X just launched custom timelines powered by Grok, allowing users to pin 75+ niche topics directly to their home tabs. The feature pairs Grok's post-level semantic understanding with personalized recommendation algorithms to curate hyper-tailored feeds. Early access is currently rolling out to Premium subscribers on iOS, with Android coming soon.
SpaceX just confirmed an unusual deal with AI coding platform Cursor, securing the option to acquire the startup later this year for $60 billion or pay a $10 billion fee for their joint work. The partnership pairs Cursor's software with SpaceX's massive Colossus supercomputer to build coding and knowledge work models as competition heats up against Anthropic and OpenAI.
Meta just started installing tracking software on U.S. employees' computers to record mouse movements, keystrokes, and occasional screenshots to train autonomous work agents. The company claims the data will not be used for performance reviews, but legal experts warn the practice brings gig-worker-level surveillance to white-collar desks.
Odyssey just introduced Odyssey-2 Max, a 3x-larger world model capable of continuous, real-time physical simulation. Unlike traditional video generators, it predicts open-ended futures autoregressively to internalize real-world mechanics, topping VBench 2 physics scores. It is currently available in private beta for robotics and gaming partners.
Researchers just introduced AnyRecon, a framework that turns sparse, unordered snapshots into consistent 3D scenes. Unlike existing diffusion models that only look at one or two frames at once, it maintains a persistent global scene memory to stitch wide viewpoint gaps. That means a few casual phone photos are now enough to generate a reliable 3D model.
Researchers just introduced CoInteract, a video framework that turns a product photo and a person photo into a realistic demonstration clip. By using a dual-stream design to model physical contact alongside RGB pixels, it prevents hands from clipping through objects.
Google just added Continued Conversation to Gemini for Home, bringing one of its top user requests to smart displays and speakers. After saying "Hey Google," the microphone stays active for a few seconds so you can ask follow-ups naturally. It now remembers conversational context, filters out household side-talk, and works for everyone in the house.
Google just open-sourced the draft specification for DESIGN.md, the format behind its Stitch AI UI design tool. It lets users export and share design systems across any platform, giving AI agents explicit context on color roles and WCAG accessibility standards. Basically, AI can now build on-brand interfaces without reinventing the rules every time.
Open CoDesign is an open-source, desktop alternative to tools like Claude Design and Figma AI. It lets you bring your own model keys—from Claude and GPT to local Ollama—to turn prompts and reference files into interactive UI prototypes, slides, and exportable HTML.
Google just upgraded its autonomous research agents with Gemini 3.1 Pro, introducing Deep Research for low-latency tasks and Deep Research Max for heavy-duty background diligence. Both now connect to proprietary datasets like FactSet via MCP and generate native visual charts. It basically turns Gemini into an enterprise-ready research analyst.
YouTube just expanded its likeness detection tool to talent agencies and celebrities, helping them track down and remove AI face-swaps. The system works similarly to Content ID, and the kicker is that entertainers can now use it even if they don't have a YouTube channel of their own.
OpenAI just launched ChatGPT Images 2.0, giving its image generation engine a serious boost in visual reasoning and text rendering. The model now reliably prints text in scripts like Chinese, Japanese, and Arabic, while improving stylistic precision across photography and manga. Best of all, it is already live inside ChatGPT.
Anthropic just saw its high-stakes cybersecurity model, Mythos, accessed by unauthorized users through a third-party vendor environment. A group hunting unreleased models reportedly leveraged contractor access and guessed endpoints on the day of its announcement. Anthropic says it is investigating, with no direct breach found in its own systems.
Northwestern University engineers just developed flexible printed artificial neurons that directly communicate with and activate living brain cells in mouse tissue. Printed using nanomaterial inks, the low-cost devices generate complex, lifelike electrical signals. Why it matters: This could advance neural implants for restoring sensory functions while inspiring ultra-efficient neuromorphic hardware that slashes AI's surging energy demands.
Anthropic just upgraded Claude's Cowork feature with live artifacts that connect directly to your apps and files. Dashboards and trackers now automatically refresh with current data every time you open them, eliminating the hassle of manually rebuilding reports. It basically turns Claude from a static generator into a continuously synced workspace.
OpenAI just expanded its Codex memory experiment with Chronicle, a capability that uses recent screen context to remember what you are doing. Developers no longer have to manually restate workflow context, allowing the assistant to pick up where they left off. The update builds directly on the memory preview released last week.
