Perplexity AI just submitted an official bid to ByteDance to merge its AI search engine with TikTok's U.S. operations and new capital partners. The proposed deal would let ByteDance's existing investors keep their stakes while funneling short-form video into Perplexity's engine. Sources value the potential deal well north of $50 billion.
Sam Altman just announced that OpenAI has finalized o3-mini and started its rollout process, targeting a launch in about two weeks. Addressing developer demands, the team plans to release the model on both the API and ChatGPT at the exact same time.
A Canadian maker named Hudzah just built a tabletop nuclear fusor capable of detecting neutrons for around $3,000, guided largely by Anthropic's Claude and OpenAI's o1 pro. Claude Projects digested forum archives to debug the build, while o1 pro handled intricate electrical wiring. Why it matters: Frontier reasoning models are actively lowering the barrier for complex, high-risk physical engineering at home.
Perplexity just acquired Read.cv, a LinkedIn alternative known for clean portfolio designs. The Read.cv team will join Perplexity's design and engineering group, while its existing platform winds down with user data exports open until May 16. It marks Perplexity's third acquisition to date.
OpenAI just updated the custom instructions UI inside ChatGPT, making it much easier to tweak how the bot talks to you. The refreshed interface lets you directly set the personality traits you want it to have, its tone of voice, and any guardrails it should follow.
Runway just launched Frames, its new text-to-image generator engineered for high-fidelity, cinematic visuals. Available to Unlimited and Enterprise subscribers, it packs 19 preset styles and lets users seamlessly turn generated stills into video using tools like Gen-3 Alpha Turbo with a single click. Runway is gunning straight for Midjourney's crown.
Cognition just released Devin 1.2, giving its AI software engineer better in-context reasoning to understand repository structures and reuse existing code. Users can now assign tasks simply by sending voice notes directly in Slack. The platform still starts at $500 a month, now with usage-based overage options.
Apple just temporarily disabled notification summaries for news and entertainment apps in the latest iOS 18.3 developer beta. The Apple Intelligence feature sparked backlash after hallucinating crucial facts, including wrongly claiming a shooting suspect shot himself in a BBC alert. Apple plans to bring it back later with italic styling, individual app toggles, and clear warnings that AI summaries "may contain errors."
Krea AI just rolled out a tool that converts flat 2D images into 3D objects within its real-time generator. Users can rotate, reposition, and scale the converted models while the background updates live. Why it matters: it makes complex scene composition and character posing much easier without manual redrawing.
Black Forest Labs just launched the FLUX Pro Finetuning API, allowing creators to customize flagship FLUX Pro models with as few as 1 to 5 reference images. The custom finetunes reliably preserve unique characters or visual styles at up to 4 megapixels and work seamlessly across the full FLUX tool suite.
Microsoft Research just published and open-sourced MatterGen, a generative diffusion model that designs novel physical materials directly from property prompts instead of screening databases. Working with SIAT researchers, the team successfully synthesized a completely new material (TaCr2O6) in the lab with under 20% error against specifications. Why it matters: It shifts material discovery from brute-force search to generative design for batteries and solar cells.
Mistral just partnered with global news agency AFP to plug up-to-date journalism directly into its Le Chat assistant. The multi-year deal gives the bot access to AFP's archives dating back to 1983 alongside 2,300 daily stories across six languages, excluding photos and videos. It marks Mistral's first media pact as it pushes beyond raw models to take on ChatGPT.
Researchers just introduced a new framework that scales diffusion models at inference time beyond traditional denoising steps. By pairing automated verifiers with search algorithms to hunt for better noise candidates, dumping more compute into generation now reliably translates into higher image quality.
Mark Zuckerberg used YouTube's fight against piracy to justify Meta's use of copyrighted datasets during a deposition in an ongoing AI copyright lawsuit. Newly unsealed court filings allege Meta trained its Llama models on pirated e-book repositories like LibGen and Z-Library despite internal legal warnings.
Luma AI just released Ray2, its next-gen video model trained with 10x the compute power of Ray1. It brings coherent physics, cinematic camera controls, 1080p output with 4K upscaling, and seamless video extensions up to 30 seconds. It is live now for paid Dream Machine subscribers on web and iOS.
