Nvidia CEO Jensen Huang just claimed on Lex Fridman's podcast that humanity has already achieved AGI. When pressed on whether AI could start and run a billion-dollar company, he quickly walked back the claim, noting that users usually lose interest in agents after a few months and the odds of 100,000 agents building another Nvidia are literally zero.
Cisco just released its free LLM Security Leaderboard, testing how top AI models hold up against adversarial attacks. Anthropic completely swept the top ranks by taking 8 of the top 10 spots, led by Claude Opus 4.5 at No. 1, while OpenAI’s GPT-5.2 finished 7th. Cisco explicitly targeted multi-turn attacks to mirror realistic jailbreak attempts over extended conversations.
Mistral AI just rolled out Voxtral TTS, its first text-to-speech model built for real-time voice agents. The 4B-parameter model runs at a 70ms latency, supports expressive speech across nine languages, and can clone a voice from a 3-second sample. It is available via API, in Mistral Studio, and as open weights on Hugging Face.
Luma Labs just introduced Uni-1, an AI model that reasons and generates pixels simultaneously. Instead of treating reasoning and visual generation as separate steps, it combines both into a single unified process that Luma calls less artificial and more intelligent.
Sakana AI just launched Sakana Chat, powered by its new Namazu alpha series that adapts open-weight frontier LLMs for Japan through post-training. The models slash DeepSeek's 72% question refusal rate on sensitive topics to nearly zero while keeping benchmark parity in reasoning and coding. The web chat also packs real-time search capabilities.
Anthropic just rolled out computer use in research preview for Claude Pro and Max subscribers inside Claude Cowork and Claude Code. Claude can now click, scroll, navigate files, and run browser tools on macOS and Windows, falling back to direct screen control when direct connectors aren't available. Paired with last week's Dispatch release, you can assign tasks on your phone and let Claude finish them at your desk.
OpenAI is dangling preferred shares with a guaranteed 17.5% minimum return to lure private equity giants like Bain Capital and Brookfield into enterprise AI joint ventures. The company wants to raise $4 billion at a $10 billion pre-money valuation to deploy AI tools across PE-owned portfolios, aiming to outflank rival Anthropic on enterprise turf.
OpenAI is in talks to purchase electricity from fusion startup Helion Energy, aiming for 5 gigawatts by 2030 and scaling up to 50 GW by 2035. OpenAI CEO Sam Altman has stepped down as Helion's board chair to avoid conflicts of interest. The deal would secure critical clean energy for AI, though commercial fusion is still unproven.
Instagram is rolling out a new "AI Transition" tool for Stories that uses AI morphing to stitch multiple still photos into a seamless video clip. Users just pick two or more images, let Meta AI generate the sequence, and can even save it to repost on Reels. It requires accepting Meta AI terms and is currently rolling out to select accounts.
Pascal Editor just launched as a 100% free and open-source 3D architecture editor running entirely in the browser. Powered by WebGPU and React Three Fiber, it supports real-time editing of walls, slabs, and zones with GPU-accelerated rendering. It brings desktop-grade architectural modeling to the web without the typical enterprise software price tag.
Silicon Valley startups and tech giants are quietly adding six-figure AI token budgets alongside traditional base salary and equity. Nvidia CEO Jensen Huang recently suggested engineers should receive roughly half their base salary again in compute tokens. Why it matters: While a massive compute allowance sounds generous, it doesn't vest or appreciate—and finance teams may expect you to double your output in return.
Elon Musk just unveiled Terafab, a new Texas chip manufacturing plant operated by Tesla, SpaceX, and xAI that targets 50x current global compute output. The vertical facility plans to fabricate chips for cars, Optimus robots, and solar-powered orbital satellites. Why it matters: Building a cutting-edge fab from scratch is an enormous gamble, but orbital AI compute might be Musk's boldest bet yet.
Elon Musk just confirmed that xAI is preparing to roll out "Grok Computer," an AI agent built to take over screen, mouse, and keyboard inputs. The leak surfaced after users spotted a hidden toggle in Grok's web code, hinting at a trial run for the rumored Macrohard project. It is currently undergoing testing.
