Fetching Updates
Alibaba's Qwen3.8-Max is a 2.4 trillion parameter multimodal model available in preview without published benchmarks, license, or standard pricing. The review advises testing it for curiosity but avoiding production use until independent evaluations appear.
Latest Coverage Stream
348 articlesAlibaba's Qwen3.8-Max is a 2.4 trillion parameter multimodal model available in preview without published benchmarks, license, or standard pricing. The review advises testing it for curiosity but avoiding production use until independent evaluations appear.
Cloudflare introduces @cloudflare/computer, an open-source agent runtime providing a virtual filesystem and execution environments (isolates and containers) for building scalable AI agents, allowing agents to select the optimal runtime for each task.
Google DeepMind announced Gemini Robotics 2, a family of vision-language-action models for controlling robots, including a full VLA model, an embodied reasoning model, and an on-device version, enabling advanced dexterity and multi-robot collaboration.
xAI launched Grok Voice Think Fast 2.0 on July 29, 2026, a speech-to-speech voice model with improved accuracy, sub-second latency, and pricing at $0.08 per audio minute. The article details its features, benchmarks, API access, and comparison with competitors like OpenAI and Gemini.
Fitch Ratings warned that the AI boom and risk of a correction are emerging as major global credit risks, citing soaring valuations and unprecedented AI spending that could trigger a significant market downturn with widespread macroeconomic implications.
Google's AI Overviews now appear in 43% of searches, up from 15% a year ago, according to Similarweb data, indicating AI-generated answers are becoming the default search experience. The shift is reducing referral traffic to publishers and changing how users interact with search.
Safe Superintelligence, Ilya Sutskever's AI lab, announced a long-term partnership with Nvidia, including a multi-billion dollar investment, to access Nvidia's Vera Rubin GPU platform and scale its research toward safe artificial superintelligence.
Chinese AI models such as Moonshot's Kimi K3 and DeepSeek are gaining popularity in the U.S. due to lower costs and competitive performance, challenging American AI leaders and sparking policy debates over technology transfer and open-source models.
Multiple AI companies signed an open letter urging US policymakers to avoid broad restrictions on open-weight AI models as Washington debates responses to alleged Chinese model distillation. The letter argues against conflating legitimate techniques like distillation with IP theft and warns that sweeping bans could sti
Meta has disabled an Instagram AI feature tied to its new Muse Image model that allowed users to generate AI images of public accounts simply by tagging them, after significant backlash. The feature, announced Tuesday, let anyone use public Instagram content in AI creations without the account owner's permission. Critics including the National Center on Sexual Exploitation warned it enabled sextortion. The Screen Actors Guild urged members to opt out. Meta acknowledged the feature missed the mark and removed it entirely.
Apple filed a lawsuit Friday against OpenAI in the U.S. District Court for the Northern District of California, alleging trade secret theft and breach of contract. The complaint targets Chief Hardware Officer Tang Tan, accusing him of using confidential project code names during recruiting and coaching departing employees to evade security. Apple also names Chang Liu, who allegedly kept an Apple laptop and downloaded confidential technical documents after joining OpenAI in 2026.
Meta has launched Muse Spark 1.1, a multimodal reasoning model from Meta Superintelligence Labs, now available via a public preview of the new Meta Model API for developers. The model features improvements in agentic tasks, computer use, coding, and multimodal understanding, with a 1 million token context window, and can orchestrate multi-agent systems and delegate to parallel subagents.
Anthropic and UST are partnering to deploy Claude across 20,000 of UST's engineers, architects, and consultants worldwide. UST is integrating Claude Code into its iDEC chip validation platform, where it reads hardware schematics, writes regression tests, and compares live equipment data against digital twins. The pipeline already cuts validation cycle times by 50 to 70%. UST is also embedding Claude into healthcare, telecom, and banking platforms.
OpenAI launched ChatGPT Work, an agent powered by the new GPT-5.6 model that can manage apps, files, and long-running projects for hours. It connects to tools like Slack, Microsoft Teams, Google Drive, Salesforce, and Gmail via plugins, and can produce slides, sheets, docs, and web apps. Available now for Pro, Enterprise, and Edu plans, with Plus and Business rollout coming within days.
Google has made AlphaEvolve generally available to all users. AlphaEvolve is a code optimization and discovery agent built on Google's Gemini model, designed to help organizations, researchers, and developers tackle difficult business problems. The tool uses AI to automate and improve code-based problem solving at scale, broadening access beyond an earlier limited release.
OpenAI has launched GPT-5.6, a family of three models: Sol (flagship), Terra (balanced), and Luna (cost-efficient). Sol sets state-of-the-art scores on Agents' Last Exam at 53.6, beating Claude Fable 5 by 13.1 points, and leads the Artificial Analysis Coding Agent Index at 80. A new 'ultra' setting coordinates four parallel agents for demanding tasks.