Keras creator François Chollet just teamed up with Zapier co-founder Mike Knoop to launch Ndea, a new research lab betting on program synthesis over raw model scaling. The team aims to combine deep learning with automated code generation so AI can adapt and reason like humans. Why it matters: Instead of throwing more compute at the same LLM architectures, Chollet is building an alternative path toward true intelligence.
Google just struck a deal with The Associated Press to integrate a real-time information feed directly into Gemini. The move aims to boost up-to-date answers for news queries, though Google hasn't revealed when the feature will launch or which regions get it first.
David Baker’s lab just used deep learning to design brand-new proteins that neutralize lethal snake venom. In mouse tests published in Nature, the computationally generated molecules yielded up to a 100% survival rate against deadly cobra neurotoxins. Traditional antivenoms rely on costly animal plasma, but these compact AI proteins can be brewed cheaply in microbes. Why it matters: While it can't neutralize full venom cocktails yet, it paves the way for accessible, life-saving antidotes in developing nations.
Scammers just used AI-generated photos and deepfake news broadcasts of actor Brad Pitt to con a woman in France out of €830,000. The scammers sent AI hospital selfies and generated news anchors to keep the romance scheme going for over a year. Why it matters: The case shows how accessible generative AI tools are supercharging romance scams, making it harder for ordinary people to spot synthetic deception.
DeepSeek just launched its official mobile app on the iOS App Store and major Android marketplaces, giving users free access to its 671-billion-parameter DeepSeek-V3 model. The app supports web search, deep thinking mode, and file and image scanning. It basically turns one of open source's hottest models into a free, pocket-sized ChatGPT rival.
LinkedIn just rolled out two free AI tools for job hunters and small businesses, including a Jobs Match feature to evaluate if postings are worth applying to and a synthetic hiring agent to triage candidates. Notably, both rely on LinkedIn's in-house AI rather than OpenAI models.
Microsoft just launched Microsoft 365 Copilot Chat, bringing metered AI agents into its free GPT-4o chat experience for commercial users. Employees can now run web searches, analyze files, and trigger custom agents without a full subscription. Why it matters: It gives companies an easy, low-friction on-ramp to deploy workplace agents at scale.
Cisco just launched Cisco AI Defense, an end-to-end security suite designed to protect enterprises building and using AI. Built directly into the network fabric, it automates model red-teaming, detects shadow AI tools, and guards against prompt injections in real time. It rolls out this March.
Hinge just rolled out an AI-powered feature called Prompt Feedback to save singles from cliché profile answers. Powered by OpenAI’s GPT-4o mini, it evaluates your prompt responses across three levels—from "Go a Little Deeper" to "Great Answer"—and suggests specific details to add instead of ghostwriting for you. Why it matters: Daters are fatigued by shallow swipe culture, and gentle AI nudges might bring back authentic conversations.
Freepik just launched an AI-powered tool to remove and replace image backgrounds. Beyond simple solid color swaps, you can generate entirely new AI scenes using Relight technology to match the lighting seamlessly. It is live now on Freepik.
MiniMax just open-sourced its new MiniMax-01 series, featuring a 456B-parameter language model and a visual multimodal model. Built on a novel Lightning Attention architecture with linear complexity, the models handle up to 4 million tokens of context with virtually zero loss. Weights are available on GitHub alongside access via MiniMax Chat.
De-extinction startup Colossal BioSciences just secured $200M in fresh funding, lifting its valuation to a massive $10.2 billion. The company will use the capital to push forward genetic engineering for the woolly mammoth, thylacine, and dodo, with tech aimed at species preservation and human healthcare.
OpenAI just announced a new content partnership with Axios, funding four new tech-enabled local newsrooms in cities like Pittsburgh and Kansas City. The deal brings Axios reporting directly into ChatGPT search with explicit citations and backlinks, as OpenAI deepens its push to turn media publishers into allies rather than copyright adversaries.
