Invideo just teamed up with Anthropic to launch an AI Motion Graphics tool that generates professional animations from a single prompt. pitching it as "vibecoding for motion design," the feature bypasses After Effects, templates, and coding altogether. It essentially lets solo creators and small teams churn out agency-grade motion assets in seconds.
Suno just launched a new tool called "Sample," allowing anyone to turn a rough voice memo, guitar riff, or uploaded audio clip into a fully produced song in minutes. The AI handles the chopping and rearrangement traditionally done manually in a DAW, letting users adjust clip length and audio influence sliders. It essentially gives non-musicians an automated beatmaker right in the browser.
Elon Musk is considering merging SpaceX with xAI or Tesla ahead of an IPO planned for later this year, according to reports from Reuters and Bloomberg. A tie-up could directly fuel SpaceX's wild plans to launch data centers into space, coming right after Tesla pledged a fresh $2 billion investment into xAI. Why it matters: Musk may be setting up his biggest corporate consolidation yet, hitching Grok directly to Starship.
Apple just acquired secretive Israeli AI startup Q.ai for nearly $2 billion, marking the tech giant's second-largest acquisition ever behind Beats. The company specializes in machine learning for facial micro-expressions and "silent" voice input. Why it matters: This could fundamentally reshape hands-free interactions for upcoming wearables like AirPods and Vision Pro without requiring users to speak out loud.
NVIDIA just introduced Cosmos Policy, a robot control model built by fine-tuning its Cosmos Predict world foundation model. Instead of patching separate action modules onto vision-language models, it treats robot actions and physical states as latent video frames inside a unified diffusion process. It scored a record 98.5% success rate on the LIBERO benchmark, with weights and code now open on Hugging Face.
Google just rolled out Project Genie, an experimental research prototype powered by its Genie 3 world model, Nano Banana Pro, and Gemini. Instead of static 3D environments, it simulates physics and renders paths ahead in real time as you walk, fly, or drive through user-prompted worlds. For now, generations are capped at 60 seconds and limited to Google AI Ultra subscribers in the U.S.
Krea AI just launched Realtime Edit to all users, letting creators tweak and transform images with complex instructions on the fly. Rather than waiting around for generation queues, users can now adjust prompts and see visual edits reflected instantaneously.
Google just expanded Gemini-powered navigation to walking and cycling routes in Google Maps across iOS and Android. Pedestrians can treat it like a personal guide to scout neighborhoods and nearby food, while cyclists can check ETAs or fire off text messages completely hands-free without letting go of the handlebars.
ByteDance and Alibaba are reportedly prepping next-generation flagship models for release during the February Lunar New Year holiday. ByteDance is lining up Doubao 2.0 alongside major image and video upgrades, while Alibaba is readying Qwen 3.5 with targeted reasoning boosts. With DeepSeek V4 also rumored for mid-February, China's frontier AI race is entering an intense new clash.
OpenAI just announced that it will retire GPT-4o, GPT-4.1, GPT-4.1 mini, and o4-mini from ChatGPT on February 13, 2026, though API access remains unchanged. The company noted that daily GPT-4o usage has plummeted to just 0.1% as users shifted to newer releases with updated conversational styles.
Anthropic just published a comprehensive guide to building Claude Skills, teaching users how to lock down repeatable workflows and pair them with MCP integrations. The guide walks through technical specs, testing, and distribution, claiming you can build your first custom skill in 15 to 30 minutes using their skill-creator tool.
Anthropic just launched contribution metrics in Claude Code in public beta, letting engineering leads see whether AI is actually moving the needle. By linking with GitHub, the dashboard tracks PRs merged and lines committed with Claude's assistance—Anthropic claims its own devs saw a 67% jump in daily merged PRs. It is rolling out to Team and Enterprise workspaces.
