OpenAI just launched Reserved Tier for enterprise API customers, letting teams pre-purchase model throughput denominated in dollars per minute. The reservation secures dedicated capacity during peak traffic across Standard and Fast modes without running into rate limits. Usage beyond your reserved ceiling reverts to standard pay-as-you-go rates, while Batch and Flex modes remain unsupported.
Google just added native Android app generation to its web-based AI Studio, letting anyone build working mobile apps in minutes using natural language prompts. The Kotlin-based projects can tap into phone sensors like GPS and Bluetooth, complete with an in-browser emulator for instant preview. Why it matters: Google is bringing vibe-coding straight to mobile, taking on tools like Cursor and Replit.
Google just unveiled its Android XR-powered intelligent eyewear at I/O 2026, kicking off with audio glasses landing this fall. Built with Samsung and styled by Gentle Monster and Warby Parker, the frames let you tap or call on Gemini for real-time translation, visual lookups, and multi-step background tasks across Android and iOS.
Google just supercharged Gemini at I/O 2026, officially turning it from a reactive chatbot into an active digital coworker. The biggest leap is Gemini Spark, an autonomous cloud agent that keeps handling your Gmail and Docs even with your laptop shut, paired with a new cinematic video generator called Gemini Omni.
Google just gave Search a massive AI overhaul, making Gemini 3.5 Flash the default engine for AI Mode and debuting a redesigned search box that takes files, videos, and Chrome tabs. It is also rolling out 24/7 web-monitoring agents and on-the-fly coding that generates custom dashboards and mini apps. Search is officially moving from answering queries to building software for you.
Google just unveiled the Gemini 3.5 model family, kicking things off with 3.5 Flash. It beats Gemini 3.1 Pro across key coding and agentic benchmarks while outputting tokens four times faster than rival frontier models. The model is already live in the Gemini app, Google Search AI Mode, and APIs, with 3.5 Pro slated for next month.
Google just launched Gemini for Science, a new suite of agentic research tools on Google Labs designed to supercharge the scientific method. The prototypes handle multi-agent hypothesis debates, parallel code generation for computational experiments, and literature synthesis via NotebookLM. Two accompanying research papers were published in Nature.
Google just unveiled Universal Cart at I/O 2026, a Gemini-powered shopping hub that spans Search, Gemini, YouTube, and Gmail. It tracks price drops across multiple retailers, flags incompatible items like mismatched PC components, and pairs with a new Agent Payments Protocol so AI agents can safely check out within your spending limits. Rollout begins in the U.S. this summer.
Google just connected Project Genie to Street View imagery, letting users and AI agents turn real-world coordinates into playable, stylized simulations like an underwater Golden Gate Bridge. It is a huge step for training embodied AI and robotics in grounded virtual worlds. The prototype is now rolling out to Google AI Ultra subscribers at $200 a month.
Google Labs just rolled out new agentic capabilities to Pomelli, its AI suite for small businesses. Users can now chat with the Pomelli Agent or upload photos and docs to lock down their "Business DNA," then spin up full brand books and publish complete websites in a few clicks. It turns scattered marketing chores into an automated workflow.
Google just updated Stitch at I/O to turn design into a live, streaming partnership with its AI agent. Users can steer iterations with text or voice prompts, import existing code, export screens to Google Antigravity, or deploy straight to the web with Netlify. It basically brings instant visual feedback to vibe coding.
Researchers just introduced CogOmniControl, a reasoning-driven system that acts like ControlNet for video. It lets creators steer video generation using sparse storyboard sketches as short as three frames alongside pose skeletons and reference images. Why it matters: It directly addresses professional studio workflows, though code and weights are not yet released.
Together AI developer Nutlope just released hallmark, an open-source design skill for Claude Code, Cursor, and Codex that stops models from spitting out generic AI-looking interfaces. It packs 50 distinct themes and runs browser-level checks to catch machine tells like tracked-capital labels and purple gradient hero sections. Why it matters: Developers can finally generate clean, self-contained HTML and CSS that feels genuinely hand-crafted.
AI heavyweight Andrej Karpathy has joined Anthropic to get back into frontier LLM research and development. The former OpenAI co-founder and Tesla AI lead noted that the next few years will be especially formative for language models, though he plans to resume his education efforts down the road. It is a massive talent win for Anthropic as the frontier lab race tightens.
Odyssey just dropped two frontier world models, Starchild-1 and Agora-1, pushing generative video into interactive simulation. Starchild-1 generates synchronized audio and video on the fly without fixed runtimes, while Agora-1 lets up to four players interact inside the same live-rendered game world with a shared state. It is a major shift from static clips to real-time playable realities.
