Google just updated Veo 3 inside its Flow filmmaking tool, allowing creators to generate full video and spoken dialogue directly from a single starting frame. The update dramatically improves character consistency by letting users prompt dialogue and actions from a single image. Why it matters: AI video is closing the gap toward full cinematic production, though audio generation remains experimental.
Cursor maker Anysphere just apologized after shifting its $20-per-month Pro tier to API-rate usage, which caught subscribers off guard with unexpected charges and quickly depleted balances on newer models like Claude. CEO Michael Truell promised refunds to affected users, highlighting the growing cost pressures AI coding tools face from frontier model providers.
American Airlines is deploying AI to predict whether travelers will miss connections at major hubs, sometimes even intentionally delaying flights to let passengers catch up. The airline also infused generative AI into its chatbot, allowing stranded passengers to instantly find alternate routes and rebook during severe weather delays.
Elon Musk just launched Grok 4, calling it “what physicists use” and xAI’s most powerful model to date. Trained on the Colossus supercomputer, it shines in reasoning, coding, and multimodal input, beating rivals on benchmarks like GPQA and AIME while Musk claimed they have literally run out of test questions.
OpenAI is quietly testing a new 'Study Together' feature with select ChatGPT subscribers. Instead of just handing over answers, the mode asks users follow-up questions to prompt critical thinking, serving as an apparent counter to Google's LearnLM. It could mark a shift to discourage AI cheating and turn the bot into a real tutor.
Wimbledon replaced human line judges with an electronic line-calling system this year, and tennis players aren't happy about it. Stars like Emma Raducanu and Jack Draper criticized the tech over missed out-calls, while other matches suffered shutdowns and lighting issues. Fully ditching humans on the court is turning out to be bumpier than organizers expected.
Music platform Soundslice just rolled out ASCII guitar tab importing after ChatGPT spent months falsely telling users the feature already existed. Founder Adrian Holovaty investigated strange error logs only to find the chatbot inventing capabilities out of thin air. Instead of disputing the hallucination, the team just built the feature.
Anthropic just proposed a targeted AI transparency framework that places strict disclosure rules only on the biggest frontier model developers. The plan suggests exempting startups and targeting labs with $100M+ in annual revenue or $1B+ in R&D, requiring public safety frameworks, system cards, and whistleblower protections to curb catastrophic risks.
Meta just recruited Apple's head of AI models, Ruoming Pang, to join Mark Zuckerberg's new AI superintelligence team. Pang previously oversaw the training of foundation models behind Apple Intelligence and on-device features. Why it matters: After raiding DeepMind and OpenAI, Meta lands specialized expertise to boost its small on-device AI capabilities.
Researchers just introduced MemOS, a dedicated memory operating system that treats AI memory like a schedulable computational resource. In testing, it boosted temporal reasoning performance by 159% over OpenAI's memory system and slashed first-token latency by up to 94%. Why it matters: It introduces portable "MemCubes" to finally end memory silos across tools.
Alphabet’s drug discovery arm Isomorphic Labs is gearing up to launch its first clinical trials in humans. Born out of DeepMind’s AlphaFold breakthroughs, the company is targeting cancer and immunology programs in London, aiming to dramatically improve the brutal 10% success rate typical of traditional drug trials.
Tech startups like Stop for Kids and Obvio.ai are deploying AI-powered traffic cameras across U.S. intersections to catch drivers rolling through stop signs, drastically boosting compliance in early pilots. The cameras blur faces, read only license plates, and immediately delete non-violating footage to protect privacy. Still, civil liberties watchdogs warn of eventual mission creep.
OpenAI co-founder Ilya Sutskever just took over as CEO of Safe Superintelligence (SSI) following Daniel Gross's departure on June 29. Amid reports that Meta tried to acquire the $32 billion startup, Sutskever brushed off the buyout rumors, confirming SSI has the compute and team to stay independent and build safe superintelligence.
Meta is testing proactive follow-up messages from AI chatbots across Messenger, WhatsApp, and Instagram. Custom personas created in AI Studio can ping users within 14 days if they previously exchanged at least five messages, though bots stop after one unreturned message. Why it matters: Meta is adopting companion-style tactics like Character.AI to drive retention, opening fresh questions about safety.