Anthropic just published a study showing that junior developers using AI assistants scored 17% lower on coding mastery tests than those coding by hand. While AI only shaved about two minutes off task completion, it created a steep drop in debugging ability—unless users actively prompted the AI for conceptual explanations rather than pure code generation. Why it matters: Relying blindly on AI tools risks creating a generation of engineers unable to catch and fix AI-generated errors.
MiniMax just launched Music 2.5, an AI music model designed to solve creative control and audio realism. It introduces 14 structural tags for surgical section-by-section arrangement, a library of over 100 instruments, and lifelike vocal synthesis capturing natural breaths and vibratos.
Meta plans to ship new AI models and products in the coming months, with Mark Zuckerberg highlighting "agentic shopping tools" that leverage personal data to buy goods. To fund its superintelligence push, the company is hiking its 2026 capital expenditure to as much as $135B, nearly double last year.
Elon Musk just announced that Tesla will discontinue the Model S and Model X in Q2 2026. The move aims to clear out factory floor space in Fremont specifically to mass-produce the Optimus humanoid robot. Musk is essentially abandoning Tesla's original luxury EV flagships to bet the entire company's future on AI and robotics.
Tesla just revealed that the third-generation, mass-production-ready Optimus humanoid robot will debut in Q1 2026 featuring an upgraded hand design. Elon Musk is even cutting Model S and X production in Q2 to clear Fremont factory floor space for a 1-million-unit annual capacity goal, though the project still faces skepticism following executive departures and earlier demo missteps.
YouTube just removed several of its biggest AI-generated slop channels, including one with over 5.9 million subscribers and 1.2 billion views. The platform confirmed it terminated multiple accounts for violating spam policies, following CEO Neal Mohan's pledge to curb low-quality AI filler. Why it matters: Even as platforms push generative tools for creators, they are finally drawing the line at automated, low-effort content farms gaming the algorithm.
Google just gave Chrome a massive AI overhaul powered by its flagship Gemini 3 model, adding a native side panel assistant and an agentic "auto browse" feature. The browser can now handle tedious multi-step workflows like researching flights, filling out forms, and finding deals across tabs. Why it matters: Chrome is evolving from a passive window into an active personal assistant, though the auto browse feature is initially restricted to U.S. AI Pro and Ultra subscribers.
Apple just rolled out major updates to Final Cut Pro and Logic Pro across Mac and iPad, weaving intelligent features directly into creative timelines. Video editors can now use natural language Visual Search and Transcript Search to locate actions or quotes in raw footage, while Logic Pro adds an AI Synth Player alongside semantic loop browsing. It essentially eliminates hours of tedious manual asset digging.
OpenMOSS just released MOVA, an open-source model that generates high-fidelity video and synced audio in a single pass. It runs on a 32B MoE architecture with 18B active parameters, handling precise lip-sync without stitching separate models together. Weights are live on GitHub and Hugging Face, though the 720p version requires serious GPU muscle.
robbyant just open-sourced LingBot-World, a real-time interactive world model built on Wan2.2. It generates controllable 480P and 720P environments from text or images at 16 FPS with sub-second latency, while retaining consistent scene memory over minutes. Why it matters: it gives developers a practical open alternative for interactive worlds and robot learning.
Google researchers just unveiled AlphaGenome in Nature, a unified DNA sequence model capable of taking 1 Mb inputs and predicting genomic regulation at single-base resolution across 11 modalities. It matched or outperformed top models in 25 of 26 variant effect benchmarks, taking under a second on an H100 GPU to decode the 98% of human DNA that doesn't code for proteins.
LM Studio just launched version 0.4.0, separating its core engine into a headless daemon named llmster that runs on cloud servers and CI pipelines without a GUI. The update also brings continuous batching for parallel inference via llama.cpp and a stateful REST API supporting local MCP tools. Why it matters: Local model testing can now scale seamlessly from personal laptops to remote headless rigs.
