18
Saturday
65 items
- model
TestingCatalog: Kimi K3 Hits AI/ML API, Benchmarked Against Fable 5
Kimi K3 is available on AI/ML API alongside Claude Fable 5, with vibe-coding benchmarks showing K3 outperforming Fable 5 at a quarter of the price.
- product
TestingCatalog: Grok Heavy Now Bundles Free X Premium+
Grok Heavy subscribers now receive X Premium+ at no extra cost by linking their X account in the Grok app.
- product
TestingCatalog: ChatGPT Desktop App Syncs Chats and Projects Across Devices
OpenAI's ChatGPT desktop app now syncs chats and projects across desktop, web, and mobile, with consistent Chat/Work mode switching.
- product
TestingCatalog: Claude Managed Agents Adds Observability Tab
Claude Managed Agents now includes an Observability tab for monitoring token usage, error rates, and other metrics.
- product
TestingCatalog: Anthropic Workbench Gets New Build Section and UI
Anthropic's Workbench has been moved to a new Build section with updated UI, Messages API testing, and Claude Fable 5 fallback model configuration.
- product
TestingCatalog: NoimosAI Launches Advisor for Automated Marketing
NoimosAI released Advisor, an AI agent that reads marketing data, identifies performance gaps, proposes actions, and executes upon one-click approval.
- product
TestingCatalog: X Launches Dedicated AI Timeline for Model Releases
X has added a dedicated AI timeline option that users can pin to follow latest AI model releases, accessible via Add Timeline settings.
- product
TestingCatalog: Kimi K3 Gets Three Reasoning Effort Levels
Kimi K3 now offers Standard, High, and Max reasoning effort levels on web and mobile, along with a new K3 doodle.
- product
TestingCatalog: Claude Fable 5 Restricted to Max and Team Premium Plans
Starting July 20, Claude Fable 5 will only be included in Max and Team Premium plans at 50% limits; Pro users get a one-time $100 credit instead.
- model
TestingCatalog: xAI's 2T-Parameter Grok Model Nears Completion
xAI's next 2T-parameter Grok model is expected to finish training next week, aiming to surpass Kimi K3 with better speed and token efficiency.
- product
TestingCatalog: Anthropic Bumps Claude Code Limits 50% Through August 19
Anthropic is raising Claude Code weekly limits by 50% through August 19 for Pro, Max, Team, and seat-based Enterprise users.
- product
TestingCatalog: Anthropic Tests New Chat/Cowork Toggle and Projects Sidebar
Anthropic is testing a repositioned Chat/Cowork toggle and a dedicated Projects section in Claude's side navigation, not yet public.
- product
TestingCatalog: Google Building Native Skills Menu for Gemini Desktop
Google is developing a native Skills menu for Gemini desktop with upload, create, and edit capabilities, plus Skill folders.
- product
TestingCatalog: Google Preps Gemini Live and Skills for Desktop Launch
Google is testing Gemini Live for web/desktop rollout, and Skills are expected to become available in normal Gemini chats.
- platformsBaseten Announces Series F to Fuel AI Inference Platform Growth
AI infrastructure startup Baseten has announced a Series F funding round. The company serves billions of LLM inference calls weekly for customers including OpenEvidence and Speechify.
Read - platformsFireworks AI Powers Drive-Thru Voice AI for Major Fast-Food Chain
A global quick-service restaurant chain is using Fireworks AI's real-time voice intelligence to speed up orders and boost revenue, with plans to scale across thousands of locations.
Read - platformsFireworks AI Announces Series D Funding and $1B ARR Milestone
The AI inference platform has closed a Series D funding round and reports hitting a $1 billion annual revenue run rate.
Read - platformsFireworks AI Launches Partner Network with Major Cloud and Consulting Players
Fireworks AI has unveiled a partner program that brings together hyperscalers, consultancies, and AI tooling companies to help enterprises deploy generative AI at scale.