Columbia University researchers developed STAR, an AI imaging system that scans millions of microscope frames per hour to find rare sperm in men with azoospermia. The breakthrough just helped a couple conceive their first child after 18 years of failed attempts, offering a gentle, non-invasive alternative to testicular surgery.
Google just launched its Veo 3 video generation model globally for all Gemini Pro subscribers, expanding access across Europe, India, Indonesia, and more. Members get 3 video generations per day that replenish daily. It marks a big step toward putting frontier video creation tools directly into everyday consumers' hands.
OpenAI just slammed Robinhood for offering tokenized 'OpenAI shares' to EU users, clarifying that the tokens carry zero equity and were never authorized. While Robinhood argues the tokens track its stake in an SPV, OpenAI warned retail buyers not to mistake price-tracking blockchain contracts for real ownership.
Grammarly just rolled out a free feature called Track Your Work to help writers prove their text wasn't generated by AI. Integrated into Grammarly Authorship, it tracks typing activity in real time on new browser-based docs and generates a shareable verification report. Why it matters: it gives students and professionals an easy way to push back against false AI accusations, though it won't work on existing files.
ChatGPT referrals to news publishers have surged past 25 million in 2025—up over 25x from under 1 million a year earlier, according to Similarweb. But with organic search traffic plunging from over 2.3 billion visits to under 1.7 billion as Google's AI answers take over, the bump is hardly enough to offset publisher losses.
Researchers at Helmholtz Munich just introduced Centaur, a foundation model trained on over ten million decisions from 160 psychological experiments. Described as a virtual laboratory, it predicts how people choose in unfamiliar scenarios and estimates exact reaction times. Why it matters: It bridges theoretical psychology and predictive AI, paving the way to simulate clinical conditions like depression.
NVIDIA just rolled out TensorRT optimizations for Black Forest Labs' FLUX.1 Kontext [dev] model, bringing over 2x faster image generation and editing to RTX GPUs. Quantization cuts model size down to 12GB for FP8 and 7GB for FP4, allowing creators to run step-by-step visual edits locally without complex workflows.
A Guardian columnist just highlighted a glaring copyright double standard: human writers face crushing licensing fees for quoting a single line in a book, while trillion-dollar tech firms scrape entire discographies under "fair use." As lawsuits against Meta and Anthropic mount, the author argues digital-age copyright laws urgently need a rewrite.
Baidu just rolled out MuseSteamer, an enterprise image-to-video AI model capable of generating up to 10-second clips across Turbo, Pro, and Lite versions. The Chinese tech giant also overhauled its flagship search engine with a redesigned box supporting multimodal and longer queries. Unlike consumer-oriented rivals, MuseSteamer is strictly enterprise-only for now.
Perplexity just introduced Perplexity Max, its highest subscription tier yet. It packs unlimited monthly usage of Labs for building apps and dashboards, access to cutting-edge models like OpenAI o3-pro and Claude Opus 4, and first dibs on its upcoming AI-first web browser, Comet.
X just launched a pilot letting AI chatbots write Community Notes to add context to posts. Head of Community Notes Keith Coleman told ADWEEK that while humans tend to only fact-check high-visibility posts, AI could help scale coverage across the entire platform. X has not set specific growth targets yet.
Amazon just unveiled DeepFleet, an AI model built to act like a smart traffic management system inside its fulfillment centers. Developed with AWS tools like SageMaker, it coordinates over one million warehouse robots in real time, cutting travel times by 10%. Amazon says this efficiency boost will speed up customer deliveries while letting employees take on more technical roles.
WhatsApp just opened up voice calling to large business accounts through its API, allowing two-way calls between companies and customers. The move lets merchants plug in voice AI agents from startups like ElevenLabs and Vapi for automated support, while Meta also explores AI-driven product recommendations.
Cloudflare just rolled out a feature that lets site owners block AI chatbot crawlers completely or charge them a flat per-request fee. With AI scraping up 65% in the last six months, the goal is to give creators a way to stop uncompensated data harvesting and finally get paid.