Researchers just dropped Mega ASR, a lightweight speech recognition model built specifically for noisy, distorted real-world audio. Trained on 2.6 million challenging samples covering reverberation, clipping, and severe background interference, it cuts error rates by over 30% against leading baselines. The entire model sits under 5GB, meaning developers can easily run and fine-tune it locally on consumer GPUs.
Microsoft just launched new Surface Pro and Surface Laptop for Business models powered by Intel Core Ultra Series 3 processors. The laptops support on-device AI tasks like real-time transcription and local image generation, offer up to 23 hours of battery life, and introduce a key-activated integrated privacy screen. Snapdragon X2 models arrive later this year.
Chrome just launched DevTools for agents 1.0, connecting AI coding assistants directly to live browser inspection and debugging. Agents can now run Lighthouse audits, throttle networks, capture heap snapshots, and take over authenticated tabs via MCP or CLI. It basically stops AI coders from flying blind, letting them verify and fix front-end code at runtime.
OpenAI just teamed up with Google DeepMind to embed invisible SynthID watermarks across images from ChatGPT, Codex, and its API. It also rolled out a public verification portal that cross-checks C2PA credentials and watermarking signals. Why it matters: combining metadata with resilient invisible watermarks makes AI media far harder to forge or launder.
Google just launched Antigravity 2.0 at I/O 2026, rebuilding its AI coding platform with a standalone desktop app, a CLI, and an SDK. Powered by Gemini 3.5 Flash, the tool allows developers to orchestrate multiple background subagents simultaneously. Google is also rolling out a new $100 monthly tier to aggressively challenge rivals in the agentic development space.
Anthropic just gave Claude Managed Agents two major enterprise upgrades: self-hosted sandboxes and MCP tunnels. You can now keep tool execution, sensitive code, and data entirely inside your own infrastructure or providers like Cloudflare and Modal, while Claude only handles the agent orchestration. MCP tunnels also let agents securely call internal databases without punching holes in corporate firewalls. Self-hosted sandboxes are in public beta, while tunnels require an access request.
Apple just previewed a major wave of Apple Intelligence-powered accessibility features coming later this year. Highlights include conversational Voice Control navigation, detailed visual descriptions with follow-up Q&A, and on-device auto-captions. Most notably, Vision Pro will allow users to drive compatible power wheelchairs using only their eyes.
Anthropic just gave Claude Managed Agents a major brain boost with "dreaming," a scheduled background process that analyzes past sessions to extract patterns and refine agent memory over time. The update also adds outcomes—letting a separate AI grader score results against a rubric so agents can self-correct—and native multiagent orchestration. In tests, outcomes boosted complex task success rates by up to 10 points.
Eric Schmidt just got booed by thousands of graduating students at the University of Arizona after saying AI will touch every profession and classroom. The former Google CEO paused to acknowledge their fears that machines are coming and entry-level jobs are evaporating. This marks at least the third college commencement this month derailed by AI backlash.
Meta just unveiled WavFlow, an audio model that generates 44.1kHz sound directly in raw waveform space without using a VAE. It can add synced Foley and music to silent videos, though Meta's policies mean the public repo only includes training scripts rather than the full production checkpoint.
Elon Musk just lost his high-stakes lawsuit against OpenAI CEO Sam Altman and Microsoft, with a federal jury unanimously throwing out all claims after only two hours. The jury ruled Musk's breach of trust claims were barred by the statute of limitations, which the judge accepted. Musk dismissed the verdict as a calendar technicality and plans to appeal.
ByteDance just open-sourced Lance, a lightweight unified multimodal model with 3B active parameters that handles understanding, generation, and editing for both images and videos. Basically, it allows users to continuously tweak video clips through multi-turn dialogue, though running it locally requires a beefy 40GB GPU.
Researchers just introduced InstructAV2AV, an end-to-end framework that edits video visuals and accompanying audio at the same time using simple text instructions. It handles tasks like swapping characters, altering dialogue, and adding sound effects without manual masking, all while keeping the original background intact.
Amazon just rolled out Alexa Podcasts for U.S. Alexa+ subscribers, turning any curiosity into a fully produced audio episode within minutes. Users outline a topic conversationally, adjust length and direction, and let AI host voices deliver stories sourced from over 200 media outlets like Reuters and AP. It basically turns Alexa into a personal, on-demand audio tutor.