Read - platformsFireworks AI Launches Context-Aware Code Assistance for Enterprise Devs
Fireworks AI rolled out an AI code assistance platform built on its inference stack, promising sub-100ms streaming completions fine-tuned to a company's internal codebase. The company cites customer results from Cursor and Sourcegraph as validation.
Read - labsQwen Debuts Safety Guardrail Model and Expands AI Toolkit
Alibaba's Qwen team introduces Qwen3Guard, its first safety guardrail model for real-time AI moderation, alongside new image editing and translation tools.
Read - platformsFireworks AI Touts $1B ARR, Launches Partner Network With Major Cloud and Enterprise Players
The inference platform maker claims it has hit a billion dollars in annualized revenue and is rolling out a broad partner ecosystem spanning AWS, Google Cloud, NVIDIA, Deloitte, and more.
Read - labsOpenAI Outlines Strategies for AI Investments and Business Models in New Blog Series
OpenAI releases multiple blog posts on managing AI investments, articulating value models, and advancing agentic coding tools.
Read - labsAlibaba's Qwen team launches Qwen3Guard, a safety classifier for live AI outputs
Qwen3Guard is the first safety guardrail model in Alibaba's Qwen family, built on Qwen3 and designed to score risks on prompts and responses in real time across English, Chinese, and multilingual settings.
Read - labsOpenAI Sitemap Shows Widespread Business, Partner, and Academy Page Updates
A crawl of OpenAI's website reveals a burst of recent updates across enterprise partner pages, business solutions, Codex, and a new Academy section — but the source contains only URLs and timestamps, with no article text to report on.
Read - labsMistral unveils Robostral Navigate, an 8B model that steers robots using just a single RGB camera
Mistral has released Robostral Navigate, an 8-billion-parameter model that navigates robots through complex real-world environments using only plain RGB images and natural language instructions — no depth sensors or LiDAR required.
Read - labsOpenAI Blog Links Reveal Wave of Enterprise AI Case Studies and Internal Tools
A batch of recent posts from OpenAI's blog showcases how companies like Rakuten, Wayfair, and John Deere are deploying its models, alongside OpenAI's own internal AI assistants for sales, support, and contract work.
Read - labsOpenAI's Latest Blog Posts Show AI Tackling Diseases and Black Holes
OpenAI's blog features new articles on using AI to diagnose rare childhood diseases and simulate black holes, highlighting practical innovations in healthcare and science.
Read - labsMistral Adds Version Control and Governance for AI Prompts in Studio
Mistral now lets enterprises version, own, and trace every prompt and skill running in their AI, closing the gap between experimentation and production governance.
Read - labsMistral's Leanstral 1.5 Crushes Formal Verification Benchmarks, Finds Real Bugs
Leanstral 1.5 sets new state-of-the-art in formal verification, solving elite math problems and uncovering software bugs automatically. It's free and cost-effective, democratizing rigorous code analysis.
Read - labsMistral Overhauls Connectors With Admin Controls, Debugging, and Multi-Account Support
Mistral shipped a batch of connector upgrades aimed at making AI agents production-ready inside enterprises, including granular admin controls, scoped API keys, and a step-by-step debugger for broken connections.
Read - labsStability AI Brings Nine Image Editing Tools to Amazon Bedrock
Stability AI has launched a suite of nine professional image editing APIs on Amazon Bedrock, giving developers and enterprises granular control over creative workflows on AWS infrastructure.
Read - labsStability AI Publishes Transparency Report Detailing AI Safety Practices
The company's first major transparency report covers its approach to preventing misuse of image, video, 3D, and audio generation models, including CSAM detection, red teaming with law enforcement, and content provenance tagging.
Read - labsStability AI Ships Stable Audio 2.5, Targets Enterprise Sound Branding
Stability AI's latest audio model generates three-minute tracks in under two seconds on a GPU and adds audio inpainting, positioning it squarely for commercial creative teams.