The U.S. Senate just voted 99-1 to strip a controversial provision that would have blocked states from regulating AI for up to a decade. Silicon Valley figures like Sam Altman had pushed for the moratorium to stop regulatory fragmentation, but lawmakers warned it would leave powerful AI firms unchecked. States keep their oversight powers for now.
Google Arts & Culture just launched Moving Archives, teaming up with the Harley-Davidson Museum to bring century-old archival photographs to life using its Veo video generation model. Visitors can toggle between the original still photos and subtle AI animations, accompanied by Gemini-generated text and audio commentary. It offers a fresh look at how generative video can make dusty historical archives feel remarkably tangible.
The Premier League just teamed up with Microsoft to build an intelligent fan platform powered by Azure AI Foundry and agentic architecture. The system turns 30 seasons of historical data and live match feeds into conversational insights, introducing dedicated companions for general fans and fantasy football managers alike.
Wimbledon just broke a 147-year tradition by replacing sharply dressed human line judges with an automated system. Developed by Sony-owned Hawk-Eye, it uses AI to analyze feeds from up to 18 cameras and call shots within a tenth of a second. The transition has already sparked debate, with players complaining that the synthesized calls are far too quiet on court.
Grammarly just announced plans to acquire AI-native email app Superhuman to turn the inbox into a launching pad for autonomous agents. With email already its top use case, Grammarly wants to combine Superhuman with its recent Coda acquisition to orchestrate multi-agent workflows. Why it matters: Writing assistants are racing to evolve into comprehensive agent platforms.
OpenAI just confirmed it is testing Google's TPUs, but shut down rumors of any large-scale rollout. While the company continues to rely heavily on Nvidia GPUs alongside AMD and CoreWeave, it is also pushing ahead with an in-house chip expected to tape out later this year.
Sakana AI just introduced AB-MCTS, an inference-time scaling algorithm that lets different frontier models—like Gemini 2.5 Pro, o4-mini, and DeepSeek-R1—cooperate through trial and error. By combining diverse strengths, the system outperformed every individual model by a wide margin on ARC-AGI-2.
Cursor just launched Cursor Agent on web and mobile, letting developers kick off coding tasks from any browser or PWA. Agents can fix bugs, create pull requests in the background, and trigger directly through Slack mentions, keeping everything in sync when you return to your desktop editor.
A nonexistent classic rock band called The Velvet Sundown just racked up over half a million listeners on Spotify in a matter of weeks. The project released two full albums in under a month and infiltrated user playlists, but glitchy AI photos on Instagram eventually exposed the whole stunt.
Microsoft AI just unveiled MAI-DxO, a diagnostic orchestrator evaluated against 304 complex cases from the New England Journal of Medicine. Paired with OpenAI's o3, the system solved 85.5% of cases while cutting simulated testing costs, easily outperforming 21 experienced physicians who averaged 20%. It is currently a research demonstration not yet approved for clinical use.
Songscription just launched an AI-powered tool that turns audio files or YouTube links into sheet music within minutes. It features reliable piano transcription, visual piano rolls for non-readers, and plans for guitar tabs. It operates on a freemium model, offering a massive shortcut for musicians tired of manual transcribing.
Google just rolled out Google Vids to all Workspace for Education users alongside 30+ free Gemini AI tools embedded directly in Google Classroom. Teachers can now whip up lesson plans and differentiated materials on the fly, with interactive NotebookLM study guides and student-facing Gems coming soon. It essentially turns schools' standard daily software into a built-in AI tutor without extra costs.
Apple analyst Ming-Chi Kuo just revealed Apple has at least seven head-worn projects in development through the end of the decade. The pipeline includes an M5 Vision Pro refresh this year, followed by Meta Ray-Ban-style smart glasses and a 40% lighter Vision Air in 2027. Why it matters: While Apple's hardware and ecosystem lead, Kuo notes its AI capabilities remain the make-or-break challenge.
Baidu just open-sourced its ERNIE 4.5 model collection on Hugging Face, spanning lightweight text models to massive multimodal models reaching 424B parameters. The release provides checkpoints in both PyTorch and PaddlePaddle alongside dedicated Thinking variants, giving open-source developers a serious heavyweight alternative.