China's aerospace giant CASC just outlined plans to build space-based AI data centers over the next five years, tapping orbital solar power for an industrial-scale "Space Cloud." The move directly challenges Elon Musk's plan to launch solar-powered AI satellites with SpaceX. Why it matters: With ground-level power grids hitting hard capacity walls, low Earth orbit is fast becoming the next big cloud computing frontier.
xAI just released the Grok Imagine API, rolling out unified models to turn text prompts and static images into dynamic video with native audio. Built specifically to tackle rendering lag and heavy compute costs, it also enables localized object editing and stylistic scene transforms. Why it matters: by optimizing latency and price, xAI aims to let creative teams test parallel ideas without painful iteration bottlenecks.
Researchers just released TeleStyle, an open-source model that restyles both images and videos using reference pictures or text prompts. Built on Qwen-Image-Edit, it pairs curriculum learning with a dedicated video-to-video module to keep content and motion consistent across frames.
Alibaba's Tongyi Lab just open-sourced Z-Image, the undistilled foundation model behind its fast Turbo edition. Built specifically for customization, it fixes the Turbo version's repetitive faces and compositions while adding negative prompt support. It needs 50 steps instead of 28, and ComfyUI already supports it out of the box.
OpenAI just launched the EMEA Youth & Wellbeing Grant, committing €500,000 to support NGOs and researchers across Europe, the Middle East, and Africa. Grants will range from €25k to €100k to study AI's real-world impact on minors and develop protective tools. Why it matters: As AI moves deeper into classrooms, model makers need independent outside research to ensure safety guardrails actually work.
OpenAI just launched its EU Economic Blueprint 2.0, partnering with Booking.com to train 20,000 small businesses across six European countries. The rollout also commits a €500,000 youth safety research grant and highlights data on an AI "capability overhang" dividing member states. Why it matters: Free training and safety grants are OpenAI's latest strategy to stay indispensable amid tightening EU regulations.
Clawdbot (now Moltbot) just went viral as an open-source, self-hosted AI assistant that lets you control your local computer through Telegram or WhatsApp. It can run scripts, manage files, and even write its own code to build new skills without sending private data to the cloud.
The UK government just launched a free online AI training initiative open to all adults, offering bite-sized courses under 20 minutes to teach workplace basics like drafting text and admin tasks. The goal is to upskill 10 million workers by 2030—nearly a third of the nation's workforce—to make Britain the fastest AI adopter in the G7.
European Space Agency astronomers just deployed a neural network called AnomalyMatch across 35 years of Hubble archives. In just two and a half days, it plowed through nearly 100 million image cutouts to spot over 800 previously undocumented cosmic anomalies, from gravitational lenses to completely unclassifiable objects. Why it matters: AI is turning decades of dormant sky surveys into fresh scientific discoveries without requiring years of manual sorting.
Google just introduced Agentic Vision for Gemini 3 Flash, turning passive image parsing into an active investigation. Instead of guessing tiny details in a single glance, the model now writes and runs Python code to crop, zoom in, and draw bounding boxes on images before responding—delivering an instant 5-10% quality boost across vision benchmarks.
Google just integrated its Google Developer Program premium perks directly into Google AI Pro and Ultra subscriptions at no additional cost. Pro tiers now receive $10 per month in Google Cloud credits, while Ultra subscribers get $100 to spend on Vertex AI, Cloud Run, or Gemini API usage. The move helps developers transition prototypes from chat prompts directly to production without unexpected infrastructure bills.
Google just rolled out the ability to ask follow-up questions directly from AI Overviews, jumping straight into an AI Mode conversation while preserving context. The tech giant also made Gemini 3 the global default model powering these search summaries. Why it matters: Google Search is quickly shedding its static results heritage for a seamless, interactive chat flow.
Krea AI just kicked off beta testing for Realtime Edit, rolling out immediate access to the first 10,000 waitlist signups and all Max subscribers. The tool brings live image editing to the canvas, but everyone else will have to stay in line for now.
