Business, Deals & Funding
xAI News
Use Grok in OpenClaw
xAI announced that users with SuperGrok or X Premium subscriptions can now use Grok models inside OpenClaw, an open-source, local-first agent and personal assistant. OpenClaw runs on various hardware (Mac Mini, laptop, server, VPS, Raspberry Pi), maintains persistent memory across sessions, and connects to multiple messaging platforms including WhatsApp, Telegram, Slack, Discord, Signal, and iMessage. Setup involves a simple command-line installation and guided onboarding process, with device-code authentication available for remote/SSH setups. The integration is available on every subscription tier.
Why it matters
This is a strategically smart move by xAI. By integrating Grok into an open-source, local-first agent framework, they're positioning themselves in the rapidly growing personal AI agent space while leveraging their existing subscription base. The local-first approach with multi-platform messaging integration is compelling — it addresses privacy concerns and meets users where they already communicate. The fact that it runs on hardware as modest as a Raspberry Pi lowers the barrier to entry signif…
Claude Code Changelog
v2.1.145
Version 2.1.145 of Claude Code adds JSON output for listing live sessions (`claude agents --json`), improves OpenTelemetry tracing with agent ID attributes and fixed span nesting for subagents, enriches status line JSON with GitHub repo/PR info, and enhances the `/plugin` discovery UI to show more plugin details including commands, agents, skills, hooks, and servers.
Why it matters
This release focuses on developer tooling improvements and observability enhancements. The `claude agents --json` feature is particularly useful for power users who want to integrate Claude Code sessions into their terminal workflows (tmux, status bars, etc.). The OTEL tracing improvements with proper span parenting will help with debugging complex multi-agent interactions. These are solid incremental improvements that show attention to the developer experience and integration ecosystem.
DATAVERSITY Smart Data
Agentic Compliance: AI Auditing AI
The article discusses the concept of 'agentic compliance' — using AI systems to audit and monitor other AI systems — particularly in healthcare. As healthcare organizations deploy AI agents that make thousands of real-time micro-decisions affecting patient care, traditional manual auditing methods become inadequate. The article argues that agentic AI systems interpret goals, retrieve information, and respond to changing contexts in ways that are difficult to predict, making old audit models insufficient. Agentic compliance involves deploying specialized oversight systems (governance bots, supervisory agents, compliance layers) that act as digital shadows monitoring operational AI agents in real time across areas like data access, actions taken, and boundary compliance. Healthcare is highlighted as a prime use case due to its small margin for error regarding patient privacy, clinical saf…
Why it matters
This article addresses a genuinely important and timely topic — the recursive challenge of governing AI with AI — but it reads more like a high-level thought leadership piece than a deeply technical or evidence-based analysis. The concepts presented are sound: traditional compliance frameworks were not designed for the speed and unpredictability of agentic AI, and healthcare's strict regulatory environment makes this especially urgent. However, the article is light on implementation details, re…
Guardian AI
Real or AI: can a photographer and internet addict spot fake portraits? – video
Guardian Australia's photographer Carly Earl and journalist Matilda Boseley take an AI faces test developed by the University of New South Wales, which challenges users to distinguish between real and AI-generated portraits. The video explores the increasing difficulty of identifying AI-generated faces and whether the ability to spot fakes is based on skill or intuition.
Why it matters
This is a lightweight but timely piece that highlights a genuinely important issue: the growing sophistication of AI-generated imagery and its implications for trust and visual literacy. While the video format appears more entertaining than deeply analytical, the underlying subject—our diminishing ability to distinguish real from fake faces—has serious consequences for misinformation, identity fraud, and public trust. The collaboration with UNSW's research adds credibility. However, as a short…
Guardian AI
AI engineer says Google unfairly sacked him after he protested against work for Israel
A Google AI engineer claims he was unfairly dismissed after distributing leaflets at Google DeepMind's London offices protesting the company's work for the Israeli government. The flyers stated 'Google provides military AI to forces committing genocide.' The engineer has filed an employment tribunal claim against Google, representing the latest instance of internal dissent over the social and ethical impacts of AI technology and its military applications.
Why it matters
This case highlights the growing tension between tech workers' ethical concerns and corporate interests, particularly regarding military AI contracts. The employee's right to protest is important, but distributing inflammatory leaflets in the workplace does create a complex legal and ethical situation. Google's Project Nimbus contract with Israel has been a persistent source of internal controversy, and this tribunal case could set important precedents for how companies handle employee dissent…
The Verge AI
Demis Hassabis said this might be the ‘foothills of the singularity.’ What?