Anthropic just launched Channels for Claude Code, allowing developers to manage coding sessions directly through Telegram and Discord via MCP. You can now triage CI bugs and trigger fixes straight from your phone instead of being tethered to a terminal. It is currently in preview and requires a persistent local process.
OpenAI just launched a perk giving verified university students in the US and Canada $100 in free Codex credits, equivalent to 2,500 ChatGPT credits. Eligible students can claim the balance via SheerID using a school email, and the credits apply across all ChatGPT plans with a 12-month expiration. It’s an aggressive bid to hook the next generation of builders early.
Claude Cowork just rolled out Projects to keep tasks and context in one focused workspace. All files and instructions stay locally on your computer instead of the cloud, and you can import existing projects in a single click or start fresh.
Amazon is reportedly working on an AI-centric smartphone code-named "Transformer," over a decade after shutting down the Fire Phone. Developed under former Xbox veteran J Allard, the device takes cues from the minimalist Light Phone and may ditch traditional app stores for ChatGPT-like mini apps. No release window or pricing has been set.
GreenSock just rolled out gsap-skills, an official skill pack that teaches 40+ AI agents—including Cursor, Claude Code, and Copilot—how to write clean GSAP code. It covers timelines, ScrollTrigger, and React patterns, basically stopping your coding assistant from hallucinating outdated animation syntax.
OpenAI is shifting its core focus toward building an autonomous "AI researcher" to tackle complex science and math challenges. Chief scientist Jakub Pachocki revealed plans for an AI research intern by September that handles multi-day tasks, followed by a multi-agent system in 2028. It could reshape labs, though chaining complex reasoning without errors remains tricky.
The White House just released a light-touch AI policy blueprint, urging Congress to establish federal rules that override state-level AI regulations. The proposal targets political bias in models, calls for age-gating to protect children, and advises lawmakers against creating new AI agencies while codifying ratepayer pledges signed by Amazon, Google, and OpenAI.
OpenAI is reportedly working on a desktop “superapp” that unites ChatGPT, its Codex coding tool, and the Atlas browser into one client. An internal memo cited by The Wall Street Journal reveals that applications CEO Fidji Simo wants to curb fragmentation and drop “side quests” as competition heats up with rivals like Anthropic. The mobile app will remain separate.
Blue Origin just filed FCC plans for Project Sunrise, aiming to launch up to 51,600 sun-synchronous satellites to host orbital data centers powered by non-stop solar energy. It mirrors SpaceX's massive satellite push to bypass terrestrial grid limits, though critics like Sam Altman have already labeled space servers as pure fantasy.
Cloudflare CEO Matthew Prince just warned that AI bots will generate more online traffic than humans by 2027. While a person might browse five sites to research a camera, an AI agent often hits 5,000 for a single task. Why it matters: The web is bracing for a relentless, permanent surge in traffic that physical servers must somehow handle.
Microsoft just launched MAI-Image-2, pushing its in-house model to #3 on the Arena.ai text-to-image leaderboard. The update focuses on lifelike photorealism, cleaner in-image typography for slides and posters, and richer complex scenes. It is already rolling out across Copilot and Bing Image Creator, giving designers a direct alternative to existing image generators.
Adobe just launched Firefly Custom Models in public beta, allowing creators and brands to train custom AI image generators on their own artwork to keep visual styles and character designs consistent across projects. Models are private by default, and Firefly automatically blocks uploads marked as opted out of AI training via Content Authenticity Initiative credentials.
Uber just announced a deal to invest up to $1.25 billion in Rivian, aiming to deploy 10,000 autonomous R2 vehicles by 2031. Starting in San Francisco and Miami in 2028, the robotaxi fleet will run exclusively on Uber's app. Why it matters: Uber wants to lock in hardware partners and cement itself as the platform for autonomous fleets before self-driving rivals push it aside.
DoorDash just launched a new Tasks program and is piloting a standalone app that pays delivery couriers to collect real-world data for AI and robotics. Dashers can earn extra cash between deliveries by filming everyday tasks like washing dishes, photographing hotel entrances, or recording speech. It essentially turns an 8-million-strong delivery fleet into an on-demand data workforce to help machines understand the physical world.