Airtable just entered the crowded AI agent arena with Superagent, marking the 13-year-old company's first-ever standalone product. Instead of sequential chat prompts, Superagent orchestrates multiple specialized agents in parallel to deliver interactive, presentation-ready research with data visualizations. It signals Airtable's aggressive pivot to an AI-native workspace, backed by a cash cushion despite tech valuation drops.
Chinese robotics startup Engine AI just partnered with commercial space firm Interstellor to build the world’s first humanoid robot astronaut. The plan is to send its 40kg PM01 robot into orbit to handle risky spacewalks and exterior repairs, keeping human crews out of harm's way.
Clawdbot just rebranded to Moltbot after Anthropic nudged the team over trademark similarities to Claude. Leaning into its lobster persona, the creators decided shedding the old shell fits their growth, keeping the mission of building AI that gets things done under the new handle @openclaw.
Life Biosciences just received FDA clearance to test cellular reprogramming in humans for the first time, starting with roughly a dozen glaucoma patients. The experimental ER-100 therapy injects three reprogramming genes into one eye, using a doxycycline antibiotic switch to control cell rejuvenation. Why it matters: If partial reprogramming restores vision without triggering tumors, it could validate a completely new frontier for treating age-related diseases.
Figure just unveiled Helix 02, expanding its unified neural network to steer an entire humanoid robot's locomotion and manipulation directly from pixels. In an impressive kitchen demo, Figure 03 autonomously completed a 4-minute dishwasher loading and unloading task, even bumping drawers shut with its hip when its hands were full. Why it matters: Replacing tens of thousands of lines of fragile C++ code with end-to-end learned control is a massive step toward humanoids doing real household chores.
Yahoo just debuted Yahoo Scout, a proprietary AI answer engine powered by Anthropic's Claude and Microsoft's Bing grounding API. Available in beta to nearly 250 million U.S. users, it synthesizes web data and Yahoo content into direct answers and lists across Search, Finance, and Shopping. Why it matters: Yahoo is weaponizing decades of user data and existing traffic to challenge standalone AI search engines.
Anthropic just partnered with the UK government to build a dedicated AI assistant for GOV.UK powered by Claude. The agentic system will initially guide job seekers through career advice and training resources while remembering context across visits. It marks a major milestone in bringing frontier AI safely into national public infrastructure.
OpenAI just introduced Prism, a free collaborative workspace built for scientific writing and powered by GPT-5.2. Built on its acquired Crixet platform, it brings full LaTeX support, paper context reasoning, and real-time collaboration to anyone with a personal ChatGPT account. Curiously, a banner already notes it will shut down on January 1, 2027.
Anthropic’s Claude just helped plan NASA’s first AI-generated Mars rover drive. Engineers at JPL used Claude Code to analyze terrain images and script navigation commands for Perseverance's 400-meter trek. After human operators confirmed the route with minor tweaks, NASA estimated the AI workflow could cut planning time in half.
Moonshot AI just open-sourced Kimi K2.5, a visual agent claiming top scores across multimodal and coding benchmarks. The headline feature is an "Agent Swarm" setup that runs up to 100 parallel sub-agents to finish complex tasks 4.5x faster. It is already accessible on kimi.com, with model weights published on Hugging Face.
Tencent just introduced ScriptAgent, a multi-agent framework that turns dialogue into cinematic videos. Three specialized agents handle scripting, orchestrate video models like Sora2-Pro and Veo3.1 across scenes, and critique visual quality. Why it matters: It universally boosts script faithfulness across top models, tackling one of generative video's biggest pain points—long-horizon narrative coherence.
Letta just launched Claude Subconscious, an open-source plugin that runs a secondary agent underneath Claude Code. It quietly studies your repo, stores context across 8 persistent memory blocks, and whispers guidance back into prompts without blocking your workflow.
Google DeepMind just debuted 'Dear Upstairs Neighbors' at the Sundance Film Festival, an animated short co-created with animation veterans. Instead of relying on text prompts, researchers built custom video-to-video workflows and local refinement tools powered by Veo and Imagen. Why it matters: It shows a practical path where professional animators keep precise creative control instead of wrestling with AI randomness.