OpenAI just rolled out a beta tasks feature to ChatGPT Plus, Team, and Pro subscribers. Users can now ask ChatGPT to fire off push reminders or handle recurring requests like daily news roundups. While it can't run background searches or buy tickets yet, it's OpenAI's first real step toward independent AI agents.
Apple just joined the board of the Ultra Accelerator Link (UALink) Consortium alongside Alibaba and Synopsys, throwing its weight behind an open interconnect standard for AI data center chips. UALink is designing an open alternative to Nvidia's dominant NVLink. Why it matters: As Apple scales Apple Intelligence and reportedly develops custom server silicon, backing an open interconnect gives it more control over its future infrastructure.
Wyze just launched Cam Unlimited Pro, a new subscription tier that turns vague security notifications into detailed AI summaries. Instead of generic alerts like "pet detected," users now get specific play-by-plays like a small dog playing with a tennis ball, alongside natural-language video search.
Sony just revealed new patents exploring how AI can predict player movements to reduce online gaming lag. Using a camera to track player behavior and anticipate inputs, the company is also looking into AI to catch cheaters and detect player frustration. It is still on paper, but it hints at how PlayStation might tackle latency.
President Biden just signed an executive order to fast-track US AI data center construction by leasing federal lands. The Defense and Energy departments must identify at least three sites each by February 28th to host gigawatt-scale facilities. While developers must fund clean power and source US-made chips, the move sparks heavy backlash over grid strain and climate goals.
Microsoft just launched its Quantum Ready program, offering workshops, skilling resources, and security roadmaps to help organizations prepare for practical quantum computing. The initiative focuses on building hybrid applications that blend quantum hardware with classical supercomputers and AI, addressing a skills gap where only 12% of business leaders feel ready for quantum tech.
Samsung just dropped a teaser video hinting at major AI upgrades for the upcoming Galaxy S25. In the demo, the AI handles a multi-step spoken request—finding a specific restaurant, messaging a contact, and updating the calendar all at once. The "true AI companion" will officially debut at Galaxy Unpacked on January 22.
Amazon's major revamp of Alexa is running into serious technical roadblocks, according to the Financial Times. Rohit Prasad, head of Amazon's AGI team, admitted that hallucinations must drop "close to zero" alongside cutting response lag and integrating hundreds of third-party apps. Insiders warn a launch won't happen anytime soon.
Sakana AI just unveiled Transformer², an adaptive system that adjusts LLM weights at test time to tackle different tasks like a living brain. Using Singular Value Finetuning and RL, it beats LoRA across math and coding benchmarks while requiring far fewer parameters.
Freepik just launched AI video generation with sound directly inside its AI Suite, supporting multiple modes and formats. Instead of generating silent clips and hunting for soundtracks in separate apps, creators can now get complete audiovisual clips in a single workflow.
OpenAI just appointed Adebayo Ogunlesi, founding partner of Global Infrastructure Partners and a senior managing director at BlackRock, to its board of directors. Ogunlesi brings decades of elite experience in transnational finance and mega-scale infrastructure. As AI labs race to secure data centers and power, his arrival signals OpenAI's aggressive push into physical infrastructure.
Netflix and Eyeline researchers just introduced "Go with the Flow," a new method that steers motion in video diffusion models without changing model architectures. By warping latent noise using optical flow fields in real time, it enables precise camera moves and object control while keeping frame quality crisp.
Mayo Clinic just teamed up with Microsoft Research and Cerebras Systems to build specialized foundation models for healthcare. The projects use chest X-rays to automate radiology reporting and patient exome data to tailor treatments for diseases like rheumatoid arthritis. Why it matters: bringing multimodal AI directly to clinical scans and genomics could drastically cut diagnosis times.
Microsoft and Pearson just signed a multiyear deal to develop new AI learning tools and certifications. Pearson will rely on Azure and roll out Microsoft 365 Copilot across its entire workforce, while Microsoft extends its Pearson VUE certification contract through 2029. Why it matters: Companies want AI, but finding workers trained to use it remains the biggest bottleneck.
OpenAI just published an "economic blueprint" outlining how it wants the U.S. government to support and regulate AI. The document urges Washington to slash red tape for data centers and nuclear power, adopt voluntary federal safety standards over fragmented state bills, and protect developers' ability to train on publicly available copyrighted web data.