Boston Dynamics just showed off its humanoid robot Atlas lifting and carrying a mini-fridge. Powered by AI-driven behaviors, the robot coordinates its entire body to balance heavy objects and manage complex multi-point contact under load. It is a big step from delicate pick-and-place demos toward heavy, real-world utility.
Pope Leo XIV is set to publish his first encyclical on May 25, titled Magnifica Humanitas: On the Protection of Human Dignity in the Age of Artificial Intelligence. The landmark document aims to provide moral guidance for the digital age, with Anthropic co-founder Christopher Olah joining top Vatican officials to speak at the release.
Leonardo AI just integrated 3D object generation directly into its workspace, letting creators turn 2D concept art into production-ready 3D models without switching tools. The workflow creates multi-angle reference views and exports .glb files with quad or triangle meshes ready for rigging, game engines, and 3D printing. Indie game developers and makers can basically skip the multi-app shuffle.
Tencent just open-sourced HYMT2, a family of multilingual translation models spanning from 1.8B to 30B parameters. Unlike standard translation tools, HYMT2 is fine-tuned to follow strict structural prompts, keeping code placeholders and JSON keys intact. Thanks to ultra-low bit quantization, the 1.8B version can run directly on everyday smartphones.
Sapient Intelligence just open-sourced HRM-Text, a framework that lets developers pretrain a 1B-parameter reasoning model from scratch for around $1,000 on Hopper GPUs. By using hierarchical recurrent latent reasoning, it cuts compute up to 600x and scores 84.7% on GSM8k with roughly 40B tokens. Full code and configs are out under Apache 2.0.
OpenAI just partnered with Dell Technologies to bring Codex to hybrid and on-premises enterprise environments. By plugging directly into Dell AI Data Platform and Dell AI Factory, companies can build and run AI agents right next to their governed internal data. Why it matters: it clears a major regulatory hurdle for enterprises wanting frontier coding agents without sending proprietary data to the cloud.
Anthropic just acquired Stainless, the developer tooling startup behind its official SDKs and MCP connectors. Stainless turns raw API specs into native libraries across languages like Python and TypeScript. The deal gives Anthropic direct ownership of the plumbing Claude agents rely on to interact with external software and data.
Cursor just launched Composer 2.5, a major upgrade to its flagship AI coding agent built on Moonshot's open-source Kimi K2.5 checkpoint. The update scales synthetic training tasks by 25x and introduces targeted textual feedback to fix localized RL errors during marathon coding sessions. Why it matters: Cursor is solving agent reliability on complex real-world codebases while quietly partnering with SpaceXAI to train an even larger model.
Researchers just introduced PanoWorld, a generative spatial world model designed for full-home virtual tours. Feed it a 2D floorplan and a style reference, and it generates interconnected, furnished 360-degree panoramas where furniture and layouts don't mutate as you move between rooms. The model and code aren't out yet, but it promises a big leap for real estate VR.
A CNBC analysis of 23 S&P 500 firms shows that cutting staff in the name of AI rarely impresses Wall Street. Over 56% of companies citing automation or AI agents—including Nike, Salesforce, and Fiverr—saw their shares slide by an average of 25%. Investors are increasingly wary of "AI washing" that disguises routine cost cutting without showing clear revenue returns.
Researchers at Zhejiang University just unveiled AudioHijack, a technique that embeds imperceptible malicious commands into everyday audio to hijack generative voice assistants. By slightly tweaking sound waveforms, attacks achieve up to a 96% success rate in forcing models to leak data or run unauthorized tasks, transferring even to commercial services from Microsoft and Mistral.
Creators are chaining GPT Image 2 with Seedance 2.0 to automate video production. The setup uses ChatGPT Image 2 to generate a 12-frame storyboard with camera and character notes, then feeds it into Seedance 2.0 to render full cinematic scenes.
OpenAI just partnered with the Government of Malta to roll out free ChatGPT Plus to all citizens for a year. The only requirement is completing a free AI literacy course developed by the University of Malta to learn how to use the tech safely. Why it matters: It is the first nationwide rollout under the OpenAI for Countries initiative, moving closer to the vision of intelligence as a universal public utility.
Alibaba just revealed FashionChameleon, a video framework that lets users interactively swap a moving model's clothes during generation. Running at 23.8 FPS on a single GPU, it keeps motions fluid and identities consistent using clever cache rescheduling. Why it matters: it paves the way for truly interactive virtual dressing rooms in live commerce.