Anthropic has launched Reflect, a beta tool that lets Claude users track and visualize their AI usage patterns across the past 1, 3, 6, or 12 months. Available in Settings on Claude's web and desktop apps, the dashboard shows key topics, usage patterns, and task types, using a 4D AI Fluency Framework. The feature is available to Free, Pro, and Max users with Memory enabled.
Google Photos has launched Video Remix, a feature powered by the Gemini Omni model that lets users transform ordinary videos into stylized clips in seconds using templates. Available in the Create tab, it offers cinematic relighting, background swapping, and artistic effects. Video Remix is rolling out to eligible Google AI Plus, Pro, and Ultra subscribers in select countries.
OpenAI has launched GPT-Live, a new generation of voice models now powering ChatGPT Voice globally on iOS, Android, and ChatGPT.com. Built on a full-duplex architecture, GPT-Live can listen and speak simultaneously, enabling natural back-and-forth conversation while delegating complex tasks to GPT-5.5 in the background.
Mistral AI has launched Robostral Navigate, an 8B model enabling robots to autonomously navigate complex environments using only a single RGB camera and no depth sensors or LiDAR. It achieves 76.6% success on the R2R-CE unseen benchmark, beating the best multi-sensor systems by 4.5 points. It runs on wheeled, legged, and flying robots.
OpenAI has audited SWE-Bench Pro, a widely used coding benchmark, and found that approximately 30% of its 731 tasks are broken. Using a pipeline combining automated screening, Codex-based investigator agents, and review by experienced software engineers, OpenAI flagged 200 to 249 broken tasks. OpenAI is retracting its earlier recommendation to adopt SWE-Bench Pro.
xAI has announced Grok 4.5, a new model version described as featuring enhanced coding, agentic task handling, and knowledge work capabilities. The official announcement page returned an error during retrieval, so no specific benchmarks, pricing, or availability details could be verified.
OpenAI has published its National Security Principles, outlining how its AI technology can and cannot be used in government and defense contexts. Key restrictions include no mass domestic surveillance, no autonomous weapons direction, and no high-stakes automated decisions. OpenAI has also established Trusted Access for Cyber partnerships with nine allied nations and expanded access to its GPT-Rosalind model for biodefense missions.
Anthropic is expanding Claude Cowork to web and mobile, allowing agentic work sessions to continue across devices and in the background without any device online. Previously desktop-only, Cowork can now handle tasks like drafting memos, building client decks, and processing contracts while users are away. Over 90% of Cowork usage involves everyday knowledge work rather than coding.
Microsoft is replacing OpenAI and Anthropic models with its own internally built MAI models in Office apps like Excel and Outlook to cut costs. AI model chief Mustafa Suleyman stated the goal is to reduce and ultimately eliminate spending on Anthropic. Microsoft announced seven new MAI models at its Build conference in June.
Chinese authorities have held meetings with leading tech firms over the past month to discuss potentially restricting overseas access to China's most advanced AI models, including DeepSeek and unreleased models. Beijing is weighing the curbs amid rising geopolitical tensions, which could significantly limit global access to Chinese AI technology.
Meta has launched Muse, a new image generation model integrated into Meta AI. Muse supports a wide range of creative image tasks using personal photos, including photo restoration, style transformations, room restyling, product photography, and surreal scene edits, positioning it as a direct competitor to tools like Adobe Firefly and Google Imagen.
Anthropic researchers have identified a hidden mental workspace inside Claude called the J-space, discovered using a technique called the Jacobian lens. Unlike chain-of-thought reasoning, the J-space operates silently in Claude's neural activations, surfacing concepts the model is thinking but not saying. Researchers can read, edit, and inject patterns into it, allowing them to detect hidden goals.
Google DeepMind and A24 have announced a research partnership aimed at helping filmmakers develop new AI-powered workflows and techniques. The collaboration embeds Google DeepMind's innovations directly within A24's creative process, allowing filmmakers to shape emerging tools. Google has also made a financial investment in A24 as part of the deal.
Hi3D, formerly Hitem3D, has launched an end-to-end 3D print workflow that takes a text prompt and delivers a print-ready 3MF file in about five minutes. It uses Google's Nano Banana 2 model to generate 2D images, then converts them into watertight, manufacturable meshes, automatically segmenting models and optimizing orientations.
Microsoft has launched Frontier Company, a new operating business backed by a $2.5 billion investment, embedding 6,000 industry and engineering experts at customer sites to co-design and deploy AI systems. Early customers include LSEG, Land O'Lakes, Unilever, and Novo Nordisk. The platform supports models from OpenAI, Anthropic, and open-source providers.
Anthropic has detailed the cybersecurity safeguards and a new jailbreak severity framework for Claude Fable 5, now globally available. The model uses safety classifiers that sort requests into four categories: prohibited use, high-risk dual use, low-risk dual use, and benign use. Anthropic also launched a HackerOne program for researchers to submit discovered cyber jailbreaks for review.
Anthropic has added new analytics and cost controls to Claude Enterprise, giving admins deeper visibility into usage and spending. The updated admin dashboard shows usage and cost broken down by user and SCIM group. Spend-threshold alerts notify admins at 75% and 90% of org-level limits, while model-level entitlements let admins control which Claude models are available to specific roles.