Lovable just expanded beyond building full-stack apps by letting its AI agent run Python scripts in a secure environment. Users can now drop in spreadsheets, PDFs, or raw data to run deep analysis and generate branded pitch decks, reports, invoices, or launch videos directly in chat. Why it matters: It bridges the gap between writing code and running the actual business in a single thread.
Perplexity just launched Perplexity Health, a new feature linking electronic medical records from 1.7 million providers with wearables like Apple Health and Fitbit. Instead of generic web results, answers combine your clinical history and lab work with peer-reviewed medical citations. Health data is encrypted and won't train models, rolling out first to U.S. Pro and Max users.
OpenAI just detailed how it deploys GPT-5.4 Thinking at max reasoning effort to monitor its internal coding agents. Over five months and tens of millions of sessions, the watchdog model caught agents overly eager to bypass rules to finish tasks, though zero rogue scheming was found. It’s a rare peek at how frontier labs supervise high-autonomy tools.
OpenAI just announced it will acquire Astral, the team behind widely popular Python tools uv, Ruff, and ty. After closing, Astral's engineers will join OpenAI to embed their tooling directly into Codex, which now sees over 2 million weekly active users. The goal is to evolve Codex from a code-generating tool into an agent that handles the entire development lifecycle.
Cursor just launched Composer 2, its new frontier-level coding model now available directly in the editor. Trained via continued pretraining and reinforcement learning on long-horizon tasks, it hits 61.3 on CursorBench while cutting pricing to $0.50/M input and $2.50/M output tokens. A faster variant is also available and set as the default option.
Researchers just introduced 3DreamBooth, a framework that brings true 3D spatial awareness to subject-driven video generation. Instead of treating subjects like flat 2D images, it uses a 1-frame optimization trick to decouple geometry from motion and pairs with 3Dapter to route multi-view cues. This basically prevents models from hallucinating mismatched angles in novel views.
MetaClaw just dropped an open-source continual meta-learning framework that turns regular conversations into training signals for LLM agents. It synthesizes reusable skills immediately after failures, then triggers cloud LoRA updates while you sleep or take meetings. That gives personal agents a way to self-improve over time with zero downtime and no local GPU required.
Character.ai just launched Imagine Gallery on its mobile app, bringing every AI visual generated across your chats into a single organized grid. You can now filter creations by Persona, save favorites, and—with the new Imagine Message feature—tap any character prompt to spin up art on the spot. Plus, c.ai+ subscribers can turn images directly into custom chat backgrounds.
Google Labs just transformed its Stitch tool into an AI-native design canvas designed for "vibe design." Users can now talk directly to the canvas to critique layouts, generate high-fidelity UIs from business goals, and instantly turn screens into interactive flows. It bridges the gap from raw sketches to production-ready code via an MCP server integration.
YouTube is now surveying viewers directly, popping up polls that ask if videos they watch "feel like AI slop." Users can rate titles and thumbnails on a five-point scale from "not at all" to "extremely." While YouTube hasn't detailed how flags will alter algorithmic reach, users suspect the crowdsourced labels could also be used to fine-tune Google's own video models.
Google just completely overhauled the vibe coding experience inside Google AI Studio, supercharging it with the new Antigravity coding agent. It now natively integrates Firebase for instant database and authentication provisioning, supports Next.js, and builds real-time multiplayer apps straight from prompts. Basically, prototyping has officially turned into production-ready software development.
Apple is pushing back on vibe coding apps, quietly halting App Store updates for Replit and Vibecode. The company cites longstanding rule 2.5.2, which prohibits apps from downloading and executing code that alters their core functionality—putting app-generated live previews in the crosshairs. It marks the first major collision between App Store gatekeeping and AI-native software creation.
Runway and NVIDIA just previewed a real-time HD video model running on Vera Rubin at GTC, pushing time-to-first-frame latency under 100 milliseconds. The speed is aimed at powering Runway's General World Model, GWM-1, pointing toward truly instant and interactive AI simulation, though it is currently strictly a research preview.
Microsoft is reportedly considering suing Amazon and OpenAI over their massive cloud deal centered on enterprise agent platform Frontier. While Microsoft eased exclusive hosting locks in October, it kept a clause requiring developer access to run through Azure, warning it will head to court if breached.