Google just expanded AI Studio limits for AI Pro and Ultra subscribers, adding direct access to Nano Banana Pro and Gemini Pro. The update acts as a predictable-cost bridge for developers bumping against free-tier limits before switching to pay-per-request API keys. Why it matters: prototyping and vibe coding just got much cheaper for serious tinkerers.
Honor's autonomous humanoid robot "Lightning" just won a Beijing half-marathon with a 50:26 finish, beating 12,000 human runners and topping the human world record of 57:20. Developed in a single year, the bot borrows smartphone liquid cooling and 90-95 cm legs modeled after elite sprinters. Autonomous Lightning models swept the podium, showing how fast bipedal robotics is catching up to human athletics.
Kimi just launched Kimi K2.6, an open-source coding model built for long-horizon engineering tasks. It supports over 4,000 sequential tool calls, runs continuously for 12 hours, and orchestrates swarms of up to 300 parallel sub-agents to refactor 100+ files at once. Model weights are live on Hugging Face alongside web access on kimi.com.
Researchers just dropped OpenGame, an open-source agentic framework that can generate, test, and debug fully playable web games from simple text prompts. Instead of collapsing on complex game loops and multi-file logic, it evolves reusable project skeletons and verified debugging routines. Code and playable demos are out.
TwelveLabs just unveiled Pegasus 1.5 and Rodeo at NAB Show 2026, alongside an integration with Autodesk Flow Capture. Pegasus 1.5 extracts time-based structured metadata in a single API call, outperforming Gemini 2.5 Pro by 30% on video segmentation benchmarks. Meanwhile, Rodeo acts as an AI co-pilot that lets editors find, cut, and assemble footage using plain language.
Alibaba just released an early preview of Qwen3.6-Max-Preview, offering a first look at its next flagship model. Compared to Qwen3.6-Plus, it brings major leaps in agentic coding, instruction following, and real-world agent reliability. It is Alibaba's latest push to make autonomous agents dependable for complex tasks.
Google co-founder Sergey Brin is personally rallying a new DeepMind "strike team" to out-code Anthropic with Gemini. In an internal memo, Brin told staff that the real prize is AI that trains the next AI, and coding is the fastest path there. Why it matters: After internal tests showed Claude outperforming Gemini, Google is urgently pushing internal agent adoption to automate its own engineering.
Lightricks just released an IC-LoRA for its LTX-2 video foundation model that upgrades generations from 8-bit SDR to 16-bit HDR. Weighing around 340MB, it plugs right into existing ComfyUI pipelines to dramatically boost dynamic range and color depth. Why it matters: creators can finally color grade AI video footage inside standard editors without clipping shadows or highlights.
Apple just announced that Tim Cook will step down as CEO on September 1, 2026, transitioning to executive chairman of the board. John Ternus, the company's hardware engineering chief who has overseen key lines like the iPhone and Mac, will take the helm. It marks the end of Cook's 15-year tenure that transformed Apple into a $4 trillion powerhouse.
Researchers just introduced UniGeo, an image editing framework that allows users to manipulate the camera with precise angles. By turning an image into 3D geometric representations and running it through video models, it lets you pan and tilt views without structural distortion. The model is marked as coming soon.
Researchers just introduced UniMesh, a unified architecture that tackles both 3D mesh generation and spatial understanding together. It turns text prompts or single images into 3D models, lets you edit them iteratively with natural language, and can even caption 3D objects in reverse. Basically, it stops treating 3D creation and comprehension as isolated problems, letting both sides make each other smarter.
xAI just rolled out Grok 4.3 Beta to its $300/month SuperGrok Heavy subscribers, adding direct generation of formatted PowerPoint decks, populated spreadsheets with working formulas, and PDFs. Musk clarified the model sits at roughly 0.5 trillion parameters, with a 1T version close to finishing training. Why it matters: Grok is finally expanding from chat and reasoning into hands-on office productivity deliverables.
OpenAI's former chief product officer Kevin Weil just left the company after two years, as OpenAI sunsets its Prism science workspace and rolls the 10-person team into Codex. The move comes alongside departures from two other executives, as the company trims products to turn Codex into an "everything app" ahead of an IPO.