Anthropic CEO Dario Amodei just published an essay warning that humanity is entering a turbulent “technological adolescence” as powerful AI approaches. He warns a “country of geniuses in a datacenter” could arrive within 1–2 years, arguing for sober, surgical regulation instead of polarizing doomerism.
Microsoft just introduced Maia 200, an in-house inference accelerator built on TSMC’s 3nm process. The chip delivers 30% better performance per dollar than Microsoft’s current fleet and boasts three times the FP4 performance of Amazon’s Trainium 3. It is already deployed in Iowa to power workloads like GPT-5.2 and Copilot, signaling hyperscalers’ aggressive push to cut reliance on commercial AI silicon.
Nvidia just launched the Earth-2 family of open models, delivering the world's first fully open, accelerated software stack for AI weather and climate forecasting. The suite lets organizations run and fine-tune models on their own hardware, cutting compute time by up to 90% in operational tests by meteorological agencies. Why it matters: It shifts heavy supercomputer weather forecasting into accessible AI workflows that any enterprise or government can run locally.
Open-source local voice tool Voicebox combines instant voice cloning with system-wide dictation, racking up over 42,000 GitHub stars. It can clone voices from a few seconds of audio, type into any text field via global hotkeys, and voice coding agents like Claude Code. Basically, it lets you ditch paid voice subscriptions and run your entire workflow locally.
Luma AI just launched Ray3.14, a major upgrade to its reasoning video model that outputs native 1080p footage. It delivers four times faster generation speeds at one-third of the previous per-second cost, cutting down on flicker and drift so studios can use raw generations directly in production.
Anthropic just rolled out interactive connectors in Claude powered by MCP Apps, allowing users to build Figma diagrams, draft Slack messages, and manage Asana timelines directly inside the conversation without tab-switching. Built on an open MCP extension, the feature is now live for all users across web, desktop, and mobile.
Alibaba's Qwen team just dropped Qwen3-Max-Thinking, a new proprietary reasoning model built around advanced test-time scaling and autonomous tool use. Paired with web search, it notched a 49.8 on Humanity's Last Exam, beating both Gemini 3 Pro and GPT-5.2.
Microsoft just open-sourced VibeVoice-ASR, a 9B speech model capable of transcribing up to 60 minutes of continuous audio in a single pass. Instead of slicing audio into chunks, it unifies transcription, speaker diarization, and timestamps into one end-to-end generation task. It basically solves context fragmentation for long meetings and podcasts.
Tencent just released HunyuanImage 3.0 Instruct, an open-weights image-to-image model that uses chain-of-thought reasoning to plan edit steps and preserve untouched regions before rendering. It handles lighting, perspective shifts, and multi-image merging, though self-hosting its 169 GB weights requires massive VRAM.
Gallup just released its Q4 2025 workplace AI data, showing frequent AI use among U.S. workers inched up to 26%, with daily users reaching 12%. However, the overall user base has stalled, with 49% reporting they never use AI at work. Why it matters: AI adoption is rapidly stratifying—it is booming among tech workers and executives, while leaving deskless and retail roles behind.
OpenAI’s GPT-5.2 was just caught citing Grokipedia—xAI’s controversial AI-generated encyclopedia—on sensitive topics like Iran and the Holocaust, according to tests by The Guardian. The catch is that Grokipedia previously faced backlash for citing neo-Nazi forums and questionable web sources. OpenAI responded that its model searches broad public web sources with safety filters in place.
Pope Leo XIV just warned that overly affectionate, always-available AI chatbots risk becoming hidden architects of our emotions, urging regulators worldwide to step in before humans form dangerous attachments. The tech-savvy pontiff also pushed for clear AI content labels to protect creators and called out the handful of tech founders who hold unchecked power over human behavior.
