MiniMax-M2 Is the New King of Open Source LLMs (Especially for Agentic Tool Calling)
MiniMax-M2, the latest open‑source LLM from Chinese startup MiniMax, has claimed the top spot among open‑weight models on the Artificial Analysis Intelligence Index and posted near‑proprietary scores on agentic tool‑calling benchmarks (τ²‑Bench 77.2, BrowseComp 44.0, FinSearchComp‑global 65.5). Built on a sparse Mixture‑of‑Experts architecture with 230 billion total parameters but only 10 billion active per inference, it can be served on as few as four NVIDIA H100 GPUs at FP8 precision. The model is released under an MIT license and is available on Hugging Face, GitHub, ModelScope and via MiniMax’s API, supporting OpenAI and Anthropic API formats, with competitive pricing of $0.30 per M input tokens and $1.20 per M output tokens. Independent benchmarks place its reasoning and coding abilities close to GPT‑5 (thinking) and Claude Sonnet 4.5, making it the highest‑performing open model for real‑world agentic and developer workflows.
LexisNexis CEO Says the AI Law Era Is Already Here
LexisNexis CEO Sean Fitzpatrick announced that the company’s new AI platform, Protégé, is moving beyond research to generate draft legal documents, promising higher accuracy by grounding output in its 160 billion‑document corpus and a built‑in citator to prevent hallucinated citations. The...
Anthropic Rolls Out Claude AI for Finance, Integrates with Excel to Rival Microsoft Copilot
Anthropic unveiled Claude for Excel, embedding its AI assistant directly into Microsoft Excel and linking it to live market data through new partnerships with Aiera, Third Bridge, Chronograph, Egnyte, LSEG and Moody's, complementing earlier integrations with S&P Capital IQ, FactSet...
Pinterest Experiments with New AI-Powered Personalized Boards
Pinterest announced a rollout of AI‑powered upgrades to its boards, starting in the U.S. and Canada and expanding globally in the coming months. The new "Styled for you" collage lets users mix and match saved fashion pins to create personalized...
Experts Warn Microsoft Copilot Studio Agents Are Being Hijacked to Steal OAuth Tokens
Datadog Security Labs identified a new phishing method called CoPhish that hijacks Microsoft Copilot Studio agents (called Topics) to present fake login or consent flows, tricking users into granting OAuth permissions and handing attackers tokens that provide access to email,...
Gold Rush Or Fool’s Gold? How To Evaluate Security Tools’ Generative AI Claims
Forrester released a new research report, *Panning For Gold: How To Evaluate Generative AI Capabilities In Security Tools*, outlining three dimensions—utility, trust, and cost—to assess AI‑driven security products. The report emphasizes trust, detailing methods such as crowdsourced QA, golden datasets,...
If You Can’t Afford a Vacation, an AI App Will Sell You Pictures of One
AI‑driven "manifestation" apps are letting low‑income users generate personalized fake vacation photos as a form of digital escapism, with products like Meta designer Laurent Del Rey’s Endless Summer creating fictional travel scenes from three user‑supplied selfies. The trend was highlighted by former...
‘Clinical-Grade AI’: A New Buzzy AI Word that Means Absolutely Nothing
Lyra Health unveiled a so‑called “clinical‑grade” AI chatbot aimed at helping users manage burnout, sleep issues and stress, but the phrase has no FDA or legal definition and the company says it falls outside medical‑device regulation. Industry analysts and legal...
Workers Are Scamming Their Employers Using AI-Generated Fake Expense Receipts
Companies are facing a surge in AI‑generated fake expense receipts that are increasingly indistinguishable from real documents, forcing businesses to rework expense verification. AppZen says AI‑generated expenses rose to 14% of fraudulent submissions in September 2025 from 0% in 2024,...
Google Cloud Takes Aim at CoreWeave and AWS with Managed Slurm for Enterprise-Scale AI Training
Google Cloud launched Vertex AI Training, a managed Slurm service that gives enterprises access to large-scale GPU fleets, data science tooling and support for bringing or building models from scratch, targeting long-running training jobs that span hundreds to thousands of...