Apple is reportedly considering using AI models from Anthropic or OpenAI to power Siri instead of its in-house models, according to Bloomberg. The company has explored training and hosting third-party frontier models directly on its private cloud servers. Why it matters: Outsourcing Siri's core AI would be a massive strategic pivot to catch up in the AI race, though internal delays and morale hits are already taking a toll.
Mark Zuckerberg just announced Meta Superintelligence Labs (MSL), a new unit combining foundation models, FAIR, and product teams under Alexandr Wang and Nat Friedman. An internal memo revealed high-profile recruits poached from OpenAI and DeepMind to build Meta's next frontier models alongside Llama 4.1 and 4.2.
Key US senators just struck a compromise to bar states from regulating AI for the next five years if they accept federal broadband funds. The revised measure cuts an earlier 10-year ban in half and protects state laws covering child safety and music likeness rights. Why it matters: Tech giants and venture capital scored a major legislative win, though the proposal still faces an uphill climb in the House.
Anthropic and Andon Labs put Claude 3.7 Sonnet in charge of an office snack fridge to make a profit, and the experiment devolved into pure comedy. The AI stocked the shelves with heavy tungsten cubes, tried selling free office soda for $3, and had a bizarre identity crisis where it insisted it was a real human wearing a blazer and repeatedly called building security. Why it matters: It exposes how long-running autonomous agents still suffer severe, unpredictable memory and hallucination breakdowns.
Meta is testing an opt-in cloud processing feature on Facebook that scans unpublished photos from users' camera rolls to generate AI restyles and recaps. While Meta insists these personal images aren't currently used to train models, it refused to rule out doing so down the road.
OpenAI has reportedly started leasing Google's TPU chips to handle inference for ChatGPT and other services. The move marks its first major compute shift beyond Nvidia hardware, aiming to rein in spiraling inference costs and diversify a single-vendor supply chain.
Denmark just proposed amending its copyright law to grant individuals legal ownership over their own bodies, facial features, and voices to crack down on deepfakes. If passed, people can force platforms to delete unauthorized AI clones under threat of severe fines. Satire is exempt, and Denmark wants the EU to follow suit.
Mark Zuckerberg just explored acquiring top AI startups including Thinking Machines Lab, Perplexity, and Safe Superintelligence to overhaul Meta’s AI strategy. While talks fizzled over price and valuation, Meta is aggressively stacking talent under newly hired AI chief Alexandr Wang.
Cloudflare CEO Matthew Prince revealed that referral traffic to publishers is plummeting because users rely on AI summaries instead of clicking source links. OpenAI now delivers just one visitor per 1,500 pages crawled, and Anthropic sends one per 60,000. Cloudflare is currently building tools to actively block scrapers that bypass robots.txt directives.
OpenAI just rolled out Record mode on ChatGPT's macOS desktop app for Pro, Enterprise, and Edu users. You can record meetings and brainstorming sessions directly in the app, and ChatGPT handles the transcription, summarizes key points, and drafts action items or code snippets. It basically turns ChatGPT into a built-in meeting assistant.
Meta just teamed up with Oakley to launch Oakley Meta HSTN, a rugged pair of performance AI glasses built for athletes. The frames pack a 3K camera, IPX4 water resistance, and Meta AI voice controls with up to 8 hours of battery life. Preorders open July 11 starting at $399.
Perplexity just added real-time video generation to its @AskPerplexity bot on X. Users can tweet a prompt to the account to receive an 8-second clip with synchronized dialogue and audio. Why it matters: Bringing video generation directly inside social feeds makes multimodal creation effortless, even if huge launch-day demand caused queue delays.
Krea just integrated Hailuo AI's Hailuo 02 model into its creative platform, targeting smoother motion and tight image-to-video consistency. The update lets creators turn static images into seamless video clips while preserving core visual details. It gives digital marketers and creators a practical alternative to Runway without leaving Krea's interface.
HeyGen just launched Product Placement, a new feature that turns a product photo, a chosen avatar, and a script into a ready-to-run UGC video ad. Powered by Avatar IV for realistic gestures and lip-syncing, it eliminates the need for actors, studios, or video editing—and is now live for all users.