Claude in Excel just rolled out to all Pro plan subscribers alongside several handy workflow upgrades. The add-in now supports drag-and-drop for multiple files, automatically avoids overwriting your existing spreadsheet cells, and uses auto-compaction to keep longer data analysis sessions running smoothly.
Google DeepMind just introduced D4RT, a model that reconstructs and tracks dynamic 3D scenes directly from standard 2D video. By turning 4D perception into parallel spatial queries, it achieves up to 300x higher trajectory throughput than previous methods. Why it matters: it gives robots and AR glasses a path to real-time spatial intuition in moving environments.
Sakana AI just announced a strategic partnership with Google, alongside fresh funding from the search giant following its Series B round. The Tokyo-based lab plans to integrate Google's Gemini and Gemma models into its scientific discovery and agentic systems, while rolling out secure AI tools for Japan's banking and government sectors. Why it matters: Google gains a direct conduit into Japan's highly regulated industries through the country's leading AI lab.
Researchers just released RF-DETR segmentation under an Apache 2.0 license, introducing a new real-time vision baseline. It arrives in six model sizes, ranging from a Nano variant clocking 3.4 ms per image (40.3 mAP) to a 2XLarge variant reaching 49.9 mAP at 21.8 ms. Why it matters: Developers get an unconstrained, ultra-fast segmentation backbone built for edge devices and latency-critical vision tasks.
Runway just rolled out image-to-video for Gen-4.5, claiming only 10% of participants could distinguish its outputs from real footage. With the model topping Artificial Analysis video leaderboards, Runway warned that hyper-realism has hit a societal tipping point where seeing is no longer believing.
Alibaba just released Qwen3-TTS, an open-source speech model that clones any voice from just seconds of audio and lets you design new vocal styles using text prompts. Packing a dual-track architecture with 97ms end-to-end latency, it can fluidly generate cross-lingual speech and regional dialects for live streaming and AI agents.
Anthropic just open-sourced Claude’s Constitution under a CC0 license, an extensive manifesto written directly to the AI model rather than human readers. Instead of rigid dos and don'ts, it establishes a strict hierarchy prioritizing safety and ethics over helpfulness, openly considers whether Claude might possess consciousness, and explicitly directs the model to disobey Anthropic if ordered to do something shady.
YouTube just announced that creators will soon be able to generate Shorts using their own AI likeness, alongside text-to-game creation tools. To prevent AI slop and identity theft, the platform is pairing the feature with strict likeness-management controls. It's rolling out later this year.
Adobe just added generative AI features to Acrobat Studio, letting users convert lengthy PDFs and meeting notes into podcast-style audio summaries and presentation slides. Users can also edit documents—like swapping text or deleting pages—via chat prompts. The podcast feature currently taps into Microsoft and Google models.
Researchers just revealed a wearable AI smart choker that turns faint throat vibrations and pulse signals into fluent speech for stroke patients with dysarthria. Instead of invasive brain implants, flexible textile sensors capture silent mouthing and feed data to an LLM agent that expands brief efforts into emotionally attuned sentences. Why it matters: It offers a practical, non-invasive alternative to slow eye-tracking systems, though it is still in small-scale testing.
ElevenLabs just released The Eleven Album, a studio-grade musical release powered by its Eleven Music model and built with Grammy-winning legends. Tracks span rap, pop, and EDM, with artists retaining full copyright ownership and streaming royalties. It signals a notable pivot for AI music, moving from copyright lawsuits to permission-based industry collaboration.
Adobe just rolled out the Acrobat AI Assistant to turn static PDFs into interactive workspaces. The tool creates instant summaries with clickable citations, generates podcast-style audio recaps across multiple documents, and builds presentation decks from scratch. Existing Acrobat and Reader users can jump in right away, and Adobe guarantees your files won't train its AI models.