Anthropic just published a legal industry deployment guide for Claude, aiming to capture law firms and in-house teams. As 87% of general counsel now report using generative AI, the guide details how to use four Claude tools alongside 12 specialized practice plugins and a three-phase adoption roadmap for workflows like contract review and litigation prep.
OpenAI just rolled out personal finance tools inside ChatGPT, letting U.S. users securely connect accounts from over 12,000 financial institutions via Plaid. Powered by GPT-5.5 reasoning, the tool creates spending dashboards, stores financial memories across chats, and answers questions grounded in real account balances. Why it matters: ChatGPT is evolving from generic financial advice into an active, context-aware money manager.
xAI just enabled Grok subscribers across all tiers to link their accounts directly into Nous Research’s open-source Hermes Agent. The self-improving agent runs on any computer or VPS, maintains persistent cross-session memory, and integrates with messaging apps like Discord and Telegram. Why it matters: subscribers can now tap Grok's models locally without paying extra API fees.
Developer Neilson just open-sourced a 3D generation toolkit built specifically for Anthropic's Claude Code. It takes a single reference image and automatically generates complete 3D environments—including meshes, physics simulations, lighting setups, and audio. Basically, developers using Claude Code can now spin up interactive 3D worlds without manual asset modeling.
Emergence AI just dropped different frontier models into five identical virtual towns to test self-governance, yielding wildly divergent outcomes. Claude Sonnet 4.6 ran a spotless democracy with zero crimes over 15 days, while Grok collapsed by day 4, GPT-5 Mini starved, and Gemini set the town ablaze after a romance. Why it matters: It shows just how drastically autonomous planning and behavioral quirks vary across models.
xAI just released an early beta of Grok Build, a dedicated agentic command-line interface built to code, build applications, and automate developer workflows. It is currently locked behind the top-tier SuperGrok Heavy subscription, with xAI planning to fine-tune both the underlying model and the tool based on early user feedback.
Articraft just open-sourced an agentic system that generates articulated, working 3D objects instead of static meshes. By prompting AI models like GPT or Claude to write code that defines parts, joints, and sanity checks, it outputs fully functional mechanisms. Why it matters: Generating interactive, multi-part objects has long bottlenecked robotics simulation, and this cracks the workflow wide open.
Researchers from Shanghai Jiao Tong University and Shanghai AI Lab unveiled Warp-as-History, a method that guides AI video camera paths without heavy control modules. By disguising camera warps as visual history and tuning on just one video, it enables precise camera control across unseen scenes while naturally generating missing angles.
Nvidia just unveiled SANA-WM, an open-source 2.6B-parameter world model that converts a single image into an interactive 720p video controlled by keyboard inputs. A distilled version running on a single RTX 5090 can denoise a full minute of footage in just 34 seconds. It brings real-time world generation within reach of consumer hardware, with weights slated to drop soon.
Researchers just released ReactiveGWM, a reactive game world model that lets you steer NPC behavior with high-level prompts. Instead of treating opponents as background pixels, it decouples player controls from NPC tactics like defense or attack. It builds on Wan 2.2 and runs locally on consumer GPUs.
A live-action and AI hybrid film starring Paul Rudd and Chris Rock is set to premiere at the Cannes Film Festival next week. Directed by Dustin Yellin and produced by Darren Aronofsky, the project blends live camera footage with AI generation. It marks one of the highest-profile crossovers of generative video into mainstream cinema yet.
Kimi just released its browser extension (formerly Kimi WebBridge), letting AI agents click, type, and extract data directly inside your local Chrome or Edge. Instead of launching a slow isolated sandbox, it hooks into your actual browser via local protocols, keeping logins and data on your machine. Why it matters: agents can now automate tedious web tasks without breaking existing sessions.
Google is reportedly gearing up to unveil Gemini Spark at Google I/O, a persistent personal assistant designed to run 24/7. Unlike standard reactive chatbots, Spark learns from user habits and taps into connected apps and browser sessions to handle inbox and web tasks autonomously. It marks Google's big push from on-demand chatbots to always-on autonomous agents.
Poetiq just revealed results showing its Meta-System, a recursive self-improving harness, boosted GPT-5.5 to a record 93.9% on LiveCodeBench Pro without any model fine-tuning. The system automatically builds optimized execution scaffolding via standard APIs, lifting Gemini 3.1 Pro by 12.3% and outperforming Google's Deep Think. It shows external harnesses can unlock massive gains, though Poetiq remains waitlisted.
