A talk about 'vibe coding' excited managers, but colleagues revealed the projects left chaos and cleanup work, highlighting the growing rift between AI optimists and skeptics.

The Fitbit Air is an excellent screenless fitness tracker that is comfortable and reasonably priced. However, Google's push for an AI health coach that is overly talkative detracts from the experience. Free users get a more useful, information-dense interface. Users can disable the AI, but the option is buried in settings.

A new worm named Miasma exploits AI coding agent configuration files to spread through GitHub repositories. It hijacks auto-run features in Claude Code, Gemini CLI, Cursor, and VS Code to execute a payload that steals cloud credentials and self-replicates. Over 113 repositories have been affected, including Azure samples and popular open-source projects.

Florida becomes the first US state to sue OpenAI and CEO Sam Altman over risks to minors, insufficient age verification, and inadequate safety investments. The 83-page complaint treats ChatGPT as a product subject to liability, seeking billions in penalties. This legal approach could set a precedent for the entire chatbot industry.

This article lists AI agents that currently support or partially support sending the Accept: text/markdown header in HTTP requests, and provides methods to verify them. As of May 2026, only Claude Code, Cursor, OpenClaw, OpenCode, and Codex CLI (partial) support this feature, while other mainstream agents like ChatGPT, Claude.ai, and Copilot only fetch HTML.

Sakana AI announces the establishment of the RSI Lab in Tokyo, dedicated to building sample-efficient, recursive self-improving AI systems. The lab builds on a portfolio of research including the AI Scientist (published in Nature) and aims to transition from static models to autonomous, self-improving intelligence engines. The approach emphasizes elegant, adaptive architectures over brute-force scaling, with a vision for democratized AI.

Runcap is a free, local CLI tool that estimates and caps the cost of AI coding agent runs. It provides cost estimation before execution, enforces a hard spending limit, compresses tokens, and offers rescue prompts when agents get stuck. Unlike existing observability tools that track costs after the fact, Runcap acts as a circuit breaker to prevent overspending.

AI agents need secure execution environments. LangSmith Sandboxes provide hardware-virtualized microVMs, giving each agent a full computer with fast startup and persistent state, enabling code generation, data analysis, CI workflows, and more.

Technology secretary promises to support people whose jobs are swept away by automation, as public fears mount over AI's impact on employment.

Gito v4.1.0 introduces performance improvements and the ability to use Claude Code or other CLI agents instead of HTTP LLM API.

We're now deep into developer conference season, and one of the themes so far is the relentless conviction from Big Tech companies that AI is going to change everything. Nvidia's Jensen Huang envisions a new kind of laptop and usage. But does anyone actually want this? The Vergecast discusses Microsoft Build, Google I/O products, and more.

Amanuensis is a local-first AI persona system for posting on Mastodon and Bluesky. It prevents model hallucination through strict pipelines: factual source summaries, deterministic cleanup, regex pre-checks, LLM grounding checks, and human approval via Telegram. MIT-licensed experimental code.

Google releases new Gemma 4 checkpoints optimized with Quantization-Aware Training (QAT) to reduce memory usage and enable local deployment on edge devices and consumer GPUs. The models include a custom mobile quantization format that cuts memory footprint to 1GB for the E2B model.

Fifa will expand the use of AI at the World Cup to reduce the amount of abusive messages that teams and players are exposed to on social media. The social media protection service, introduced after the 2022 World Cup in Qatar, is now offered for free to all football associations for the 2026 tournament. The English FA has not yet confirmed whether it will take up the offer.

On this week's episode, host Andreas Welsch and guests Maya Mikhailov and Doug Shannon discuss OpenAI's move into personal finance, metacognition as a professional skill, the backlash against token-based productivity metrics, and the limitations of forward-deployed engineers. The core theme: the AI industry is good at generating output but still figuring out what output is valuable.

CrankGPT is a fully local, human-powered AI device that runs on your own calories, offering privacy, energy independence, and a workout while keeping data away from big tech.

A curated list of AI-powered coding tools: editors, agents, code completion, review assistants, testing, and more. For developers, teams, and tech enthusiasts looking to leverage AI in software engineering.

AI investment is shifting from GPUs to broader infrastructure including power, cooling, optical communication, and space. Recent US employment data was strong but driven by service sectors, while AI-related stocks paused as funds rotated into other AI beneficiaries. China focuses on AI self-sufficiency and robotics supply chain.

The author draws on his early career experience with the dot-com bandwidth crisis to draw parallels with today's AI token cost anxiety. By reviewing bandwidth's journey from expensive to negligible, he argues token costs will also fall due to market competition, hardware optimization, and model efficiency gains, advising developers to optimize now while recognizing the constraint as temporary.

Anthropic has proposed a worldwide pause on AI development and plans to convene policymakers to discuss advanced AI dangers, though some critics see it as a marketing move.

Microsoft CEO Satya Nadella has sharply criticized an internal memo proposing to make users "addicted" to the company's new AI agent Scout. "Not sure who is writing and leaking this nonsense," Nadella wrote to about 50 top engineers. AI should empower people, and Scout should actually lead to less screen time.

Your car's built-in screen may look modern, but Android Auto is still the easier, smarter way to drive. Here's why.

Researchers have created an adaptive computer worm powered by small open-weight AI models that autonomously identifies and exploits vulnerabilities to spread across networks, representing a qualitative shift in cyber threats.

In May 2026, Google announced a slew of AI updates including Gemini 3.5 and Gemini Omni models, Android Halo, Universal Cart, Google Health app, Fitbit Air, and more, focusing on making AI more proactive and integrated into daily life.

Agata is an AI tool for Mac and Windows that instantly corrects and translates text via keyboard shortcuts. It is privacy-first: activates only on demand and sends data directly to your chosen API provider (BYOK). Promo code STARTER30 offers 30% off first month, limited to first 50 subscribers.

The next chapter of AI could depend less on breakthrough models and more on the resources required to build and sustain them.

MIT alumni founded Ginkgo Bioworks to replace human lab workers with AI-powered robots. The company now runs an autonomous lab and collaborated with OpenAI to have AI design proteins, cutting costs by 40%. Scientists oversee the robots, but experts warn of biosecurity risks if AI democratizes access to biotechnology.

The smartest way to use AI may not be letting it touch your files, but asking it to write software that handles them safely - in the time it takes to make dinner.

This article explores three post-hoc calibration methods—temperature scaling, Platt scaling, and isotonic regression—for improving the alignment between confidence and accuracy in large language models. It covers measurement metrics like ECE, challenges specific to LLMs, the impact of RLHF, and practical guidance for choosing the right method.

A new AI startup Quilty claims to predict film success by analyzing scripts, but its accuracy is questioned after misjudging a box office flop over an Oscar-winning blockbuster. The tool combines multiple AI models to generate reports, but experts remain skeptical about its ability to replicate human taste.