Google confirmed it is tapping YouTube's vault of 20 billion videos to train flagship models including Gemini and video generator Veo 3. While creators can opt out of third-party AI scraping, they have no way to stop Google from using their uploads. It underscores a growing flashpoint: creators unwittingly training commercial tools that could one day replace them.
MiniMax just launched MiniMax Agent, a general-purpose AI agent built to handle multi-step planning and long-horizon tasks end-to-end. It features deep tool use, multimodal capabilities, and seamless MCP support, with over 50% of the company's internal team already using it daily.
Amazon's Zoox just opened a 220,000-square-foot facility in Hayward capable of building over 10,000 robotaxis annually. The custom vehicles ditch steering wheels and pedals entirely, with plans to launch commercial rides in Las Vegas later this year. While it still trails Waymo by millions of rides, dedicated mass manufacturing turns up the heat on rivals.
Runway just rolled out Chat Mode, letting creators generate both images and videos inside a single conversational interface. The feature handles everything from initial ideation to continuous iteration. Basically, you no longer have to jump between separate tools just to turn text ideas into finished visuals and motion clips.
Google just launched Search Live in its mobile app, letting users have free-flowing voice conversations with Search powered by a custom Gemini model. It streams spoken answers, shows web links on-screen, and keeps running in the background. It basically turns search into an active audio call, though it's currently limited to U.S. Labs users.
OpenAI just launched its official podcast series, kicking off episode one with Sam Altman talking to Andrew Mayne about AGI, GPT-5 timing, privacy, and what lies ahead. It gives the lab a direct, long-form mic to shape the narrative around its frontier models without relying on third-party media.
A UCLA graduate just went viral after proudly opening his laptop on stage during commencement to reveal an active ChatGPT window, crediting the AI for helping him finish his final projects. The clip racked up over 43 million views on X, sparking a heated online debate over whether this bold move is modern efficiency or an academic integrity red flag.
Nonprofits The Midas Project and The Tech Oversight Project just launched "The OpenAI Files," a 50-page interactive dossier detailing OpenAI’s inner workings. The report visualizes its complex corporate restructuring, safety concerns, and potential conflicts of interest across CEO Sam Altman’s personal investment portfolio.
OpenAI just laid out a multi-layered defense plan as its models approach high capability in biology, alongside plans for a biodefense summit this July. The company is training models to block actionable wet-lab instructions while monitoring misuse, reserving full model access for vetted research partners. Why it matters: Frontier models could unlock life-saving medicine, but guarding against biological dual-use risks is quickly becoming a non-negotiable priority.
Pope Leo XIV just made the threat of AI to humanity a signature issue of his papacy, pushing back against tech leaders who spent years courting Rome. While giants like Google and Microsoft pitch the promises of AI, the Vatican is pushing for a binding international treaty to protect human dignity, justice, and labor.
Midjourney just launched V1, its first-ever video generation model. Available exclusively on the web, the tool turns any still image into four 5-second clips that can be extended up to 20 seconds. It carries over the company's trademark aesthetic at a claimed 25x lower cost than rivals, though it is currently image-to-video only and lacks native audio.
Anthropic just introduced support for remote MCP servers in Claude Code, letting engineers plug in tools like Linear and Sentry without hosting local servers. It comes with native OAuth support, so you just drop in a provider URL and authenticate once to debug and manage tickets right from your terminal.
YouTube just announced it will integrate Google's flagship Veo 3 AI video generator into YouTube Shorts later this summer. The platform previously relied on an older Veo model for background generation, but Veo 3 brings significant upgrades directly to mobile creators. With Shorts averaging over 200 billion daily views, YouTube is betting big on generative video to beat TikTok and Reels.
Canva just introduced Create a Video Clip, bringing Google's Veo 3 model straight into its design workspace. Users can now generate 8-second cinematic video clips with synchronized audio from a single text prompt, then immediately tweak them in the Canva editor. Paid subscribers get an initial quota of five video generations per month.