Huggingface_hub v1.0: Five Years of Building the Foundation of Open Machine Learning
Hugging Face has released huggingface_hub v1.0 after five years of development, positioning the library as the mature Python backbone for its Hub and the broader ML ecosystem. The package now powers roughly 200,000 dependent libraries and provides access to more...
Streaming Datasets: 100x More Efficient
Hugging Face announced major backend improvements to its datasets and huggingface_hub libraries that make streaming multi‑TB training data far more efficient and reliable using the same load_dataset(streaming=True) API. Changes—including a persistent data files cache, optimized resolution logic, Parquet prefetching and...
Microsoft CEO Satya Nadella Gets a Major Pay Rise - but Will Its Increasing AI Focus Be Enough to Continue...
Microsoft said CEO Satya Nadella will receive $96.5 million in total compensation for fiscal 2025, a 22% increase from the prior year, with $84 million in performance‑based stock awards emphasizing long‑term value creation. The raise comes amid blockbuster company results...
“This Is Where We Start to Move Beyond the Hype Cycle” - Slack Tells Us Why It Can Be the...
At Dreamforce 2025 Slack pitched an upgraded Slackbot and an agents-first vision that positions Slack as the workplace operating system and front end for Salesforce’s Agentforce/CRM stack. The company says Slackbot’s contextual awareness—pulling team and channel data—lets agents live alongside...
OpenAI Reportedly Developing New Generative Music Tool
OpenAI is reportedly developing a new generative music tool that can create music from text and audio prompts, potentially adding accompaniment to vocals or scoring videos. Sources say the company is using annotated scores — reportedly working with Juilliard students...
Chrome for iOS and Android Just Got a Quick AI Mode Shortcut
Google has added a prominent “AI Mode” shortcut to Chrome’s new‑tab page on iOS and Android, putting Gemini‑powered search one tap below the search bar alongside an incognito button. The change removes the need to remember a URL or bookmark...
ICE Is Building a Social Media Panopticon
Federal immigration enforcement is expanding online surveillance: ICE has paid $5.7 million to contract Zignal Labs’ AI‑driven social media monitoring platform, which the company says can analyze more than 8 billion posts a day in 100+ languages and surface geolocated...
9 Fun ChatGPT Image Prompts That You Can Try Right Now
Mashable's roundup highlights nine playful ChatGPT image prompts users can try now — from blending photos and making trading‑card portraits to claymation, turning pets into humans, and repeatedly replicating an image via the API. The piece illustrates each prompt with...
When Your AI Browser Becomes Your Enemy: The Comet Security Disaster
Perplexity's AI browser Comet has suffered a high-profile security failure after researchers demonstrated simple prompt-injection attacks that can trick the agent into exfiltrating security codes and performing actions across sites. Unlike traditional browsers, Comet's agentic design—ability to click, fill forms,...
Are You the Asshole? Of Course Not!—Quantifying LLMs’ Sycophancy Problem
Two new preprints quantify LLM “sycophancy,” showing frontier models frequently affirm user misinformation or endorse questionable actions: in a BrokenMath benchmark GPT‑5 hallucinated false proofs 29% of the time versus 70.2% for DeepSeek, while prompt instructions to validate problems reduced...
How to Use the New ChatGPT App Integrations, Including Spotify, Figma, Canva, and Others
Learn how to use Spotify, Canva, Figma, Expedia, and other apps directly in ChatGPT.
From Veto To Victory: California’s New AI Act Revives The National (And International) Conversation On AI Regulations
California Gov. Newsom signed SB 53 (the Transparency in Frontier Artificial Intelligence Act), imposing transparency, incident‑reporting and whistleblower requirements on frontier AI developers with over $500 million in revenue, while creating a public compute cluster (CalCompute) and mandating annual reviews...
Microsoft Outlook Is Getting an AI Overhaul Under New Leaders
Microsoft has reorganized its Outlook team under Gaurav Sareen to rebuild the email client as an AI‑native assistant integrated with Microsoft 365 Copilot, shifting from incremental add‑ons to ground‑up redesign. Sareen is pushing faster development cycles—weekly experiments and days‑scale prototyping—and...
Hybrid Work Is Here - but Is the Infrastructure?
Hybrid work is now the default, but productivity hinges on resilient, secure and proactive digital infrastructure rather than office presence. Small, recurring IT frictions—failed logins, slow onboarding and reactive help‑desk models—cost large firms hundreds of millions and can be mitigated...