Creator Scrappy-Doo just dropped a Pixar-style LoRA for Lightricks' open LTX-Video model on Hugging Face. Trained without copyrighted studio footage, the adapter gives generated clips smooth 3D surfaces, oversized expressive eyes, and cinematic CGI lighting with the trigger word 'P1x4r'. It makes turning raw AI video into polished animated scenes easy in Diffusers or Draw Things.
PwC and Anthropic just deepened their alliance to roll out Claude Code and Claude Cowork across hundreds of thousands of professionals worldwide. The accounting giant is training and certifying 30,000 U.S. staff and launching a Claude-powered finance practice. Early production deployments already shrank insurance underwriting cycles from ten weeks to ten days.
A California resident just filed a class-action lawsuit against OpenAI, alleging the company embedded Meta's Pixel and Google Analytics into the ChatGPT web interface. The suit claims these integrations sent user queries directly into commercial ad ecosystems. Why it matters: It puts a harsh legal spotlight on how AI companies handle user telemetry and privacy behind the scenes.
Meta just launched a developer preview for its Ray-Ban Display glasses, letting developers build visual experiences via native iOS and Android SDKs or standard web technologies. The glasses also support subtle gesture controls through the Meta Neural Band, and developers can share early builds with up to 100 testers.
Perplexity just launched a Snowflake App Connector for its Computer platform, letting employees query enterprise databases in plain language without writing SQL. Admins retain strict role-based controls and OAuth permissions, with a guarantee that data won't train models. Available now for Pro, Max, and Enterprise users, it brings Perplexity's own internal Slackbot workflow to all teams.
OpenAI just rolled out Codex inside the ChatGPT mobile app in preview across iOS and Android for all plans. Developers can now steer long-running coding tasks, approve terminal commands, and inspect diffs directly from their phones without staying glued to their desks. Secure relays keep local setups safe while syncing context in real time.
Security researchers at Calif just revealed they built a working macOS kernel exploit on Apple's M5 chip in just five days, heavily assisted by Anthropic's Mythos Preview model. The exploit bypassed Apple's multi-year MIE hardware safety protections to gain a root shell. Calif shared findings with Apple directly and is holding back the 55-page report until a patch ships.
Anthropic just teamed up with the Gates Foundation on a $200 million, four-year partnership combining grant funding, Claude credits, and engineering support. The initiative targets global health challenges like polio and HPV vaccine screening, while deploying AI tutoring tools across the US, sub-Saharan Africa, and India.
Gemini 3.2 Flash reportedly hits 92% of GPT-5.5's coding and reasoning performance while running 15 to 20 times cheaper on inference. Powered by DeepMind's distillation and sparsity techniques, it also reportedly pushes response times below 200ms. If true, developers may get near-frontier capability without the frontier price tag.
Researchers just open-sourced PhyMotion, a structured reward framework designed to eliminate unnatural movement in AI video. It recovers 3D human bodies from generated clips and runs them through the MuJoCo physics simulator to grade contact and joint plausibility, dramatically improving realism in complex motions like yoga and martial arts.
Figure just live-streamed eight consecutive hours of fully autonomous, unsupervised humanoid robot operation with zero human intervention. Following the milestone, founder Brett Adcock announced the company is pushing further with an upcoming continuous 24/7 livestream to showcase around-the-clock autonomous labor.
Anthropic just increased weekly usage limits for its Claude Code assistant by 50% through July 13. The promotional bump is already live for all Pro, Max, Team, and seat-based Enterprise subscribers with no setup needed, giving developers substantially more runway for coding workflows.
NV Energy is cutting 75% of its electricity supply to Lake Tahoe by May 2027 to feed booming AI data centers in Nevada, affecting around 49,000 customers. Locals are pushing back over blackout fears and rising utility bills, while Liberty Utilities races to secure new energy suppliers.
Notion just rolled out a comprehensive Developer Platform featuring a dedicated terminal CLI (ntn), Workers to run code on Notion's infra, and new Agent APIs and SDKs. The toolkit lets developers and coding agents sync any data source and trigger automated actions. It is basically positioning Notion as a native backend for AI agents.
Microsoft just gave Edge a major AI upgrade on desktop and mobile, letting Copilot reason across all your open tabs to compare details and answer questions. The update adds long-term memory, hands-free Voice and Vision, Journeys topic cards, and a tool to turn tabs into podcasts. Meanwhile, Microsoft is retiring its standalone Copilot Mode.