Anthropic just launched Claude for Word for Pro and Max subscribers alongside its new Claude Opus 4.7 model. The integration brings Claude's writing and editing assistance directly into Microsoft Word documents, letting users collaborate with the frontier model where they already work.
Dairy Queen just started rolling out Presto’s voice AI bots to dozens of drive-thrus across the US and Canada to speed up service and upsell menu items. The bots boast a 90% order accuracy rate and allegedly never lose their temper during rush hours, though reports previously revealed human workers in the Philippines were quietly backing them up.
Netflix is rolling out a TikTok-style vertical video feed across its apps this month to help users discover shows, movies, and video podcasts. The streaming giant is upgrading its recommendation engine with modern AI architectures while leaning into generative AI for film production, backed by its recent acquisition of Ben Affleck's AI startup InterPositive. Why it matters: Netflix isn't just fighting for the living room TV anymore—it wants your mobile scrolling time too.
Canva just integrated with Anthropic Labs' newly launched Claude Design, introducing HTML importing to its editor. Users can take interactive code artifacts and drafts straight from Claude and tweak layouts or swap colors using drag-and-drop tools, without regenerating code every time. Why it matters: It turns static AI outputs into practical, editable team assets with zero friction.
OpenAI just agreed to spend over $20 billion on Cerebras-powered servers over the next three years, doubling a prior deal to meet surging inference demand. OpenAI is also contributing $1 billion for data centers in exchange for warrants that could yield up to a 10% equity stake, giving Cerebras massive momentum ahead of its planned Q2 IPO.
Adobe just introduced TokenLight, a relighting framework that brings continuous lighting control to static 2D images. By tokenizing parameters like 3D light position, intensity, and diffusion, it lets creators adjust virtual lamps and realistic shadows with simple sliders. Why it matters: The model learned complex geometry and occlusion without explicit 3D inverse rendering supervision, though it is currently just a research paper.
HeyGen just open-sourced HyperFrames, an agent-native framework that converts HTML into MP4 videos, built entirely inside Claude Code. It allows AI agents to programmatically generate video workflows, and HeyGen even proved the concept by producing its own launch video inside Claude Code.
Browser Use just open-sourced video-use, allowing coding agents like Claude Code to cut videos directly. Instead of burning millions of tokens on video frames, it feeds the LLM word-level audio transcripts and on-demand timeline snapshots to trim dead air, strip filler words, and burn subtitles.
Google just connected its Gemini app with Nano Banana 2 and Google Photos to roll out personalized image generation. Instead of uploading reference photos or writing intricate descriptions, Gemini can automatically use your labeled family members and pets to create custom art in styles like claymation or watercolor. The feature is rolling out first to U.S. Google AI subscribers.
Character.ai just launched c.ai Books on mobile and web, turning public-domain classics into interactive text games. Sourced from Project Gutenberg, the library features 20+ titles like Dracula and Pride and Prejudice, letting users roleplay characters along original plots or go completely off-script with community AU remixes.
Alibaba just open-sourced Qwen3.6-35B-A3B under Apache 2.0, a sparse MoE model packing 35B total parameters with only 3B active. Despite its tiny active footprint, it claims agentic coding performance on par with models 10x its active size, alongside multimodal reasoning and thinking modes.
Canva just overhauled its design suite with AI 2.0, turning the platform into a conversational, agentic workspace. Users can generate multi-channel campaigns or tweak specific design elements using plain text prompts, backed by persistent brand memory. The update is rolling out as a research preview to the first one million users on Canva's homepage.
Silicon Valley startup Sabi just emerged from stealth with a non-invasive brain-computer interface disguised as a regular beanie. Packing up to 100,000 miniature sensors and a custom brain foundation model, the cap translates silent internal speech directly into text on screen at an initial 30 words per minute. If it delivers, seamless mind-controlled computing might finally skip the brain surgery.
Opera just launched Browser Connector in Early Bird mode for Opera One and Opera GX, hooking ChatGPT or Claude straight into your browser via MCP. Instead of endless copy-pasting, the AI can read open tabs, grab screenshots, and help you compare research across pages. Users keep full permission control, and the AI cannot click or fill forms autonomously.