OpenAI and Google are now buying YouTubers' unpublished, raw video footage for $1 to $4 per minute to train video generators like Sora and Veo. Licensing middleman Troveo AI has already paid creators over $5 million for unused 4K and drone clips. Why it matters: The race for video training data is turning cutting-room-floor scraps into an unexpected cash windfall.
Israeli startup FireDome has developed an off-grid wildfire defense system inspired by Israel's Iron Dome. The setup combines AI cameras with a mechanical launcher that fires eco-friendly retardant capsules up to a quarter-mile away when flames are detected. Field tests are scheduled to begin in May 2025.
Adobe just rolled out Firefly Bulk Create in beta, letting users resize or swap backgrounds across up to 10,000 images at once. The web tool packages Firefly developer APIs into simple "Remove Background" and "Resize" features aimed at marketers. It relies on consumption-based credits, with PSD export coming later.
Nvidia just publicly pushed back against the Biden administration's new 200-page "AI Diffusion" rule, calling it misguided overreach. The chip giant argues the sweeping restrictions micromanage how semiconductors and even consumer gaming PCs are sold globally, hurting American competitiveness rather than protecting it.
Mercedes-Benz just partnered with Google Cloud to power its next-gen MBUX Virtual Assistant using Gemini. The smart navigation agent can handle complex multi-turn chats and remember conversation context throughout a trip, debut set for the new CLA running MB.OS later this year.
The Biden-Harris Administration just released an Interim Final Rule on AI diffusion, reshaping how computing power moves across the globe. It lifts export restrictions for 18 core allies and creates verified end-user statuses, while strictly clamping down on closed-weight model transfers to adversaries. Basically, Washington wants the world's AI running exclusively on American rails.
OpenAI is officially getting into physical hardware. Robotics hardware lead Caitlin Kalinowski just posted openings for the company’s first hardware-focused roles, recruiting senior sensing engineers, mechanical designers, and a TPM manager to set up a new training lab. OpenAI isn't content with just supplying the AI brains; it wants to build the bodies too.
Vectara's benchmark shows Zhipu AI's GLM-4-9B-Chat and Google's Gemini-2.0-Flash-Exp tied for the lowest hallucination rate at just 1.3%, followed by OpenAI's o1-mini at 1.4%. The numbers prove that smaller, specialized models can easily match or beat giant frontier LLMs when it comes to factual accuracy.
NVIDIA just released its latest retail industry report, revealing that nearly 90% of retailers and CPG brands are now using or testing AI. Over 94% reported cuts in operating costs, with generative AI largely deployed for marketing content and predictive analytics.
UC Berkeley’s NovaSky team just open-sourced Sky-T1-32B-Preview, a reasoning model that rivals OpenAI's o1-preview while costing under $450 to train. Fine-tuning Qwen-2.5-32B on 17K curated reasoning samples took 19 hours on 8 H100 GPUs. Best of all, the entire pipeline, data, and weights are publicly available.
Adobe Research and HKUST just open-sourced TransPixar, a new model that solves a major VFX headache: generating videos with built-in alpha channels. Instead of solid backgrounds, it generates transparent RGBA elements like smoke and portals ready for compositing. It is an early-stage research project, but the code and interactive demo are already live.
xAI just released a standalone iOS beta app for Grok in the US. The app delivers the same real-time info, answers, and image generation without needing to open X. After going free last month, Elon Musk's chatbot is ready to take on ChatGPT and Claude directly.
VideoLAN just showed off built-in, real-time AI subtitling and translation for VLC at CES 2025. The feature runs entirely locally and offline without cloud services, with plans to support more than 100 languages. There is no official release date yet, but it marks a massive privacy-friendly upgrade for local media playback.
Google just rolled out Daily Listen in Search Labs, an experimental feature that turns your search habits and Discover feed into a customized five-minute AI podcast. It uses the same Audio Overview magic found in NotebookLM, complete with scrubbers and full transcripts. For now, it's strictly an opt-in test for iOS and Android users in the US.