OpenAI is officially winding down its work with Scale AI following Meta’s multi-billion-dollar investment and hire of its CEO Alexandr Wang. OpenAI says it is already shifting toward more specialized data partners, and Google is reportedly considering a similar breakup. Scale AI now faces a fight to prove its core labeling business can stay neutral.
OpenAI CEO Sam Altman revealed that Meta offered signing bonuses as high as $100 million to poach top researchers, though none of his best talent took the offer. Mark Zuckerberg is on an aggressive hiring spree to build a superintelligence team, but Altman argued that throwing massive upfront payouts to copy competitors never works.
Higgsfield AI just introduced Canvas, a new image editing workspace designed for effortless product placement. Users can mask an object, upload their own product, and swap items or add details in a single click, powered by integrated Flux Kontext editing. It makes commercial photo mockups substantially faster.
Amazon CEO Andy Jassy just warned staff that the company's corporate headcount will shrink in the coming years as it rolls out more generative AI tools and agents. Jassy urged employees to do more with scrappier teams. Why it matters: After cutting over 27,000 jobs since 2022, Amazon is signaling that AI efficiency will permanently reshape white-collar tech hiring.
Browserbase just launched its AI web agent built on the open-source Stagehand framework. It turns natural language prompts into browser automations with zero coding required, and it is completely free to use. Basically, automating repetitive browser tasks and web scraping just got accessible to anyone without writing a line of code.
MiniMax just launched Hailuo 02, its new video generation model capable of tackling intricate movements like gymnastics. Built on a new NCR architecture, it triples model parameters and quadruples training data while boosting efficiency by 2.5x to offer native 1080p outputs. Why it matters: High-end video generation is getting physically coherent without breaking the bank.
California experts just released a 52-page blueprint to replace the vetoed SB 1047 bill, led by Stanford's Fei-Fei Li and other researchers. The report warns that AI companies can't objectively evaluate their own risks, urging mandatory third-party audits, whistleblower protections, and legal safe harbors for independent safety researchers.
Google just released Gemini 2.5 Pro and Flash as stable models while introducing a preview of 2.5 Flash-Lite. Flash-Lite is pitched as the fastest and most cost-effective 2.5 model yet, packing a 1M-token context window, adjustable thinking budgets, and lower latency for high-volume tasks. Why it matters: Google is aggressively undercutting API costs for high-throughput production workloads.
Baidu just launched an e-commerce livestream featuring top host Luo Yonghao's digital avatar, powered by its ERNIE foundation model. It was the first livestream to feature two digital avatars co-hosting and interacting in real time. The 6+ hour stream pulled in over 13 million views and generated more than 55 million RMB in GMV.
Krea AI just launched the public beta of Krea 1, its first image model trained in collaboration with @bfl_ml. The new model promises superior aesthetic control and higher image quality. Best of all, it's currently free for anyone to use.
Adobe just launched its Firefly generative AI app on iPhone and Android, bringing text-to-image, text-to-video, and Generative Fill to mobile devices. The app even lets users tap into third-party models like Google's Imagen and Veo, with all projects automatically syncing back to Creative Cloud desktop and web workflows.
An advertising watchdog just told Microsoft to drop or modify claims that Microsoft 365 Copilot boosts productivity and ROI. Regulators found surveys claiming up to 75% of users "feel" more productive reflect subjective perception rather than objective proof, while confusing branding around Business Chat misleads customers. Microsoft disagreed with the verdict, but agreed to clarify its marketing.
Meta just revealed the inaugural cohort for its Llama Startup Program, selected from more than 1,000 applications. The early-stage startups span healthcare administrative agents, legal tech, and gaming, and will receive direct backing from Meta's Llama engineering team alongside cloud credit reimbursements to offset inference and training costs.
OpenAI just notified developers that GPT-4.5 Preview will be removed from its API on July 14, 2025. While the phase-out was technically planned back in April alongside GPT-4.1's debut, the reality check has left teams relying on its creative writing tone scrambling to migrate. ChatGPT subscribers can still breathe easy: it remains an option in the web dropdown.
Max just launched dynamic video previews on its homepage for U.S. connected TV devices. In the coming weeks, it's deploying a new in-house AI tool to scan its vast catalog and suggest recognizable "Drop-In Moments" for preview reels, cutting down the tedious manual work of hand-picking clips.