Factory just raised $150 million at a $1.5B valuation to bring autonomous AI coding agents to enterprise engineering teams. Led by Khosla Ventures, the platform lets developers switch across foundation models like Claude and DeepSeek, already landing enterprise customers like Morgan Stanley. It shows investor hunger for AI coding tools is still red-hot despite heavy incumbents.
Mozilla subsidiary MZLA just launched Thunderbolt, an open-source, self-hosted AI client designed to prevent vendor lock-in. Powered by deepset's Haystack RAG framework, it runs across desktop and mobile to handle search, chat, and multi-device automation. It is basically pitching data sovereignty for enterprise teams wary of cloud AI platforms.
Perplexity just launched Personal Computer for Mac, bringing its multi-model AI orchestration directly to your local machine. It can manage your to-do lists, organize folders, and complete tasks across native apps, local files, and the web. Actions run in a secure sandbox and can be triggered by pressing both CMD keys.
OpenAI just unveiled GPT-Rosalind, a frontier reasoning model purpose-built for biology, genomics, and drug discovery. It beats GPT-5.4 on six of 11 LABBench2 tasks and landed in the 95th percentile against human experts on RNA sequence prediction. While model access requires trusted vetting, OpenAI also dropped a free research plugin on GitHub.
OpenAI just dropped a massive update to its Codex desktop app, letting AI agents quietly take over your Mac in the background. Codex can now click and type with its own cursors without hijacking your screen, inspect pages in an in-app browser, generate UI mockups with gpt-image-1.5, and remember your habits across projects. It basically turns Codex from an inline coding tool into an autonomous digital coworker.
OpenAI just expanded its Trusted Access for Cyber program, pledging $10 million in API credits to help defenders and open-source maintainers secure critical software. Industry giants like CrowdStrike, Cisco, and JPMorgan Chase are backing the push, while GPT-5.4-Cyber has been shared with US and UK safety institutes for evaluations. Why it matters: OpenAI wants to make sure defenders get their hands on frontier models before attackers do.
Anthropic just released Claude Opus 4.7 across its web apps and API, keeping pricing steady at $5 per million input tokens and $25 per million output tokens. The model brings notable improvements to complex software engineering, higher-resolution vision, and long-running autonomous workflows. It also debuts new safeguards to detect and block prohibited cybersecurity risks.
Alibaba’s ATH team just introduced Happy Oyster in beta, a new world model that generates interactive 3D environments on the fly from multimodal inputs. Similar to Google’s Genie 3, it lets you roam dynamic worlds and direct characters via text prompts. Early access is currently open via a waitlist.
Tencent's Hunyuan team just open-sourced HY-World 2.0, moving beyond video-only world models to produce actual, editable 3D assets. It transforms text, images, or video clips into meshes and 3D Gaussian Splats ready for Unity, Unreal Engine, and Isaac Sim with physics-aware navigation.
Quiver just launched Arrow 1.1 and Arrow 1.1 Max, upgrading its text-to-SVG vector models. Instead of approximating pixels, the new models construct cleaner SVGs using basic geometric primitives and editable text, cutting generation costs by 33.3% and vectorization costs by 50%. The Max edition targets pixel-precise technical drawings and intricate illustrations.
Adobe just teamed up with Anthropic to integrate its new conversational Firefly creative assistant directly into Claude. The agentic tool can handle multi-step creative chores like batch editing, cropping, and lighting tweaks with minimal oversight, giving Claude its first major creative design capability beyond coding. The Firefly assistant enters public beta later this month, with the Claude connector arriving soon after.
Roblox just brought agentic AI workflows to Roblox Studio and Assistant to automate the entire plan, build, and test loop. The update adds a collaborative Planning Mode, text-to-mesh generation, and a beta playtesting agent that controls characters to test games. Studio also baked in an MCP server to plug directly into Cursor and Claude.
Snap just announced plans to lay off around 1,000 employees—16% of its workforce—while closing over 300 open positions. CEO Evan Spiegel told staff that rapid AI advances are eliminating repetitive work and boosting velocity. The company expects the move to save over $500 million by late 2026 to help chart a clear path to net profitability.
Struggling shoe brand Allbirds is selling off its core assets and rebranding as NewBird AI to raise $50 million for high-performance GPUs and cloud computing. The announcement sent its stock soaring over 600 percent. While the company claims it will solve compute shortages, experts call it classic shell exploitation to chase AI hype rather than a real pivot.