Edelman just released its curated AI Top Voices list to help brands communicate AI initiatives through trusted figures. With 74% of people trusting individuals over institutions, the agency vetted B2B creators across LinkedIn and YouTube based on credibility, engagement, and ethical alignment.
Garmin just won a CES Innovation Award for its next-gen Unified Cabin solution, built alongside Qualcomm's Snapdragon Cockpit Elite platform. The setup uses generative AI to generate personalized cabin themes, brings dedicated voice assistants and audio to every passenger seat, and relies on UWB sensors to detect seat occupancy.
Startup Satellai just unveiled the Satellai Dog Tracker, combining AI and satellite connectivity to track pets in remote areas without cellular coverage. Powered by Qualcomm's 9205S modem and Skylo's NTN network, it lets owners map virtual boundaries even deep off-grid.
Stability AI and NVIDIA just rolled out SPAR3D at CES, turning single 2D images into detailed 3D objects in just 0.7 seconds. The architecture pairs a point diffusion model with a triplane transformer, letting creators edit point clouds and rebuild meshes in 0.3 seconds. Weights are open on Hugging Face for RTX AI PCs.
Microsoft just released the official weights for its 14B-parameter Phi-4 model on Hugging Face under a permissive MIT license. Previously restricted to the Azure AI Foundry platform, developers can now freely download, fine-tune, and deploy the high-reasoning model locally for commercial applications.
NVIDIA just expanded its NIM microservices and AI Blueprints to RTX AI PCs, pairing prepackaged model APIs with RTX 50 Series FP4 compute. By cutting VRAM requirements by more than half, developers can run complex agentic workflows locally without painful optimization. Hardware support arrives next month.
Polish director Patryk Vega just set his $15M biopic Putin for a Jan 10 premiere across 64 countries. The production used AI visual effects to superimpose the Russian president's real face onto an actor over his body language. Why it matters: It brings deepfake-level AI straight into feature-length cinema, sparking fierce ethical debate along the way.
McAfee just unveiled McAfee Scam Detector at CES 2025, an AI-powered tool built to identify scams across text messages, emails, and online video streams in real time. The proactive system flags deceptive and deepfake threats before victims can be duped. It will roll out at no extra cost to McAfee customers starting this spring.
Delta CEO Ed Bastian just unveiled Delta Concierge, an AI assistant launching this year to rebook delayed flights and guide passengers to their gates. Instead of generic chatbots, the airline is targeting real operational pain points, using AI behind the scenes to optimize gate assignments and flight connections.
Panasonic just unveiled Umi at CES 2025, a family-focused wellness digital assistant powered by Anthropic’s Claude. The mobile app lets households set health goals, manage routines, and monitor aging parents via voice and group chat. It is scheduled to launch in the U.S. later this year.
Philips Hue just announced an upcoming generative AI lighting assistant arriving in the first quarter of 2025. Users can simply type or speak prompts like "Give me a scene for a garden party" to generate custom color palettes or adjust existing fixtures. The feature will be compatible with all Philips Hue smart lights.
Samsung just teamed up with Instacart to turn its 2025 Bespoke smart fridges into automated grocery restock hubs. Powered by AI Vision Inside food recognition, users can order missing groceries right from the fridge screen for same-day delivery. Existing AI Family Hub+ units will also receive the feature via network firmware updates in 2025.
Nvidia just unveiled Mega at CES, an Omniverse blueprint designed to test and optimize physical AI and industrial robot fleets in digital twins before real-world deployment. It combines Isaac robotics and Omniverse sensor simulation, allowing operators like KION Group to simulate infinite warehouse scenarios. Why it matters: Bringing software-defined simulation to physical factories finally lets logistics scale without trial-and-error downtime.
Nvidia just unveiled Project Digits at CES 2025, a $3,000 personal AI supercomputer powered by its new GB10 Grace Blackwell Superchip. The compact workstation packs 128GB of memory, letting researchers and developers run models with up to 200 billion parameters right on their desks—or link two together for 405-billion-parameter beasts. It begins shipping in May.