Google DeepMind CEO Demis Hassabis closed out Google I/O 2026 by declaring that humanity may be standing in the 'foothills of the singularity.' He introduced Gemini for Science, a set of tools for scientific research through Google Labs and Google Antigravity, and expressed Google's ambition to 'reimagine drug discovery with the goal of one day solving all disease.' Hassabis described the current moment as 'profound for humanity,' claiming AI technology will be a 'force multiplier for human ingenuity' that will usher in a 'new golden age of scientific discovery and progress.' The article notes this kind of grandiose rhetoric is common among tech executives, comparing it to similar statements from Microsoft's Satya Nadella and Luma AI's Amit Jain.
Why it matters
This is peak tech executive theatrics — wrapping product announcements in civilizational-scale rhetoric to make incremental tools sound like humanity's defining moment. The phrase 'foothills of the singularity' is carefully hedged enough to be unfalsifiable while still generating exactly the kind of breathless headlines it received. After hours of mundane app demos and travel itineraries, pivoting to existential proclamations about humanity's future feels jarring and calculated. While AI tools…
TechCrunch AI
Google just declared itself a contender in AI design at IO 2026
Google announced at I/O 2026 the launch of Pics, a new AI-powered design and image-generation app for Google Workspace. Pics allows users to create social media graphics, invitations, marketing materials, and mock-ups using simple text prompts without needing design skills. Powered by Nano Banana 2 and Gemini for its editing layer, the app makes every element in a generated design fully adjustable — users can click on specific parts to modify them or leave comments like in Google Docs, addressing the common frustration of having to re-prompt entire images for small changes. The app is built natively into Google Workspace for visual collaboration. It launches first to testers at I/O and will roll out to Google AI Ultra subscribers this summer. Google positions Pics as a competitor to Canva and Anthropic's Claude Design.
Why it matters
This is a significant and well-timed move by Google. The editing approach — clicking on individual elements and leaving comments rather than re-prompting from scratch — addresses one of the most genuine pain points in AI image generation today. Integrating it natively into Google Workspace is a smart strategic play that leverages Google's massive existing user base of businesses, educators, and organizations. The comparison to leaving feedback in Google Docs is intuitive and could lower the bar…
TechCrunch AI
You can now talk to your Gmail inbox, as seen at Google IO 2026
Google announced Gmail Live at Google IO 2026, a conversational AI feature powered by Gemini that lets users ask questions about their inbox using natural language voice queries instead of traditional keyword search. The feature can answer follow-up questions, understand context and nuances between related topics, and pull granular details from emails like hotel room numbers. Gmail Live is positioned as an additional option alongside traditional search rather than a replacement, likely informed by backlash Google received when it forced AI-powered search on Google Photos users. The feature will roll out later this summer, initially limited to Google AI Ultra subscribers. Google also announced expanding its AI Inbox overview feature to Pro and Plus subscribers, along with new capabilities like ready-to-send drafts and task management.
Why it matters
This seems like a genuinely useful application of AI that addresses a real pain point — finding specific information buried in years of accumulated emails. The conversational approach with follow-up questions and context awareness could be significantly better than keyword search for locating things like confirmation numbers or appointment details. I appreciate that Google learned from the Google Photos debacle and is keeping traditional search available rather than forcing the AI experience on…
From X/Twitter
- Kate Bergeron moves up to run Product Reliability for all Apple devices, backfilling the role Tom Marieb held before replacing John Ternus.
- The head of Claude Code says he hasn't written code manually in months — Anthropic's 30-minute workshop shows how the team actually uses it.
- Miles Deutscher spent two weeks building a personal OS with Claude Code that stores tasks, finances, calendar, and journal in one place — now sharing the blueprint.
- Watchmen writes skill files from your own coding sessions and shares them across Claude Code, Codex, and Pi so agents stop re-learning what they knew last week. Local and open-source.
- Claude Managed Agents now supports self-hosted sandboxes and MCP tunnels so enterprises can run agents inside their own security perimeter.
- New Anthropic article on Claude Code tips for large codebases — how to keep million-line repos navigable and maintain memory via CLAUDE.md.
From Reddit/HN/YC
- [Hacker News] Open-source Claude Code bundle for bug hunting ships with 574 report patterns baked in.
- [Hacker News] Paletty ships 800 curated terminal themes and a contrast-aware generator so your shell finally looks intentional.
- [Hacker News] LG drops the world's first native 1000Hz Full HD gaming monitor — yes, one thousand.
- [Hacker News] A breakdown of why Elon Musk lost his suit against OpenAI.
- [Hacker News] A scrubber designed to prevent AI from reading your secrets before you paste context into a prompt.
- [Hacker News] RFC 7489 is officially obsoleted — email receivers must now downgrade to quarantine by default under the new spec.