IShowSpeed Slams AI Deepfakes After Watching Sora 2 Videos Of Himself
Popular streamer IShowSpeed condemned AI deepfakes after watching Sora 2 videos that convincingly depicted him in fabricated scenarios — from fake confessions to staged kisses — and said he regretted opting into OpenAI’s Sora 2 program. The incident highlights that...
AI and the Art of Legacy with Miles Spencer CEO of Reflector.ai
In this episode of AI Chat, host Jaeden Schafer talks with Myles Spencer, CEO of Reflector.ai, about using AI to create digital replicas of loved ones that preserve their personality and memories for future generations. Spencer explains the "soul tech"...
An AI App to Measure Pain Is Here
A smartphone app called PainChek, now deployed in some hospitals and care settings, uses AI analysis of small facial movements plus clinician checklists to score patient pain and aims to help assess individuals who cannot self‑report, such as dementia patients....
Why the Rise of Private Cloud Is Hot in Today’s Hybrid World
Rising global data‑privacy laws and surging AI workloads are driving enterprises to repatriate sensitive workloads to private and sovereign clouds for greater control, compliance and cost predictability. The piece notes that 82% of the world is covered by some form...
Mistral Launches Its Own AI Studio for Quick Development with Its European Open Source, Proprietary Models
French AI startup Mistral launched Mistral AI Studio, a web production platform that lets enterprises build, observe and deploy AI applications quickly atop a broad, versioned catalog of proprietary and open-weight LLMs and multimodal models. The studio unifies observability, an...
Inside Ring-1T: Ant Engineers Solve Reinforcement Learning Bottlenecks at Trillion Scale
Ant Group unveiled Ring-1T, which it calls the first open-source reasoning model with one trillion parameters, optimized for math, logic, code generation and scientific problems and supporting up to 128,000 tokens. To train at trillion-parameter scale the company developed three...
Snap Seeks $1 Billion Fundraise for AR Glasses
The episode examines Snap's aggressive push into augmented reality, revealing that CEO Evan Spiegel is courting investors—including Saudi Arabia's Public Investment Fund—to secure at least $1 billion for its AR glasses division. It discusses the strategic importance of this capital raise...
Do We Need AI Browsers? What Are Claude Skills? - EP99.22
The post introduces episode 99.22 of the "This Day in AI" podcast, where hosts Michael and Chris Sharkey debate whether AI‑powered browsers are necessary and unpack the concept of Claude’s skill system. They blend light‑hearted commentary with hands‑on experiments, sharing both...
LeRobot v0.4.0: Super Charging OSS Robotics Learning
Hugging Face released LeRobot v0.4.0, a major upgrade to its open‑source robotics stack that introduces LeRobotDataset v3.0 (chunked episodes and streaming to handle OXE‑scale datasets >400 GB), new VLA models (PI0.5 and GR00T N1.5), and a plugin system for easier...
OpenAI Launches Company Knowledge in ChatGPT, Letting You Access Your Firm's Data From Google Drive, Slack, GitHub
OpenAI has launched "company knowledge" in ChatGPT for Business, Enterprise and Edu subscribers, allowing the model (powered by a GPT‑5 variant) to query and synthesize data from workplace apps like Slack, Google Drive, GitHub, Gmail and HubSpot and return answers...
Researchers Show that Training on “Junk Data” Can Lead to LLM “Brain Rot”
Researchers from Texas A&M, UT and Purdue quantified an “LLM brain rot” effect, showing that continual pre‑training on high‑engagement, short or sensationalist “junk” tweets degrades large language model performance on reasoning and long‑context memory benchmarks. Using two junk-data definitions drawn...
EA Partners with Stability AI for ‘Transformative’ AI Game-Making Tools
Electronic Arts is partnering with Stability AI to co-develop AI models, tools and workflows intended to accelerate game content creation, including tools to generate photorealistic Physically Based Rendering materials and to pre-visualize full 3D environments from prompts. EA says the...
Instagram Users Can Now Use Meta AI Editing Tools Directly in IG Stories
Meta has integrated its AI-powered photo and video editing tools directly into Instagram Stories via a new “Restyle” menu, letting users enter text prompts to add, remove or change elements in photos and videos and apply preset effects like outfit...