Xiaomi just officially launched MiMo-V2-Pro, its flagship agent model that previously dominated OpenRouter under the stealth codename "Hunter Alpha." It packs over 1T total parameters (42B active) and a 1M context window, claiming to outcode Claude 4.6 Sonnet on serious software tasks. Why it matters: Hardware giants are racing to build the central brains for autonomous AI agents.
Late actor Val Kilmer is returning to the silver screen, with generative AI recreating his likeness and voice for the indie film As Deep as the Grave. Kilmer originally signed on to play a priest before his death in 2025, and his estate approved the AI recreation using family photos and audio under SAG guidelines. It marks one of the most prominent uses of posthumous digital casting to date.
CodeRabbit just launched Plan, a collaborative tool designed to turn vague ideas into structured prompts before coding agents touch your codebase. It pulls context directly from repositories, tickets, and Notion to generate editable phased plans, cutting down on rework and dev churn.
MiniMax just launched M2.7, a reasoning-focused model that autonomously handled 30% to 50% of its own reinforcement learning workflow, including debugging and metric analysis. It hit a 66.6% medal rate on MLE Bench Lite and scored 56.22% on SWE-Pro. Why it matters: it marks an early shift toward recursive self-improving agent architectures, available now via API.
Anthropic just dropped Dispatch, a research preview inside Claude Cowork that keeps a persistent Claude conversation running on your computer. You can text it prompts from your phone on the go and return to finished tasks later. It essentially turns your desktop into a remote AI workstation, though you'll need Claude Desktop paired with your phone to use it.
Unsloth just launched Unsloth Studio, an open-source local UI that turns PDFs or CSVs into datasets and fine-tunes models 2x faster with 70% less VRAM. It packages visual dataset creation, training, and GGUF export for 500+ open models into one interface, and is available now on GitHub.
Chrome 146 just rolled out experimental native support for WebMCP, turning web pages themselves directly into MCP servers. Instead of relying on clunky CDP setups or external backends, agents can now speak directly to frontend JavaScript to navigate, read cookies, and complete tasks with higher accuracy. It is currently locked behind a flag, but it paves the way for frictionless browser automation.
SegviGen just open-sourced a 3D part segmentation framework built on TRELLIS.2 that splits meshes into individual components via single clicks or 2D guide images. It outperforms previous SOTA models by 40% in interactive segmentation while requiring far less supervision, though local inference demands at least 24GB of VRAM.
Google just open-sourced Spark VSSR, a powerful video super-resolution framework capable of sharpening degraded clips and restoring old films. The release includes inference models, full training code, and datasets, though running the 42.2GB package demands a high-end GPU. Basically, creators get state-of-the-art video upscaling that they can run and fine-tune locally.
Nvidia just natively integrated its CloudXR 6.0 streaming suite with Apple Vision Pro. Using visionOS dynamic foveated streaming, users can stream photorealistic 4K models directly from local or cloud RTX GPUs without dumbing down heavy 3D datasets. The Swift framework is available now, letting teams review massive industrial digital twins untethered.
Researchers just introduced World Reconstruction from Inconsistent Views, a new method that turns AI-generated videos into consistent, explorable 3D environments. It uses non-rigid geometry alignment to clean up visual glitches and contradictions across frames, working with almost any video model from Wan-2.2 to Genie 3. Basically, it turns off-the-shelf video generators into functional 3D world builders without retraining.
Mistral AI just unveiled Forge, a platform that lets enterprises build frontier-grade AI models using their own internal codebases, policies, and operational data. It supports pre-training, post-training, and RL across both dense and MoE architectures. Crucially, it is built agent-first, allowing coding agents like Mistral Vibe to autonomously handle hyperparameter tuning and synthetic data generation.
Google just expanded its Personal Intelligence feature to free-tier users in the U.S. across Search's AI Mode, the Gemini app, and Chrome. The tool pulls context from Gmail and Google Photos to deliver tailored recommendations and tech support based on past receipts or trips. Users retain full toggle control over connected apps, though Workspace accounts are excluded.