Read - labsStability AI and NVIDIA Ship Optimized Stable Diffusion 3.5 NIM for Enterprises
Stability AI partnered with NVIDIA to release a NIM microservice for Stable Diffusion 3.5 Large, delivering 1.8x speed gains and consolidated deployment for enterprise teams.
Read - researchResearchers Launch First Benchmark for Debugging LLM Multi-Agent Systems
A team spanning Penn State, Duke, and Google DeepMind has defined the problem of automated failure attribution in multi-agent LLM systems and released the Who &When benchmark to tackle it.
Read - industryEnterprise AI is flying blind: four gaps exposed in new VentureBeat survey series
VentureBeat's multi-part survey of over 150 enterprises reveals that organizations are deploying AI agents and infrastructure faster than they can secure, evaluate, or even measure them.
Read - researchAI Agent Tops Kernel Benchmark With Single-Kernel Design, Outpacing Rivals by Wide Margin
Fable wrote a GPU kernel that achieved an 18.7x speedup over a PyTorch baseline, crushing competitors and signaling a leap in AI's ability to automate core research tasks.
Read - researchNVIDIA's ENPIRE Lets Robots Teach Themselves in the Real World
NVIDIA has built a software framework that lets physical robots autonomously experiment, fail, and improve without human intervention — a closed loop borrowed from coding agents, now applied to real hardware.
Read - researchAI Out-Persuades Expert Humans in Massive Study, Raising Societal Power Questions
New research shows AI systems reliably outperform even elite human persuaders in text-based conversations, including raising real money for charity, with implications for how influence will be distributed in society.
Read - researchAlignment Researchers Launch Nonprofit Sequent to Tackle Superintelligence Safety
Veterans from the UK AI Security Institute and startup Timaeus have formed Sequent, a nonprofit pursuing high-confidence alignment techniques for superintelligence, with plans to raise $100-150M initially.
Read - researchOpen AI Model Ecosystem Now Runs on Far More Than a Handful of Chinese Labs
A new analysis shows the open-model landscape has diversified sharply over the past year, with dozens of organizations across categories releasing models — and the trend is accelerating.
Read - researchZ.ai's GLM-5.2 stuns AI community, rivaling Claude and OpenAI on top benchmarks
The open-weight GLM-5.2 model is the first to crack into top-tier coding agent performance alongside Anthropic and OpenAI's best — and it landed right after the U.S. effectively banned Claude Fable 5.
Read - industryGoPro Max 2 Bundle Throws In $100 Worth of Accessories for Just $70 More
The GoPro Max 2 accessory bundle is now $369, packing a pole, extra battery, SD card, and mounts that would cost $75 plus on their own.
Read - industryGovernment Pilots AI for Medicare Prior Authorization, but Doctors Fear More Denials
A new federal AI pilot aims to cut waste in Medicare claims processing. Physicians worry it will just automate the denial of necessary care.
Read - industryGoogle-backed FireSat satellites launch to spot wildfires before they spiral
The first three FireSat satellites are now in orbit, promising to detect fires as small as a backyard before existing systems can see them.
Read - industrySan Francisco AG Demands Apple and Google Remove 13 Nudify Apps From Their Stores
San Francisco's attorney general sent cease-and-desist letters to Apple and Google, ordering them to pull 13 apps that use AI to generate non-consensual nude images of real people.
Read - industryWildfire Smoke Threatens World Cup Final — and FIFA Has No Air Quality Plan
As Canadian wildfire smoke blankets the East Coast, the Argentina-Spain World Cup final outside New York faces an air quality crisis FIFA never planned for. Other leagues have protocols; soccer's governing body does not.
Read - industryGoPro Max 2 Accessory Bundle Drops to $369 with $100 Worth of Extras
GoPro's Max 2 360-degree camera bundle is selling for $369 — just $70 more than the bare camera — and packs in an extension pole, extra battery, microSD card, and more.