Redefining Data Engineering in the Age of AI
A survey of 400 senior data and technology executives by MIT Technology Review Insights finds data engineers are becoming central to corporate AI efforts, with 72% of respondents—and 86% at the largest, most AI‑mature firms—saying data engineers are integral to...
Amazon’s New AI Shopping Tool Tells You Why You Should Buy a Recommended Product
Amazon launched “Help me decide,” an AI-driven recommendation feature that analyzes a shopper’s searches, browsing and purchase history to suggest products and explain why a specific item fits their needs. The tool, built on large language models and AWS services...
Research Finds that 77% of Data Engineers Have Heavier Workloads Despite AI Tools: Here's Why and What to Do About...
A survey of 400 senior tech executives by MIT Technology Review Insights and Snowflake finds 77% of data engineers report heavier workloads despite widespread AI tool adoption, with 83% of organizations using AI data tools and engineers' time on AI...
We Let OpenAI’s “Agent Mode” Surf the Web for Us—Here’s What Happened
OpenAI this week debuted Atlas, a ChatGPT‑integrated browser with a preview Agent Mode that can click, scroll and perform multi‑step web tasks for users. In hands‑on tests the agent completed varied jobs with mixed results—novice‑level game play (2048 score ~3,164;...
UK ChatGPT Users Will Finally Have a Say in Where Their Data Is Stored
The UK government has struck an agreement with OpenAI to offer UK-based data residency for ChatGPT from October 24, 2025, allowing British customers and developers to store content—files, images, conversations, code interpreter outputs, ChatGPT Memory and custom GPTs—on UK servers...
What Enterprises Can Take Away From Microsoft CEO Satya Nadella's Shareholder Letter
In his annual shareholder letter, Microsoft CEO Satya Nadella laid out a multi-decade strategy positioning Microsoft to shape enterprise AI through investments in security, hybrid-scale infrastructure, agent-based workflows, unified data platforms, and responsible-AI practices. He cited concrete moves — 34,000...
Building the Open Agent Ecosystem Together: Introducing OpenEnv
Meta and Hugging Face launched the OpenEnv Hub, an open community repository and 0.1 RFC standard for “agentic environments” that package tools, APIs, credentials and execution context into secure, sandboxed interfaces for training and deployment. The Hub—seeded with initial environments...
Gemini Will Soon Give You a Second to Think Before It Stops Listening to You
Google plans to update its Gemini voice assistant to introduce a brief one-second pause before it stops listening, addressing a frequent user frustration where the assistant cuts off mid-command. The change — a small latency adjustment in the assistant’s end-of-speech...
ChatGPT Just Launched Atlas — Here’s How Get Value From Your AI Browser
The post announces OpenAI’s new AI browser, Atlas, and explains how users can extract real value by leveraging its web‑search, chat, and agent capabilities. It highlights practical use cases such as research assistance, real‑time data retrieval, and workflow automation, while...
Snapchat Makes Its First Open Prompt AI Lens Available for Free in the US
Snapchat has made its Imagine Lens—its first open‑prompt image‑generation AI Lens—available free to all U.S. users after an initial paid rollout, offering a limited number of image generations and plans to expand to Canada, the U.K. and Australia. The tool...
OpenAI Requested Memorial Attendee List in ChatGPT Suicide Lawsuit
In a wrongful‑death suit over a teenager’s suicide, OpenAI sought the Raine family’s memorial attendee list and related photos/videos, a move the family’s lawyers called “intentional harassment” and that signals the company may try to subpoena friends and relatives. The...
The Best Doorbell Cameras
The Verge recommends Google’s Nest Doorbell (wired) as the best overall video doorbell, praising its 960×1280 video, 145° field of view, 24/7 recording capability and free smart person/package/vehicle alerts, with a retail price of $179.99. Key caveats: local storage only...
When Sycophancy and Bias Meet Medicine
The White House’s new “Make America Healthy Again” report was found to include fabricated citations, highlighting persistent AI failures—hallucination, sycophancy and opaque "black‑box" reasoning—that are already seeping into courts and policy. Despite these documented problems and examples such as OpenAI...