NVIDIA just opened early access for its Omniverse Sensor RTX APIs, targeting AV and robotics developers. The tool simulates physically accurate data for cameras, radar, and lidar, letting teams safely generate rare edge cases without risking real-world hardware. Partners like Accenture, Foretellix, and Nuro are already plugging it into their testing pipelines.
Nvidia just released a new AI Blueprint for video search and summarization, designed to let visual agents ingest surveillance and industrial footage at 30x real-time speed. Powered by Cosmos Nemotron models, the agents pack reasoning and tool-calling skills to flag factory defects or analyze athletic form. Why it matters: it unlocks actionable data from billions of unwatched cameras.
NVIDIA just unveiled its GeForce RTX 50 series GPUs at CES, doubling generative AI performance with FP4 support. The new quantization cuts model memory needs in half, allowing models like FLUX.1 to run locally on under 10GB of VRAM in just five seconds. The flagship RTX 5090 launches January 30.
NVIDIA just introduced the Llama Nemotron and Cosmos Nemotron model families, custom-built to power autonomous AI agents across enterprises. Available in Nano, Super, and Ultra sizes, the open models use distillation and pruning to run everywhere from edge PCs to multi-GPU clusters. Why it matters: It gives developers purpose-built open building blocks to deploy multi-agent workflows without bleeding compute budgets.
NVIDIA just unveiled Cosmos at CES, a platform of open generative world foundation models designed for physical AI like autonomous vehicles and robots. Built into its three-computer AV pipeline, Cosmos turns thousands of human-driven miles into billions of synthetic driving miles to stress-test edge cases. The models are already available under an open license.
Nvidia just teamed up with orchestration leaders like LangChain and CrewAI to release new enterprise-ready Agentic AI Blueprints. The catalog includes pre-built templates for turning PDFs into conversational podcasts, automating code docs, and orchestrating multi-agent systems, letting developers spin up reasoning workflows in a single click via NVIDIA Launchables.
NVIDIA just unveiled its Isaac GR00T Blueprint for synthetic motion generation alongside the new Cosmos world foundation models at CES. Developers can capture human demos via Apple Vision Pro and multiply them into massive synthetic datasets in Omniverse. Robotics giants Boston Dynamics and Figure are already onboard.
NVIDIA just launched the DRIVE AI Systems Inspection Lab to help automakers and suppliers verify self-driving tech against strict international safety standards. Accredited by ANAB, the lab evaluates partners' software on the DRIVE AGX platform across functional safety, cybersecurity, and AI compliance, with Continental and Sony among the first onboard.
Apple just announced it will update iOS to clearly label Apple Intelligence notification summaries after repeated complaints from the BBC. The beta feature had been badly butchering news alerts, including falsely claiming a shooting suspect shot himself. You can already turn off notification summaries under Settings > Notifications if you do not trust them.
Samsung just confirmed that its rolling AI companion robot Ballie will actually ship to consumers in 2025. Equipped with a built-in projector, onboard speakers, and visual AI, Ballie can cast screens onto walls or floors and manage smart home devices. Why it matters: After five years of CES demos, Samsung is finally putting its adorable concept bot to the real-world test.
AMD just previewed its next-gen Radeon RX 9070 and 9070 XT GPUs at CES, built on a new 4nm RDNA 4 architecture. The biggest shift is FSR 4, which now relies on machine learning upscaling powered by dedicated on-die AI accelerators. Both cards drop in Q1, though specs and prices remain under wraps.
Google DeepMind is building a new team dedicated to AI that simulates the physical world, led by former OpenAI Sora co-lead Tim Brooks. The group plans to build on Gemini, Veo, and Genie to advance visual reasoning, robotics planning, and real-time generation, viewing world models as essential on the path to AGI.
Google just previewed Gemini AI features for Google TV at CES 2025, turning the living room screen into a conversational hub. The update lets you search via natural conversation, control smart home devices, and generate a daily "News Brief" sourced from the web and YouTube. It is slated to roll out to select devices later in 2025.