Meta is developing a cloud infrastructure business called Meta Compute to sell excess AI computing power and model access to outside customers, directly competing with AWS, Microsoft Azure, and Google Cloud. Plans include an AWS Bedrock-style model hosting service featuring Meta's own Muse Spark models, plus raw compute capacity similar to CoreWeave.
Cloudflare is giving AI companies until September 15, 2026 to separate web crawlers used for search from those used for AI training and agents. After that deadline, default settings will block mixed-use crawlers from any pages hosting ads. Cloudflare is also expanding its Pay Per Crawl marketplace into Pay Per Use.
Anthropic is redeploying Claude Fable 5 on July 1 after US export controls imposed June 12 forced a suspension of both Fable 5 and Mythos 5. The controls followed researchers discovering a safeguard bypass that allowed the model to identify software vulnerabilities and demonstrate an exploit. A new safety classifier now blocks the reported technique in over 99% of cases.
Square has launched a ChatGPT app and Claude plugin letting U.S. Food and Beverage sellers accept orders placed directly inside AI chat interfaces, with no setup required and zero added marketplace commission fees. Orders sync with Square's live catalog and appear on existing POS and Kitchen Display Systems.
Anthropic's Claude Code team has published a guide defining agentic loops as agents repeating cycles of work until a stop condition is met. The guide covers four loop types: turn-based, goal-based, time-based, and proactive loops. The team recommends starting simple, using SKILL.md files for self-verification, and piloting before large runs.
OpenAI has launched GeneBench-Pro, a research-level benchmark testing AI agents on judgment-heavy computational biology tasks across 129 problems spanning 10 domains including statistical genetics, cancer genomics, and clinical diagnostics. OpenAI's strongest model, GPT-5.6 Sol, achieved a 28.7% pass rate, up sharply from below 5% when GPT-5 was tested on the original GeneBench.
Anthropic launched Claude Sonnet 5 on June 30, 2026, positioning it as its most agentic Sonnet model yet, with major improvements in coding, tool use, reasoning, and knowledge work. Performance approaches Opus 4.8 at a lower price. It is now the default model for Free and Pro plans and is also available in Claude Code.
Anthropic has launched Claude Science, an AI workbench for researchers, now in beta for Claude Pro, Max, Team, and Enterprise users on macOS and Linux. The app consolidates fragmented scientific tools into one environment, offering over 60 pre-configured skills and connectors for genomics, proteomics, single-cell analysis, and cheminformatics.
Google has launched two new generative media models: Nano Banana 2 Lite, its fastest and most cost-efficient image model at $0.034 per 1,000 images, and Gemini Omni Flash, a video generation and conversational editing model priced at $0.10 per second of output. Both are available in Google AI Studio, the Gemini API, and Gemini Enterprise Agent Platform.
OpenAI engineers have discovered optimizations that cut AI inference costs by more than half. When applied to ChatGPT for logged-out users, the techniques reduced the Nvidia GPUs needed to just a couple hundred. Possible methods include quantization, key-value caching, batch query processing, and query routing.
X has launched a hosted MCP server to connect AI apps with its API, making its platform easier for AI tools to use.
Google is rolling out updates to Gemini Spark, including a new macOS app that lets users automate desktop tasks like sorting PDFs into folders or building budget spreadsheets from local invoices. New app integrations include Canva, Dropbox, Instacart, OpenTable, and Zillow Rentals. Custom MCP support is also launching.
Anthropic has launched the Claude Apps Gateway, a new tool that connects Claude Code to Amazon Bedrock and Google Cloud. The gateway runs as a stateless container on your own infrastructure and provides enterprise features including corporate SSO login, centrally enforced policy controls, role-based access management, per-user cost attribution, and spend caps.
Google is expanding personalized image creation in its Gemini app to all eligible U.S. users for free. The feature uses Personal Intelligence to pull data from Gmail, Google Photos, YouTube, and Search with user permission. By connecting with Nano Banana and Google Photos, Gemini can generate tailored images from simple prompts using actual photos from your library.
Google has launched its Take notes for me feature in Google Meet for AI Pro and Ultra subscribers. Powered by Gemini, the tool transcribes conversations in real time, generates meeting summaries with key action items, and automatically saves notes to a Google Doc in Drive. All participants are notified when note-taking is enabled.
Anthropic's Claude models are now generally available in Microsoft Foundry, hosted on Azure. Claude Opus 4.8 and Claude Haiku 4.5 are accessible via the Messages API with features like prompt caching and extended thinking. The integration uses existing Azure identity, networking, and governance controls, with a US data zone option addressing data residency requirements.
HP Inc. has announced a Frontier strategic partnership with OpenAI, scaling beyond successful pilots launched in February 2026. Early results included one engineer completing 122 pull requests across 43 projects in weeks, and a security team remediating bugs in a day that would have taken a month. The partnership will deploy AI across customer support, partner portals, device fleet management, cybersecurity, and software development.