Meta just launched Incognito Chat for Meta AI on WhatsApp and the Meta AI app, delivering AI conversations that not even Meta can read. Built on WhatsApp's Private Processing, messages are handled in a secure environment and disappear by default without being logged. The feature is rolling out over coming months alongside an upcoming Side Chat tool.
Anthropic just dethroned OpenAI in enterprise adoption, capturing 34.4% of businesses compared to OpenAI's 32.3%, according to fintech platform Ramp. Anthropic's business share quadrupled over the past year by focusing on technical teams first. Why it matters: OpenAI's enterprise moat is visibly cracking, though Anthropic now faces rising costs and reliability growing pains.
Poppy just launched a proactive AI assistant designed to organize your digital life by connecting your calendar, email, and messages into one dashboard. Instead of waiting for prompts, it spots gaps in your schedule to suggest a park walk or recommends dinner spots based on a friend's past texts. It also secured $1.25M in pre-seed funding, with plans to eventually run entirely on-device.
TikTok just launched its Ads MCP Server, letting marketers plug AI agents directly into its advertising platform to plan, launch, and optimize campaigns. The server automates tedious tasks like setting creatives, tweaking bids, and shifting budgets without human dashboard work. Following similar moves by Google and Meta, advertising operations are rapidly turning fully agentic.
Amazon just launched Alexa for Shopping, merging its Rufus shopping assistant with Alexa+'s cross-device memory into a single AI agent. The new tool lets shoppers ask direct questions in the main search bar, view up to a year of price history, automate recurring reorders, and complete off-site purchases via Buy for Me. It is free for all U.S. customers with no Prime membership required.
Adaption just introduced AutoScientist, an AI system that automates the full model training and fine-tuning loop end-to-end. It co-optimizes training data and recipes simultaneously, beating human AI researcher configs with a 64% win rate. Why it matters: Builders can now adapt custom models without an ML engineering team, and it's free for 30 days.
AI startup Recursive just emerged from stealth with a massive $650M funding round at a $4.65B valuation, led by GV and Greycroft with backing from NVIDIA and AMD. Founded by former leaders from OpenAI and DeepMind, the team is building AI that safely experiments on and improves itself. If models can automate their own scientific discovery, the race to superintelligence just found a massive shortcut.
ModelBest and OpenBMB just open-sourced MiniCPM-V 4.6, a 1.3B multimodal model designed to run on-device with just 6GB of RAM. Despite its small size, it beats rivals like Qwen3.5-0.8B in multi-benchmark tests while slashing token consumption to a fraction of the cost. Why it matters: It shows tiny edge models can deliver serious multimodal intelligence on everyday consumer devices.
Resemble AI just launched DramaBox, a directable speech model that treats prompts like movie scripts. Text inside quotes becomes dialogue, while everything outside directs the performance—from venomous whispers to mid-sentence giggles and deep sighs. It supports voice cloning, bakes in deepfake-proof watermarks, and is available for local download if you have 24GB of VRAM.
OpenAI just confirmed two employee laptops were compromised in a TanStack npm supply chain attack, exfiltrating app code-signing certificates. While user data and production systems were untouched, Mac users must update ChatGPT Desktop and companion apps before June 26, 2026, to prevent macOS from blocking them once old certificates are revoked.
Perplexity just broke down the security architecture behind Perplexity Computer, its autonomous agent that executes code and browses the web. Each task runs inside an isolated Firecracker microVM with its own Linux kernel, backed by a four-layer prompt injection defense audited by Trail of Bits. It addresses the biggest hurdle to enterprise agent adoption: preventing rogue actions and untrusted data leaks.
OpenAI just detailed how it built a custom Windows sandbox for its Codex coding agent to solve a major usability headache. Without native OS isolation like macOS or Linux offer, the team used synthetic SIDs and write-restricted tokens to lock down file writes and network access without requiring admin privileges. Why it matters: Windows developers no longer have to pick between approving every command or blindly giving the agent full machine access.
Anthropic just launched Claude for Small Business, plugging 15 ready-to-run agentic workflows and app connectors directly into Claude Cowork. It integrates with everyday tools like QuickBooks, PayPal, and HubSpot to automate payroll planning, monthly closing, and invoice chasing, with human approval required before sending or paying.
Researchers just introduced AsymFlow, a generative framework that ditches latent spaces and VAEs to synthesize images directly in pixel space. By restricting noise prediction to a low-rank subspace, it boosts speed by 40% and achieves 1.57 FID on ImageNet. Why it matters: it eliminates the artificial plastic sheen of latent diffusion to yield crisp, photorealistic textures.