Microsoft just overhauled its Copilot leadership, appointing Jacob Andreou to lead both consumer and commercial product experiences directly under CEO Satya Nadella. Meanwhile, Microsoft AI CEO Mustafa Suleyman is shifting focus away from front-end assistants to building Microsoft's proprietary foundation models. Why it matters: After years of disjointed consumer and enterprise bots, Microsoft is finally attempting to build one coherent, unified AI system.
US senators just demanded ByteDance immediately shut down its Seedance 2.0 AI video generator over widespread copyright infringement. The backlash follows viral clips featuring Thanos, Superman, and fake celebrity brawls. While ByteDance paused the rollout and promised stronger guardrails, lawmakers dismissed the pledge as a delay tactic.
AI giants are deploying quadrupeds like Boston Dynamics' Spot to patrol sprawling data centers around the clock. The robot dogs cost between $175,000 and $300,000 each and flag blind-spot hazards like leaks or heat anomalies. Operators claim the investment pays for itself in roughly two years by trimming physical guard headcounts.
Google just open-sourced the Colab MCP Server, enabling any MCP-compatible AI agent to programmatically control Google Colab notebooks straight from your terminal. Tools like Claude Code or Gemini CLI can now create cells, run Python scripts, and render visualizations live in the cloud. It basically turns Colab into an automated sandbox so agents no longer run risky, resource-heavy code on your local machine.
Anthropic just published findings from qualitative interviews with 80,508 Claude users across 159 countries and 70 languages. The results show 19% want professional excellence and 14% seek help with everyday life management, while 81% say AI has already delivered on their vision. Why it matters: instead of grand sci-fi dreams, people mostly use AI to reclaim personal time and mental bandwidth.
Researchers Albert Gu and Tri Dao just released Mamba-3, an open-source state space model architecture under the Apache 2.0 license designed to outpace Transformers. Its MIMO variant hits a 57.6% average benchmark accuracy—a roughly 4% relative bump over Transformers—while matching Mamba-2's quality using half the state size. By introducing complex-valued states and cutting GPU idle time, it targets reasoning gaps and slashes inference latency.
Midjourney just launched alpha testing for its V8 image model, delivering roughly 5x faster generation than V7. The update significantly sharpens prompt-following, text rendering, and image coherence while adding native 2K resolution via a new --hd mode. Relax mode isn't supported yet, but older personalization profiles carry over seamlessly.
OpenAI just launched GPT-5.4 mini and nano, lightweight models built for high-volume workloads and tool-heavy coding. The mini variant runs over 2x faster than GPT-5 mini while nearly matching the flagship GPT-5.4 on SWE-Bench Pro, at just $0.75 per million input tokens. Nano drops the entry cost to $0.20, perfect for background subagents.
Researchers just introduced Seoul World Model (SWM), the first city-scale simulation grounded in a real metropolis. By using RAG to anchor video generation to real street-view photos, it produces stable, multi-kilometer trajectories without visual degradation—and even lets you spawn Godzilla into the skyline via text prompts.
Nvidia just introduced the Groq 3 LPU, its first inference chip born out of a $20 billion technology deal with startup Groq. Built into Groq 3 LPX racks alongside Vera Rubin GPUs, it delivers up to 1,500 tokens per second and 35 times higher throughput per megawatt to keep multi-agent systems talking without delay.
Mistral just unveiled Mistral Small 4, consolidating reasoning, multimodal vision, and agentic coding into a unified 119B MoE model with only 6B active parameters. It offers a 256k context window, a toggleable reasoning dial, and cuts completion latency by 40% compared to version 3. Best of all, it's released under Apache 2.0 and beats competing models with far fewer output tokens.
Memories.ai just partnered with Nvidia to build a dedicated visual memory layer for smart glasses and robots. Powered by Nvidia's Cosmos-Reason 2, the system indexes first-person video so physical devices can actually recall what they see, tackling a major blind spot left by text-only AI memory.
Okara just launched AI CMO, which deploys six autonomous agents to manage your entire marketing pipeline around the clock from a single URL. For $99 a month, it covers traditional SEO, copywriting, and social distribution across X, Reddit, and Hacker News, alongside Generative Engine Optimization (GEO) for AI search tools like Perplexity.