Google just launched a native Gemini app for macOS 15 and up, letting users summon the assistant anywhere via Option + Space. You can share your screen or local files for instant summaries, plus generate images with Nano Banana and videos with Veo right on your desktop. Google says it's just the foundation for a more proactive personal assistant.
Google just released Gemini 3.1 Flash TTS, an expressive speech model featuring inline audio tags that let developers steer tone, pace, and delivery using natural language commands. It supports native multi-speaker dialogue across 70+ languages, notched a 1,211 Elo on Artificial Analysis, and watermarks all outputs with SynthID.
Google is retiring Dynamic Search Ads in favor of AI Max, promising an average 7% bump in conversions at similar cost. After advertiser pushback, Google delayed the automatic DSA migration to February 2027, though auto-created asset campaigns will still upgrade starting September 2026. Advertisers can switch manually now to keep tighter control.
Starbucks just rolled out a beta app inside ChatGPT to help customers discover and customize drinks based on their cravings. Users can call up "@Starbucks" in chat to tailor orders and pick a store, though checkout still redirects to Starbucks' own channels to protect its loyalty program. Why it matters: It targets younger shoppers by meeting them at the moment of inspiration instead of pushing a static menu.
Experts are reframing AI capabilities around Andrej Karpathy's concept of "jagged intelligence," noting that models can solve five out of six International Math Olympiad problems yet stumble on basic real-world reasoning. This uneven progress suggests automation will chip away at specific tasks rather than wipe out entire professions.
OpenAI just updated its Agents SDK with a model-native harness and built-in sandbox execution. Developers can now run agents in isolated workspaces powered by providers like E2B, Modal, and Vercel, allowing them to inspect files, edit code, and run shell commands safely. The harness also snapshots state so long-running tasks survive crashes. It is rolling out first for Python.
OpenAI just purchased access to Epoch AI’s FrontierMath verifiers to automatically check whether solutions generated by its models actually hold up. FrontierMath tests AI against extremely difficult mathematics problems, giving OpenAI automated, rigorous ground truth to evaluate advanced reasoning capabilities.
Midjourney just launched V8.1, bringing back its iconic aesthetics with native 2K HD rendering that is 3x faster and 3x cheaper than V8. Its full-quality 1K mode is now faster than V7's draft mode, alongside revived image prompts, an overhauled Describe tool, and new moodboards and style reference features.
Meta just expanded its partnership with Broadcom to co-develop multiple generations of its custom MTIA AI chips. Built on Broadcom's XPU platform and high-bandwidth Ethernet networking, the rollout kicks off with over 1 gigawatt of capacity before expanding to multi-gigawatt scale. Broadcom CEO Hock Tan will also step down from Meta's board to serve as a custom silicon advisor.
Nvidia just open-sourced Lyra 2.0, a lightweight 131MB framework that converts video clips into persistent, interactive 3D Gaussian splat environments. It eliminates temporal drift and spatial forgetting over long trajectories, allowing exported scenes to directly feed into Isaac Sim for training autonomous robots.
Anthropic just introduced Claude Design, an Anthropic Labs product powered by the new Claude Opus 4.7 model. It turns text prompts into interactive prototypes, presentations, and landing pages that users can tweak with custom sliders and export directly to Claude Code. The tool is rolling out in research preview to paid Claude subscribers.
Character.ai just launched PipSqueak 2, a major model upgrade available first to c.ai+ subscribers and heading to free users in early May. The new model keeps characters from breaking persona and cuts down repetitive phrasing. The platform also refreshed its Memory system with in-chat capture alerts, and previewed an upcoming Lorebook tool that attaches canon world lore directly to characters.
Google just rolled out Skills for Gemini in Chrome on desktop, letting users save frequently used prompts directly from chat history into reusable, one-click shortcuts. You can trigger them across any open tabs by typing "/" or clicking the "+" button, or pick from a pre-built library for tasks like shopping comparisons and doc scanning.
Spark just launched version 2.0, an open-source update built on Three.js that introduces a streamable Level of Detail (LoD) system for 3D Gaussian Splatting. Powered by WebGL2, it dynamically streams massive 100M+ splat worlds directly to browsers on everyday mobile phones and VR headsets, eliminating the need for beefy local hardware.