Read - industryChina Launches Parallel AI Order with 29-Nation Group
China and 29 nations formed the World Artificial Intelligence Cooperation Organization, a bid to build AI governance outside Western influence, announced alongside training slots for the Global South.
Read - industryTrump Administration's AI-Driven Prior Authorization Pilot Raises Patient Access Concerns
A new government program uses AI to streamline Medicare coverage decisions, but critics warn it could lead to more wrongful denials of necessary care.
Read - industryGoogle-Backed FireSat Satellites Launch to Catch Small Wildfires Before They Spread
Three FireSat satellites built to detect small wildfires that other systems miss have reached orbit, backed by Google and the Bezos Earth Fund.
Read - industrySan Francisco AG Orders Apple and Google to Remove AI Nudify Apps From Their Stores
San Francisco's attorney general sent cease-and-desist letters to Apple and Google demanding removal of 13 nudification apps, calling the companies' profits from these tools illegal under California deepfake laws.
Read - industryOpen-Weight AI Models Are Closing the Cyber Gap With Frontier Systems, AISI Finds
The UK's AI Security Institute reports that open-weight models like GLM-5.2 and DeepSeek V4-Pro now match the cyber capabilities of closed frontier models from just four to seven months ago — down from a six-to-ten-month gap at the start of 2025.
Read - industryChina narrows AI gap with US as startup launches world's largest open model
A Chinese startup has released the world's largest open AI model, directly competing with US frontier systems and sending AI stocks tumbling.
Read - industryNavy Declares Itself 'AI-First,' Plans to Run Language Models Directly on Warships
The Department of the Navy signed a sweeping data-and-AI strategy that prioritizes speed over perfection, aiming to double its AI workforce by 2029 and deploy large language models on ships even when communications are jammed.
Read - industryAnthropic cuts Claude Fable 5 limits, pushes Pro users to API pricing
Anthropic is slashing Fable 5 usage in its subscription plans by up to 50 percent while nudging Pro and Team Standard users toward pay-per-use API pricing. The move comes as competitive pressure mounts from cheaper models like OpenAI's GPT-5.6 Sol.
Read - industryEU Orders Google to Share Search Data and Open Android to Rival AI
Brussels hits Google with binding DMA rules forcing it to share search data with competitors and allow deeper integration of rival AI assistants on Android.
Read - industryMusk Quietly Bought a $1 Billion Gas Turbine Company to Power Grok
Elon Musk acquired fossil fuel firm APR Energy in May, revealed via an FTC filing. The most likely purpose: fueling energy-hungry AI data centers.
Read - industryChinese Startup Drops World's Largest Open AI Model, Shaking US Stocks
Moonshot unveiled an open AI model that rivals Anthropic and OpenAI's offerings, sending AI and semiconductor stocks sliding and signaling China's aggressive open-source bet.
Read - industryMusk Quietly Bought a $1 Billion Gas Turbine Firm to Power Grok's AI Data Centers
Elon Musk's acquisition of fossil fuel company APR Energy, hidden until an FTC filing surfaced, signals a massive bet that AI's future runs on gas turbines.
Read - researchNo Story Found — Source Page Returns 404 Error
The provided link leads to an IEEE Spectrum 404 error page with no article content, only site navigation and a handful of trending-story teasers.
Read - researchSource page returned no article content
The provided source material is an IEEE Spectrum 404 error page with no substantive news content to report.
Read - researchSource Material Contains No Article Content
The provided source is an IEEE Spectrum 404 error page with no substantive news article to rewrite — only site navigation and headline teasers for unrelated stories.
Read - researchIEEE Spectrum 404 page recommends articles on tech failures
The IEEE Spectrum 404 error page highlights articles about historical tech failures, including VR goggles, a transatlantic cable, and a CIA spy drone.
Read - industryWired's tests find the Urevo CyberPad is the best walking pad for WFH workouts
Wired's remote team tested underdesk treadmills and named the Urevo CyberPad their top pick for its powerful incline and solid build, recommending walking pads as a tool for productivity and health while working from home.
Read