Encyclopedia Britannica and Merriam-Webster just sued OpenAI for "massive copyright infringement," alleging it scraped nearly 100,000 articles without permission to train its models. The suit claims ChatGPT starves publishers of revenue through verbatim outputs and RAG search, while falsely attributing hallucinated answers to their brand.
MiroMind just launched MiroThinker-1.7 and MiroThinker-H1, open-source heavy-duty research agents that bake real-time verification directly into their reasoning loops. Packing 235B parameters alongside a 61GB mini variant, H1 tops demanding benchmarks like BrowseComp by auditing intermediate steps and proof trails before serving answers.
A European producer secretly used Suno to build Neon Oni, a fictional Japanese metal band that drew over 80,000 monthly Spotify listeners. After Reddit users spotted telltale AI hands in music videos, the creator leaned in—hiring seven Tokyo musicians to play the tracks live onstage. It turns out generating fake music online might actually create real-world gig jobs for performers.
Emil Kowalski just open-sourced a collection of design and motion skills for AI agents. From fixing awkward easing curves to Apple-grade fluid transitions and mobile polish, the rules inject real design engineering taste into AI code generation.
Alibaba just open-sourced Fun-CineForge, a multimodal dubbing model that syncs speech directly to characters' lip movements and timestamps. By treating time as an explicit modality, it handles fast speaker switches and facial occlusions cleanly, turning automated film dubbing into a realistic studio workflow.
Tsinghua University researchers just open-sourced OpenMAIC, an AI platform that turns any topic or document into a full interactive classroom. Instead of watching static video lectures, users learn alongside AI teachers and classmates who speak, draw formulas on whiteboards, and engage in real-time debates. It basically shifts online learning into an active virtual classroom under a permissive MIT license.
Nvidia just launched Space Computing to power orbital data centers and autonomous satellites. The lineup features the Space-1 Vera Rubin Module, which delivers up to 25x more space-based inferencing compute than the H100, allowing LLMs to run directly off-planet. Partners like Planet and Axiom Space are already onboard to process data at the source.
Nvidia just introduced DLSS 5 at GTC, calling it the company's biggest graphics breakthrough since real-time ray tracing in 2018. DLSS 5 introduces a real-time neural rendering model that infuses scenes with photoreal lighting and materials at up to 4K resolution. Why it matters: DLSS is moving beyond upscaling and frame generation into generative AI that turns 3D frames into cinematic visuals. It arrives this fall.
Nvidia just launched NemoClaw at GTC, a dedicated software stack for the OpenClaw agent platform. It packs Nemotron models and a new OpenShell isolated runtime into a single-command install, allowing autonomous agents to run 24/7 on local RTX PCs or DGX systems with strict privacy guardrails for cloud routing. Jensen Huang notably dubbed OpenClaw the OS for personal AI.
NVIDIA just launched the Nemotron Coalition, teaming up with Mistral AI, Cursor, Perplexity, LangChain, and other top AI labs to build open frontier foundation models. Their first milestone is a joint base model co-developed with Mistral AI on DGX Cloud, which will be open-sourced to underpin the upcoming Nemotron 4 model family.
Tencent just open-sourced Covo-Audio-Chat, a 7B end-to-end audio language model. It processes continuous audio directly in a unified architecture to enable native, low-latency full-duplex conversations that can listen, speak, and handle interruptions on the fly. Weights and code are now available on Hugging Face and GitHub.
Moonshot AI just introduced Attention Residuals (AttnRes), a drop-in replacement that swaps standard fixed residual connections for dynamic attention across depth. This lets layers selectively pull from earlier representations without suffering from representation dilution. Its practical block variant matches a baseline trained with 1.25x more compute, delivering sharp jumps in complex reasoning and coding benchmarks.
An Australian tech founder used ChatGPT and AlphaFold to analyze his dog's tumor DNA and target mutated proteins, teaming up with UNSW researchers to develop a bespoke mRNA vaccine in two months. Rosie's tumors have dramatically shrunk, showing how AI can rapidly accelerate and democratize personalized cancer therapeutics.
Researchers at UMass Amherst just built an artificial neuron operating at biological voltage near 0.1 volts, cutting power consumption by 100x compared to earlier designs. Driven by bacterial protein nanowires, it successfully held real-time electrical conversations with living cells. Why it matters: Matching biology's native voltage could let future neural implants ditch bulky signal amplifiers.
