Amazon's Zoox just opened a 220,000-square-foot facility in Hayward capable of building over 10,000 robotaxis annually. The custom vehicles ditch steering wheels and pedals entirely, with plans to launch commercial rides in Las Vegas later this year. While it still trails Waymo by millions of rides, dedicated mass manufacturing turns up the heat on rivals.
Apple just introduced AI-generated App Store tags in the iOS 26 developer beta to boost app discoverability. Instead of basic OCR, Apple uses AI models to pull context from screenshots, descriptions, and other metadata without forcing developers to keyword-stuff their visuals. The tags are human-reviewed, and they haven't rolled out to the public search algorithm yet.
NVIDIA and Stability AI just introduced TensorRT FP8 optimizations for Stable Diffusion 3.5, slashing Large model VRAM needs from over 18GB down to 11GB while boosting performance by up to 2.3x. Developers can also grab the new standalone TensorRT for RTX SDK with JIT compilation. This basically puts heavyweight image generation within reach for standard consumer RTX rigs.
Elon Musk just shared that Tesla tentatively plans to kick off its robotaxi service on June 22 in Austin, Texas. The rollout will start small with 10 to 20 Model Ys across a limited radius, though Musk warned the date could shift as the team remains paranoid about safety.
Microsoft just rolled out an AI-powered search for browsing history in Edge Beta. It can find previously visited pages even if you use synonyms, vague phrases, or typos. Best of all, it runs entirely on an on-device model and your data never leaves the machine—sidestepping the privacy backlash seen with Recall.
Qualcomm just unveiled its Snapdragon AR1+ Gen 1 processor at AWE USA, shrinking chip size by 26% while powering on-glass generative AI without relying on a phone or cloud connection. During a live demo on RayNeo X3 Pro frames, it ran a local Llama 1B model entirely offline, taking smart glasses closer to true standalone devices.
Apple just launched the Foundation Models Framework at WWDC, giving third-party developers direct access to the on-device models behind Apple Intelligence. It supports offline tasks and tool calling with zero inference costs, with a public beta rolling out next month.
Apple researchers just introduced STARFlow, an image generation system combining normalizing flows with autoregressive transformers to rival DALL-E and Midjourney. Unlike iterative diffusion models, it enables exact maximum likelihood training in continuous spaces. Why it matters: it gives Apple a unique technical path that could play to its strengths in on-device AI and fine-grained control.
Apple just upgraded Apple Intelligence across its lineup, introducing systemwide Live Translation and on-screen visual intelligence. Crucially, third-party developers can now tap directly into Apple's on-device foundation model for free offline inference via Swift. Public rollout arrives this fall.
Microsoft just rolled out AI-powered Relight and natural language search to the Photos app on Windows 11 Copilot+ PCs. Users can place up to three customizable virtual lights on any shot and search their local library using plain descriptions like "sunset at the beach." For now, Relight is rolling out to Insiders on Snapdragon PCs, with AMD and Intel versions coming in the next few months.
Apple is set to expand Apple Intelligence at WWDC 2025 with major upgrades to translation and developer tooling, according to Bloomberg. Users could get automated Message translation, live phone call translation, and real-time conversation translation right through AirPods, alongside new APIs opening Genmoji-style tools to third-party apps.
Samsung just teased the next chapter of its Ultra lineup, promising to pack its top-tier specs and foldable-optimized Galaxy AI into a lighter, more compact body. The device promises voice-driven multitasking on the go and adaptive AI workspace features when unfolded. Ultra is finally going folding with smarter onboard models.
Smart pet tech company Fi just launched the Series 3+, a smart dog collar that uses AI to detect behaviors like barking, scratching, licking, eating, and drinking with 80% claimed accuracy. It pairs with Apple Watch and uploads vet records for context, aiming to catch illness before costly vet emergencies.
Google just quietly dropped Google AI Edge Gallery, an experimental app that lets users download and run open-source AI models directly on Android devices. It taps into Hugging Face to handle image queries, chat, and text tasks completely offline without cloud servers. Why it matters: It brings on-device AI to regular phones with zero privacy compromises, and iOS support is coming soon.
Microsoft just rolled out a suite of AI features to Windows 11 Insiders across classic native apps. Paint gains a prompt-based sticker generator and smart object selection, Snipping Tool introduces an AI-powered smart-snapping screenshot tool, and Notepad adds inline generative text drafting via Ctrl+Q.
Mistral just unveiled Devstral, a 24B coding model built with All Hands AI under a fully commercial Apache 2.0 license. It fits on a single RTX 4090 or a 32GB Mac, beating Google's Gemma 3 27B and DeepSeek's V3 on SWE-Bench Verified. Why it matters: Developers finally get a capable, agent-ready code model they can run locally without restrictive licenses.
NVIDIA just released Llama Nemotron Nano 4B, an open reasoning model tailored for edge agents. It delivers higher accuracy and 50% faster throughput than other sub-8B open models across coding, math, and function calling. It runs directly on Jetson and RTX GPUs for low-cost local reasoning.
Google just shared a closer look at Android XR, bringing Gemini directly onto headsets and lightweight smart glasses. The glasses use cameras and microphones to match your vantage point, demoing live in-person translation and hands-free directions with an optional in-lens display. Google is partnering with Gentle Monster and Warby Parker for consumer frames, with developer access opening later this year.
Google just unveiled Gemma 3n, a mobile-first open model designed to run rich multimodal AI entirely on phones and laptops. Thanks to a new Per-Layer Embeddings technique, its 5B and 8B models operate with dynamic memory footprints of just 2GB and 3GB. Why it matters: it makes complex, private local AI actually fit on everyday hardware.
Microsoft just baked its Phi-4-mini model directly into the Edge browser alongside new experimental AI APIs for prompting and text generation. Web developers can now tap local AI capabilities right on the client side without racking up backend costs or sacrificing privacy. The update also rolls out full-document PDF translation and brings Copilot agents directly to the Edge for Business sidebar.