Qualcomm just expanded its Snapdragon X lineup with a cheaper Arm laptop chip, bringing Copilot Plus PCs down to around $600. It features an eight-core Oryon CPU and a 45 TOPS NPU to keep all of Microsoft's on-device AI features intact. Why it matters: Qualcomm is aggressively taking Windows on Arm mainstream to squeeze Intel where it hurts.
Brisk It just unveiled the Zelos 450, a compact $399 smart pellet grill powered by its upgraded Vera 2.0 generative AI platform. Users can snap photos of raw ingredients to generate custom recipes that the grill executes automatically, with the AI even adjusting temperatures to salvage common grilling blunders. It is set to hit shelves in Q1 2025.
Intel just unveiled a massive PC chip refresh at CES 2025, led by its Core Ultra 200 series across desktop and mobile. The new silicon features dedicated NPUs, AI-based power management, and up to 3.3x faster local Llama 3 8B performance on the 285H. Systems begin shipping later this month.
Halliday just debuted a pair of retro-style smart glasses at CES 2025 that skip traditional waveguide lenses entirely. Using a tiny optical module called DigiWindow, the device projects a discreet 3.5-inch virtual screen directly into your field of vision without a camera, paired with a proactive AI agent that assists during conversations.
Timekettle just rolled out Babel OS, its first operating system built specifically for AI-driven simultaneous interpretation in earbuds. It uses semantic segmentation to predict speech completions with near-zero latency, adds voice cloning to preserve tone, and supports custom lexicons across 40+ languages. The software is rolling out via free OTA updates to products like the W4 Pro and X1 Hub, with offline edge AI on the way.
OpenAI CEO Sam Altman just admitted that the company is currently losing money on its $200-per-month ChatGPT Pro plan because heavy users are running it far more than anticipated. Altman confessed he personally chose the price thinking it would turn a profit, and OpenAI is now reportedly weighing subscription price hikes or usage-based billing.
Aiper just unveiled the IrriSense Smart Irrigation Sprinkler at CES 2025, a $399 plug-and-play yard sprinkler that doesn't require buried pipes. Users can map their lawn in an app to avoid spraying sidewalks, while weather tracking and soil sensors skip routines if it just rained. It basically turns lawn care into a precision task.
Current Backyard just unveiled the Model P, a $599 indoor-outdoor electric pizza oven that reaches 850°F to cook a pie in two minutes flat. It uses proprietary algorithms to distribute heat evenly without requiring manual rotation, and its app calculates optimal cooking times based on your dough and toppings.
Withings just revealed Omnia, a futuristic smart mirror concept that packs a smart scale base and an embedded AI voice assistant. The device measures weight, heart, and metabolic metrics, pulls data from other wearables, and talks you through daily health feedback with optional telehealth tie-ins. It is currently in development with no set release date.
VanceAI just launched Video AI, an all-in-one video enhancement tool capable of upscaling footage to 8K. It also packs frame interpolation up to 120 FPS, SDR-to-HDR conversion, denoising, deinterlacing, and stabilization into a single suite for restoring and polishing clips.
Elon Musk just revealed that xAI's upcoming Grok 3 has finished pretraining, using 10 times more compute than Grok 2. He added that the new flagship model is coming soon. Why it matters: xAI is rapidly scaling its infrastructure to challenge OpenAI and Google at the very top of frontier AI.
Meta researchers just detailed how generative retrieval can predict user intent instead of relying on traditional database searches. By introducing frameworks called LIGER and Mender, the system treats recommendations as sequence predictions, slashing storage costs while using LLMs to capture subtle user preferences. Why it matters: It shows generative models could soon replace the classic embedding systems powering the internet's biggest feeds.
OpenAI just shared personal reflections on the two years since launching ChatGPT, detailing the company's turbulent scaling and internal drama. The leadership noted they now know how to build AGI, predicting the first AI agents will "join the workforce" in 2025 before setting sights on superintelligence.
Sonus AI just launched its Sonus-1 family of language models, offering Mini, Air, and reasoning-boosted Pro tiers. The flagship Sonus-1 Pro with reasoning claims a 91.8% score on MATH-500 and 90.0% on HumanEval, taking aim at frontier proprietary models across math and code.
