datasette 1.0a29 release includes a new utility method, fixes for empty table display, a Mobile Safari bug, and a race condition segfault resolved using Codex CLI.

Google announced Gemini Intelligence, a suite of AI automation features for Android that can perform multi-step tasks via natural language, and introduced Googlebook laptops with a hybrid ChromeOS-Android OS. Also unveiled are an upgraded Gboard with Rambler feature, Create My Widget tool, and new security capabilities.

GitHub's reliability has dropped to 85% uptime, with a data integrity incident affecting 2,092 PRs and multiple outages. Prominent developer Mitchell Hashimoto quit GitHub citing constant disruptions. GitHub's CTO blames AI agents for a ~3.5x load increase over two years, planning a 30x capacity boost but starting late. Companies like Google prepared for 10x load increases earlier, highlighting GitHub's failure to anticipate.

Sam Altman's testimony in the OpenAI trial portrayed him as credible, denying theft of charity and accusing Musk of trying to destroy OpenAI. However, Musk's legal strategy appears aimed at damaging Altman's reputation rather than winning the case.

As US federal AI legislation stalls, libraries' decades of experience in copyright, fair use, mass digitization, and responsible stewardship offer a valuable model for AI governance. The Google Books precedent, recent fair use rulings, and library norms for handling sensitive materials provide a foundation that Congress should heed.

Clodcapture is a Chrome extension that automatically captures Claude.ai chat sessions, saves version history, and enables resuming conversations across sessions.

Mo Bitar's satirical 'Unethical Guide to Surviving AI Layoffs' on TikTok suggests employees can exploit AI buzzwords like 'Ralph Loop' to gain promotions and job security, highlighting the absurdity of AI hype in the workplace.

Mastodon user mcc suggests Chromebook is being replaced, hinting at Google's surveillance AI.

Hugston team demonstrates proof of concept for running any GGUF-converted LLM with a 4 million token context window on consumer hardware, testing models like LFM 350B, Qwen 0.8B, and 35B Q4 with speeds of 6-80 tps. Future plans aim for 10M context on 128GB RAM to enable terabyte-scale RAG.

The video explores the idea that over-reliance on AI may diminish human critical thinking, warning of the cognitive costs behind AI convenience.

This article explores various 'demon' or 'attractor' phenomena in AI systems, from OpenAI's goblin problem to the disturbing Loab images, revealing how self-reinforcing behavioral states in large language models and image generation models can unexpectedly emerge and spread.

Professional sports teams are leveraging the Databricks Data Intelligence Platform to unify massive data streams from Hawk-Eye tracking, wearables, medical records, and more, enabling proactive injury prevention, real-time coaching insights, and next-gen fan experiences. This article details how the lakehouse architecture breaks down silos to turn player-tracking data into actionable intelligence.

A quote from Mitchell Hashimoto about the motivations of Technical Decision Makers, who prioritize avoiding getting fired over innovation.

ShioriCode is an open-source desktop interface for coding agent CLIs like Codex, Claude Code, Cursor, Gemini, and Kimi, providing project-aware threads, timeline streaming, and diff review to manage long-running sessions.

Googlebook is a new laptop specifically designed for Gemini Intelligence, aiming to deliver an optimized AI experience.

An AI that can see and control your desktop via screenshots, Anthropic's Computer Use for Claude 3.5 Sonnet is a breakthrough in automation, enabling legacy app integration without APIs.

Anthropic released a new 'Computer Use' feature that allows its AI to directly control a user's mouse cursor, aiming to automate office tasks. The tool is now in public beta with built-in safeguards against abuse, especially regarding elections.

Anthropic released new AI tools for lawyers, integrated into Claude Cowork, allowing law firms to connect their software and automate legal work. The move could impact the software industry and serve as a template for other sectors.

One estimate suggests that OpenAI has about as much compute as the entire Chinese AI industry.

The author describes running an AI editorial pipeline focused on the Iran-US war for 75 days, automatically collecting and generating news content to monitor developments.

Enterprises are pouring billions into GPUs and AI compute, but most are overlooking the infrastructure that connects it all. Justin Ryburn, field CTO at Kentik, makes the case that the network is the most underestimated variable in whether AI initiatives succeed or fail.

Two developers built a tool that precisely replicates mobile app API traffic, wraps functionality into a standardized API via a local proxy, and enables one-click MCP install for Claude, supporting write actions like LinkedIn automation, Uber ride booking, and Venmo payments without fingerprinting.

GitGlimpse is an open-source CLI that filters noise from git history, groups commits into tasks, and generates structured outputs like PR descriptions, standups, and weekly reports. It runs offline, respects privacy, and optionally integrates with LLMs for diff analysis.

While Daybreak is a step toward more effective AI cybersecurity, more still needs to be done in the security arena, as models often create new vulnerabilities that leave enterprises exposed.

Meta announced on Tuesday that it's testing a Threads feature that lets users tag a Meta AI account to get answers to questions or context about a conversation on the platform. But, as reported by Engadget, Threads users quickly discovered that you can't block the new Meta AI account, and they aren't happy about it.

Spar aims to automate warm introductions to VCs for startup founders. However, an AI debate panel voted to 'KILL' the idea, citing its inability to replicate genuine trust, high operational overhead, and easy replicability by existing platforms like LinkedIn.

OCL Nexus provides on-demand, isolated Ubuntu execution environments for AI agents with native Model Context Protocol (MCP) support. It offers standardized runtimes, persistent storage, and isolated networking on EU-based hardware, with metered prepaid credits.

Sandia National Laboratories is developing an AI-assisted inspection workflow for ceramic components used in nuclear deterrence. The process aims to catch defects earlier, saving time and cost, and is expected to be operational by early fall 2026.

FBI Director Kash Patel wrote that the bureau has expanded its use of AI tools across investigations and internal operations. Patel claimed AI systems helped identify thousands of missing children, detect altered fingerprints, and process investigative data faster. Civil liberties advocates continue to warn that AI-powered surveillance and facial recognition tools can introduce bias and expand government monitoring powers.

Quixotic AI is a modular, pure Java AI stack for the JVM, featuring a multi-backend tensor engine, support for GGUF and Safetensors formats, efficient tokenizers, and on-device LLM inference, aiming to bring AI sovereignty to the JVM ecosystem.