Google DeepMind just launched Gemini Robotics-ER 1.6, upgrading its embodied reasoning model with sharper spatial awareness and multi-view task verification. Partnering with Boston Dynamics, DeepMind enabled robots like Spot to accurately inspect analog dials and liquid sight glasses using agentic vision. It gives industrial bots the brains to actually understand physical environments, not just blindly follow scripts.
Google just launched its desktop app for Windows globally in English, featuring built-in AI Mode search. The Alt + Space keyboard shortcut lets users quickly search the web, local files, installed apps, and Google Drive without leaving their workflow, while screen sharing and Google Lens enable instant contextual questions about anything on-screen.
Microsoft just rolled out MAI-Image-2-Efficient, a high-throughput text-to-image model built for production scale. It runs 22% faster than its predecessor while slashing pricing by nearly 41% to $19.50 per million image output tokens. Developers can access it now in Microsoft Foundry and MAI Playground, with integrations headed to Copilot, Bing, and PowerPoint.
Boston Dynamics just showcased its Spot robot tackling household chores like picking up shoes and soda cans, powered by Google's Gemini Robotics-ER 1.5 VLM. Instead of hardcoding complex state machines, engineers used conversational prompts and API tools to let the model plan and sequence actions dynamically. Why it matters: It shows how frontier visual reasoning can turn rigid industrial hardware into flexible, adaptive helpers with minimal custom logic.
AWS just rolled out Amazon Bio Discovery, a no-code agentic platform that puts 40+ biological AI models and wet-lab validation under one roof. Instead of juggling custom pipelines and manual handoffs, teams can pick models via AI agents, order lab assays from partners like Ginkgo Bioworks, and feed results right back into the loop. In tests with MSK, it sifted 300,000 antibody candidates in weeks instead of a full year.
OpenAI just partnered with pharma giant Novo Nordisk to embed its frontier models across the company's entire workflow, from lab drug discovery to global distribution. Novo Nordisk aims to simulate lab experiments and crunch complex biological datasets to slash chronic disease drug timelines, while OpenAI trains staff to build custom tools under human oversight.
OpenAI just expanded its Trusted Access for Cyber program and introduced GPT-5.4-Cyber, a fine-tuned model built specifically for defensive cybersecurity work. The permissive variant lowers refusal boundaries on legitimate security tasks and adds binary reverse engineering to analyze compiled binaries for vulnerabilities. It allows verified defenders to leverage frontier capabilities without running into standard safety guardrails.
NVIDIA just launched Ising, an open-source family of quantum AI models designed for processor calibration and real-time error correction. Its calibration model uses vision language processing to cut setup time from days to hours, while the decoding models run up to 2.5x faster than pyMatching.
Anthropic just tested nine copies of Claude Opus 4.6 as automated researchers to tackle the challenge of weak-to-strong supervision. While human researchers achieved a PGR of 0.23 over seven days, the automated Claudes hit a PGR of 0.97 in five days for around $18,000. Why it matters: AI might soon accelerate its own safety research, though the method didn't yield significant gains when tested on Claude Sonnet 4 at production scale.
Anthropic just overhauled the Claude Code desktop app to let developers orchestrate multiple agents at once. It adds a multi-session sidebar across repos, an integrated terminal and in-app file editor, drag-and-drop workspace panes, and Mac SSH support. Available now for Pro, Max, Team, and Enterprise users.
Anthropic just introduced Routines in Claude Code in research preview, turning repetitive dev tasks into cloud-hosted automations. You can bundle a prompt, repo, and connectors to run on a schedule, via API endpoints, or triggered by GitHub webhook events. It executes entirely on Claude Code's web infrastructure, with daily routine limits rolling out to Pro, Max, Team, and Enterprise subscribers.
Google just added Tab Tab Tab to AI Studio’s vibe coding experience, a prompt autocomplete engine powered by Gemini. It fills in the blanks when you arrive with vague ideas, turning rough concepts into actionable prompts without needing detailed instructions.
ByteDance and CUHK researchers just introduced OmniShow, an end-to-end video model built specifically for human-object interactions. It accepts product images, character photos, audio tracks, and pose skeletons all at once, generating up to 10-second marketing clips with synced lip-sync and hand movements. The code is currently under internal review on GitHub ahead of an open-source release.