Google just launched free, full-length SAT practice tests directly inside Gemini, partnering with The Princeton Review for vetted test materials. Students can take exams on demand, receive immediate scores with answer explanations, and get personalized study plans. Why it matters: It opens up typically expensive test tutoring to anyone at zero cost, with more standardized tests planned.
Anthropic CEO Dario Amodei just dropped a striking timeline at Davos, saying AI could handle most, if not all, software engineering end-to-end within 6 to 12 months. With tools like Claude Code already reshaping workflows, that puts a surprisingly tight deadline on the future of development.
Researchers just introduced VideoMaMa, a model that uses generative video diffusion priors to turn coarse segmentation masks into pixel-accurate alpha mattes. Even with only synthetic training, it handles in-the-wild video effortlessly and powers a massive 50K-video dataset. All models and data are slated for public open-source release.
Researchers just open-sourced Motion 3-to-4, a framework that synthesizes coherent 4D dynamic objects from a single monocular video. By decoupling static 3D shapes from motion latent representations, it recovers per-frame vertex trajectories and can even transfer character motion into entirely new videos.
Researchers just introduced OmniTransfer, a Wan 2.1-based framework that unifies spatial appearance and temporal video transfer. It replicates complex VFX, fluid character motion, and cinematic camera moves directly onto new clips without manual pose or trajectory extraction.
Lightricks just rolled out an audio-to-video feature that lets creators generate video directly from speech, music, or sound effects. The model understands beats and rhythm to keep visuals in sync with the track, making it easy to generate music clips, b-roll, and digital avatars. Basically, audio is becoming the new visual prompt.
Adobe just launched an on-device AI Object Mask tool for Premiere, letting editors track complex moving subjects with a simple hover and click. Redesigned shape masks track up to 20x faster, and editors can directly import assets from Firefly Boards. After Effects 26.0 also launched with native 3D parametric meshes and variable font animation.
Overworld just dropped Waypoint-1, an interactive video diffusion model you can explore in real time using a keyboard and mouse. Powered by its new WorldEngine inference library, the 2.3B model hits up to 60 FPS on an RTX 5090 with near-zero control lag. Visual consistency and hallucinations are still rough around the edges, but you can already test the demo or run it locally.
LightOn just open-sourced LightOnOCR-2-1B, an Apache 2.0 end-to-end vision-language model that turns PDFs and scans into clean text without multi-stage pipelines. Despite its tiny 1B footprint, it beats Chandra-9B on OlmOCR-Bench and runs 1.73x faster than DeepSeekOCR. The release also includes bounding-box variants and a massive 16M-page training dataset.
Remotion's Agent Skills now lets AI coding agents like Claude Code script, animate, and render polished motion graphics entirely from plain text prompts. In a recent demo, Claude generated React components, orchestrated 3D camera flips, and delivered an 8-second promo without touching After Effects. Basically, product demos and UI animations can now be vibe-coded on demand.
Meta just open-sourced ShapeR, a conditional 3D generation framework that reconstructs casual video captures into individual, metrically accurate 3D meshes. By combining SLAM points with rectified flow transformers, it avoids baking an entire scene into a single uneditable blob. Why it matters: Creators and roboticists can now pull individual, real-scale 3D assets straight from phone-recorded video.
FlashLabs just open-sourced Chroma 1.0, a 4B end-to-end spoken dialogue model that directly understands audio inputs and replies with natural speech in real time. Built on Qwen2.5-Omni and Llama 3 under an Apache-2.0 license, it skips traditional TTS pipelines and supports instant voice cloning with a short audio prompt. It currently supports English and is hosted on Hugging Face.
Researchers just introduced VIGA, a multimodal agent that turns 2D images into fully editable 3D scene programs inside Blender. By combining interleaved multimodal reasoning with graphics engine feedback, it can assemble scenes from basic geometric primitives or plug in existing 3D tools, beating standard baselines by over 124% on BlenderBench.
Researchers just open-sourced CoDance, a new animation model that can animate multiple subjects from a single, potentially misaligned pose sequence. It breaks rigid spatial constraints and accurately rebinds motion to target characters using text prompts and masks. All code and model weights are now publicly available.