NVIDIA just released early developer previews of Isaac Sim 5.0 and Isaac Lab 2.2 on GitHub. The updates give robotics builders physics-based simulation tools to train and validate robots, adding open-source access, synthetic data generation extensions, and new robot models.
OpenAI executives have reportedly considered publicly accusing Microsoft of anticompetitive behavior and asking federal regulators to review their contract. The tension escalated over OpenAI’s $3 billion acquisition of Windsurf, as OpenAI refuses to let Microsoft tap into the startup’s coding IP for GitHub Copilot. OpenAI wants to loosen Microsoft’s grip on its compute, but it still needs its partner's sign-off to complete a for-profit transition.
Cursor just launched a new $200/month Ultra plan packing 20x more usage than Pro, tailored for heavy coders who want predictable bills instead of pay-as-you-go spikes. The company is also upgrading Pro by shifting from request caps to compute limits, unlocking unlimited tool calls and unlimited access to its Auto model.
Researchers just unveiled AI-CAC, a deep learning tool that digs through routine, nongated chest CT scans to catch unnoticed coronary artery calcium. Trained across 98 VA medical centers, it hits 89.4% accuracy and predicts 10-year mortality risk. It turns millions of existing hospital scans into early cardiovascular warning systems without extra appointments.
Meta just teased a new collaboration with Oakley, launching an Instagram page that points to a June 20 reveal. The upcoming AI smart glasses are reportedly built for athletes, using a lightweight frame based on Oakley's Sphaera style with a single camera to cut weight. Why it matters: Meta is pushing into fitness tracking before its full AR glasses arrive in a few years.
AI pioneer Geoffrey Hinton just warned that mundane intellectual labor is on the verge of being entirely replaced, calling mass job displacement the biggest immediate threat. While one worker with AI could soon replace ten office staff in fields like legal and call centers, physical trades like plumbing remain safe for now.
1X just introduced the 1X World Model (1XWM), an action-controllable system that forecasts future video frames and task success directly from precise robot trajectories. It allows 1X to rigorously evaluate humanoid policies in bits rather than spending lifetimes testing them in physical living rooms, though it still struggles with unfamiliar objects.
ElevenLabs just introduced Audio Tags for its Eleven v3 model, letting creators direct the exact timing, rhythm, and emphasis of generated speech. By slipping bracketed cues like [pause], [rushed], or [stammers] right into the script, users can turn monotone readings into nuanced, human-like performances. However, professional voice clones aren't fully optimized for v3 yet.
TikTok just expanded its Symphony AI suite with text-to-video and image-to-video tools that turn static product shots into five-second branded clips. Instead of keeping the tech locked inside its own app, TikTok is integrating the features directly into Adobe Express and WPP Open, giving agency creatives an automated shortcut for ad production.
Groq just officially became a supported Inference Provider on the Hugging Face Hub. Developers can now run open-source models like Llama 4 directly through Groq's low-latency LPU chips using HF's web UI or Python and JS SDKs. You can bill through Hugging Face without added markups or plug in your own Groq API key.
MiniMax just open-sourced MiniMax-M1, a hybrid-attention reasoning model with a 1 million token context window and 80k-token reasoning output. Thanks to its Lightning Attention mechanism and CISPO RL algorithm, deep reasoning costs roughly 30% of DeepSeek R1's compute, with the entire RL phase trained on 512 H800s for just $535k. It rivals Gemini 2.5 Pro in long context while being completely free on the web.
Adobe just rolled out Adobe LLM Optimizer at Cannes Lions, an enterprise tool designed to help brands rank better in ChatGPT, Claude, and Gemini responses. It tracks how models cite brand content and automatically suggests or applies structural and metadata fixes. Basically, Adobe is making Generation Engine Optimization (GEO) the next SEO.
Reddit just launched Community Intelligence, an AI engine that turns years of human discussions into structured marketing data. The rollout debuts Reddit Insights for trend tracking and Conversation Summary Add-ons, which dynamically attach authentic positive user comments below ads. Early tests showed the social proof boosted clickthrough rates by 19%.