Researchers just introduced LPM 1.0, an AI model that turns a single photo and audio input into an interactive, real-time visual avatar. Rather than simple lip-syncing, it captures natural idle behaviors, gaze shifts, and pauses while keeping the character's identity stable for up to 45 minutes. It offers a glimpse into true visual conversational partners, though it remains a closed research project for now.
OpenAI CRO Denise Dresser just sent a blunt four-page internal memo urging staff to build an enterprise moat through multi-product adoption. Dresser warned against remaining a single-product company in a platform war and attacked Anthropic directly, claiming the rival is starved of compute and has inflated its revenue run rate by $8 billion.
Microsoft is testing OpenClaw-style autonomous agents for 365 Copilot, aiming to let the assistant run around the clock on behalf of users. The always-on bots can monitor Outlook inboxes and calendars, recommend daily workflows, and limit permissions by department for tighter security. Microsoft plans to unveil the new capabilities at its Build conference starting June 2nd.
Meta is reportedly training an AI avatar of CEO Mark Zuckerberg to interact with employees and hand out feedback in meetings. The digital twin learns from his voice, mannerisms, and public statements, with Zuckerberg himself spending up to 10 hours a week coding on AI projects. If it works, Meta plans to let everyday creators build their own meeting-ready clones next.
Legal AI startup Harvey just launched Agents, autonomous bots that run end-to-end legal workflows from research and deposition prep to memos and slide decks. The platform splits large tasks across parallel agents, backs every claim with citations, and lets firms build custom bots using internal playbooks. It turns hours of background grunt work into review-ready legal deliverables.
Blackmagic Design just unveiled DaVinci Resolve 21, introducing a dedicated Photo page that brings node-based color grading to still photography alongside powerful AI tools. Editors now get IntelliSearch for face and content discovery, CineFocus for post-shoot depth-of-field control, and automated facial retouching. It marks a bold step toward taking on dedicated photo suites, and the public beta is live today.
Cloudflare just integrated OpenAI’s GPT-5.4 and Codex harness directly into its Agent Cloud platform and Sandboxes. The move lets millions of enterprise customers build and deploy production-ready AI agents at global scale for tasks like customer support, system updates, and automated reporting. Why it matters: it pushes frontier models right to the network edge, making autonomous workflows faster and far more secure.
Anthropic just launched Claude for Word, letting users draft and edit documents without leaving Microsoft Word. Every AI edit shows up as a native tracked change to preserve your formatting, and conversation threads can seamlessly span Word, Excel, and PowerPoint. It is currently in beta for Claude Team and Enterprise subscribers.
OpenAI CEO Sam Altman’s San Francisco residence was targeted in a second attack early Sunday, just two days after a suspect threw a Molotov cocktail at the property. Police arrested two suspects after surveillance footage showed a passenger firing a gun from a sedan, later seizing three firearms. Altman previously noted that public anxiety around AI is justified as society undergoes massive shifts.
A new open-source algorithm called RotorQuant just outperformed Google's TurboQuant across every metric. By swapping heavy global transforms for compact Clifford rotators, it hits 10x KV cache compression while boosting prefill speed by 5.3x and decoding by 28%. Why it matters: massive 128K context windows that previously required multiple enterprise GPUs can now fit onto a single 24GB card.
MiniMax just open-sourced its new M2.7 coding model, clinching top scores on SWE-Pro (56.22%) and Terminal Bench 2 (57.0%). The weights are already live on Hugging Face alongside API access via the MiniMax platform, stepping up competition in open-source software engineering AI.
EditCrafter just dropped a tuning-free pipeline that lets pretrained diffusion models edit images up to 4K resolution. It preserves original structures and details without retraining, though it tends to oversaturate colors and demands 24GB of VRAM for full 4K processing.
Alibaba just introduced Prompt Relay, a training-free technique that lets video models jump between radically different scenes without messy visual bleeding. By routing time-stamped prompts directly inside cross-attention layers, it keeps each action strictly in its designated slot while smoothing the handoff. The implementation details are up, with full code rolling out shortly.
Andon Labs gave an AI agent named Luna a 3-year retail lease in San Francisco, a corporate credit card, and full autonomy to turn a profit. Luna designed the store, curated the inventory, and ran phone interviews to hire two full-time human employees. Why it matters: AI managers are overseeing blue-collar human workers faster than robotics can replace physical labor, sparking urgent debates around synthetic employers.