Replit just launched Mobile Apps, allowing anyone to build full-stack mobile applications using only natural language prompts. You can test your app in real-time on your phone by scanning a QR code with Expo Go, and push it directly to Apple's App Store in a few clicks. It basically turns mobile development into a conversation, bypassing native toolchains entirely.
Google just introduced TranslateGemma, an open translation model family in 4B, 12B, and 27B sizes built on Gemma 3. Distilled from Gemini, the 12B model outperforms the earlier 27B baseline across 55 languages while keeping multimodal support to translate text inside images.
HeartMuLa just dropped as a powerful open-source music foundation model, generating full songs with clean vocals and accompaniment directly from text and lyrics. The 3B parameter version supports multilingual lyrics, remembers melodies across song sections, and matches top commercial tools in lyric accuracy. It basically proves full-song generation is no longer exclusive to closed APIs.
Black Forest Labs just launched FLUX.2 klein, delivering sub-second image generation and editing on consumer GPUs with only 13GB of VRAM. It comes in an Apache 2.0-licensed 4B version and an open-weights 9B model, available via API, local runtimes, and a free web demo.
OpenAI just launched ChatGPT Translate, a standalone web tool supporting over 50 languages to directly challenge Google Translate. The tool lets users customize translation styles like "academic" or "business formal," with voice input enabled on mobile browsers. It's OpenAI's latest push to unbundle ChatGPT features into dedicated web utilities, though dedicated mobile apps and image support are still missing.
Black Forest Labs just launched FLUX.2 [klein], a speedy image generation family claiming sub-second inference that beats rival models by over 30%. It arrives in 4B and 9B variants, with the lightweight 4B edition released under a permissive Apache 2.0 license runnable on 8.4GB of VRAM.
NVIDIA just unveiled PersonaPlex, an end-to-end conversational model that listens and speaks simultaneously with seamless interruptions and backchanneling. Unlike past full-duplex systems locked to a single voice, it can adopt any persona and vocal style using simple text and audio prompts. Why it matters: It finally bridges human-like conversational rhythm with complete personality customization.
Researchers just introduced FlowAct-R1, a streaming humanoid video generation model capable of synthesizing infinite-duration video in real time. It achieves 25fps at 480p with a 1.5-second first-frame latency while maintaining natural behaviors. Why it matters: This drastically cuts latency for responsive AI avatars and seamless live streaming.
Thinking Machines Lab co-founders Barret Zoph and Luke Metz just left Mira Murati's fledgling startup to rejoin OpenAI, alongside early staffer Sam Schoenholz. The sudden move comes amid allegations that Zoph shared confidential data with competitors, though OpenAI dismissed the concerns and put him in a leadership role. Losing key technical founders is a massive blow to Murati's $12B venture in the middle of a brutal talent war.
Google just launched Personal Intelligence in beta, letting Gemini reason across Gmail, Google Photos, YouTube, and Search without requiring manual prompts. The tool pulls specific personal context from texts, photos, and videos to answer everyday questions. It is off by default, does not train models on your raw inbox, and is rolling out to U.S. Google AI Pro and Ultra subscribers.
Columbia Engineering researchers just built a flexible silicone robot face powered by 26 tiny motors that learned to lip-sync by watching YouTube videos instead of following rigid rules. After practicing its own facial mechanics in a mirror, its vision-to-action model beat five existing systems across 11 languages. Why it matters: Paired with ChatGPT, it could finally cure the stiff, uncanny lip-syncing that has plagued humanoids for years.
Browser Use just launched the beta of BU, a web agent platform aimed squarely at replacing Meta's newly acquired $2B startup, Manus. BU pairs browser automation with full local computer access, cutting task times to 33.4 seconds and costs down to 1.9 cents per run. It is an aggressive push to prove open-source roots can beat expensive proprietary wrappers.
