered." Now consumers can spot AI-generated content instantly, and the novelty that once signaled innovation increasingly signals laziness. The brands treating AI like a marketing strategy are learning it's not one.

This issue of Import AI covers Fable's record-breaking GPU kernel (18.71x speedup), a significant rise in AI automation on the Remote Labor Index (from 2.5% to 16.1%), the OSWORLD 2.0 benchmark for long-horizon computer tasks, and JD's Oxygen AI Item Center managing billions of SKUs. These developments signal AI's accelerating penetration into research, economy, and business operations.

Anthropic's Fable 5 promises mythic AI power, but surprise restrictions make me wonder if it's more trouble than it's worth for day-to-day use. The author explains five reasons for sticking with Opus 4.8: volatility, moving guardrails, higher cost, potential slowness, and the imminent arrival of Opus 5.

A developer explores why it's crucial to question AI coding assistants, sharing experiments with Claude that reveal the need for explicit guardrails to prevent unintended actions.

This article explores five concrete ways small language models (SLMs) are being used in next-generation AI agents, including handling repetitive tasks, running on-device, fine-tuning for tool calling, collaborating with larger models in heterogeneous systems, and keeping sensitive data local. It provides research backing, practical tools, and cost comparisons to help decide when an SLM is sufficient over a frontier model.

In the near future, AI-powered surveillance systems will track our public and private lives. But we can make policy choices to reject it.

To compare rates for serverless and reserved GPUs, look at your application's peak-to-average ratio.

Big tech companies have recently changed their stance on AI causing massive job losses, now suggesting AI will augment rather than replace human workers.

A personal rant expressing boredom with AI discussions, comparing them to vapid vaping enthusiasts and scolding health advocates, with a wish for the AI bubble to burst.

A comprehensive guide on how to maintain tool selection accuracy in AI agents as the tool catalog scales, covering techniques like gating, retrieval-based selection, semantic routing, planning, fallback logic, and benchmarking.

The FCA's landmark review warns that AI will amplify cyber-crime and fraud risks, and urges ministers to toughen the City regulator's powers to protect consumers.

Researchers from Oxford and Potsdam universities found that AI tools used to redraft messages on sensitive political topics inject political biases, which could spread rapidly and reshape long-term public opinion.

China's new AI companion regulations take effect July 15, requiring anti-addiction systems, mandatory notifications, and instant exit mechanisms. Major apps like Doubao and Qwen have shut down companion features to comply, sparking user backlash over data loss.

This article presents a proposal called TLBIC (Town-Level Basic Income Coin) that leverages AI to address despair through localized basic income. Full details are available in the attached PDF.

SvelteChatKit is a universal, provider-agnostic AI chat UI kit for SvelteKit that supports OpenAI, Ollama, Dify, and custom backends through a unified interface, featuring streaming, persistence, markdown rendering, and a zero-config demo mode.

This article describes how to use AI assistants like Gemini and Kilo Code to quickly find API endpoints, parameters, and whether an API key is required. It provides an example using the CoinGecko API to fetch Bitcoin price in ZAR on the VALR exchange, showing curl commands for both free and pro tiers. The author advocates relying on AI over manual doc reading, with a security note about using environment variables for API keys.

A private, intelligent AI journal app that supports text, voice, links, and more. Automatically categorizes and tags entries, syncs across platforms, and allows data export or deletion anytime. Your data is never used for AI training.

Sakana AI launches Sakana Translate, a new feature in Sakana Chat powered by Namazu, its model series adapted for Japanese. It handles bidirectional translation between Japanese, English, and Chinese with three modes: Translate, Proofread, and Ask. Available for free as a web app.

PAI is a personal assistant AI for Mac that follows Unix/Linux philosophy, drops LLM into terminal, avoids polling, and manages calendars and emails.

This paper from NUS and UC Berkeley explores the near-term possibility of AI agents that can economically sustain and replicate themselves without human intervention. It defines self-sovereign agents (SSAs) and outlines three core mechanisms: economic, replication, and adaptation loops. The authors discuss societal implications including economic displacement, security risks, and governance challenges, and propose a four-level roadmap for assessing progress toward fully self-sovereign systems.

Venice AI raises $65M Series A, achieving unicorn status. Its privacy-first platform allows users to freely use AI models without restrictions while preserving privacy.