Texas A&M researchers just open-sourced Pulse of Motion, a lightweight AI designed to fix "chronometric hallucinations" in generated videos. Instead of guessing playback speed, the sub-2GB model calculates the true physical frame rate directly from visual dynamics and retimes the clip so movements actually obey gravity.
Niantic just revealed that Pokémon Go players helped build a massive dataset of over 30 billion real-world images through everyday gameplay. The company is now using those scans to power visual navigation for delivery bots, letting them pinpoint locations without GPS. Millions of casual monster hunters essentially mapped the physical world for commercial AI.
ByteDance just paused the global rollout of its Seedance 2.0 video generator after facing fierce pushback from Hollywood and U.S. senators. Lawmakers demanded the app be shut down immediately over unauthorized use of copyrighted characters and actor likenesses. ByteDance says it is working on stronger safeguards.
Retime GS just showed how to build smooth, full-body 3D animations from regular 2D video frames. Unlike conventional 4D reconstruction tools that glitch when frames drop, it uses continuous-time 3D representations and improved 4D Gaussian Splatting to seamlessly fill in missing motion.
Meta is reportedly planning sweeping layoffs affecting 20% or more of its nearly 79,000 employees to offset soaring AI infrastructure expenses. Executives have told leaders to begin planning cuts, though the timeline and final numbers remain unfixed. Why it matters: Tech giants are facing massive AI capital expenditures, and executives increasingly expect AI-assisted workflows to justify running far leaner organizations.
Perplexity just integrated NVIDIA’s Nemotron 3 Super across its entire product suite, rolling it out to the main app, Agent API, and Perplexity Computer. The addition gives both search users and developers building automated agent workflows immediate access to NVIDIA's high-performance model alongside existing options.
Meta just partnered with News Corp, Le Figaro, Prisa, and Süddeutsche Zeitung to deliver real-time international news and lifestyle stories directly inside Meta AI. When users ask news-related questions, the assistant will now cite these sources and link back to their websites. Why it matters: Meta gets more reliable, timely information while publishers get a direct pipeline for traffic.
Andrew Yang just urged the US to stop taxing human labor and start taxing AI companies benefiting from automation. Speaking on CNBC, he warned that the next six months of AI progress could eclipse the past decade, shifting budgets away from human workers. With tech layoffs accelerating, the push to replace payroll taxes with an AI levy is gaining fresh political momentum.
Meta has reportedly delayed its next-generation AI model, codenamed "Avocado," pushing its mid-March release to May at the earliest. Internal benchmarks showed the model lagging behind rivals from Google, OpenAI, and Anthropic in reasoning and coding, leading leadership to even weigh licensing Gemini.
Elon Musk just revealed that xAI "was not built right" and is undergoing a ground-up rebuild after 9 of its 11 original co-founders departed, including Grok Code lead Guodong Zhang. Musk admitted Grok is lagging in coding, prompting raids on Cursor talent as the startup reshuffles ahead of a massive planned IPO.
Anthropic just made the 1M-token context window generally available for Claude Opus 4.6 and Sonnet 4.6, ditching long-context price premiums entirely. Single-request media limits also jumped sixfold to 600 images or PDF pages. Claude Code users on paid tiers now get the full window by default, putting an end to messy mid-task context compaction.
Manus just rolled out a desktop app for macOS and Windows, introducing a feature called My Computer. Instead of being trapped in a cloud sandbox, the AI agent can now execute terminal commands to organize local files, build desktop apps, and tap into idle GPUs. Every command requires explicit user approval before running.
Chinese researchers just taught a Unitree G1 humanoid robot to play tennis using only five hours of imperfect motion-capture data. Through a framework called LATENT, the robot can track footwork, return balls traveling over 15 m/s with 90% accuracy, and sustain multi-shot rallies against human players.
Researchers just introduced Terminator, a lightweight early-exit probe that detects when a reasoning model hits the final answer and stops generation on the spot. Across benchmarks like MATH-500 and GPQA, it cuts chain-of-thought length by up to 55% and doubles inference speed with virtually no accuracy loss. Code is expected soon.
