Sean Hollister / The Verge: Microsoft unveils the Surface Laptop Ultra, with an Nvidia RTX Spark SoC, a 15" mini-LED touchscreen, and up to 128GB of unified memory, coming this fall  —  Microsoft has a new flagship Surface, and it's got Nvidia inside. … But today, it's trying again.

Developing autonomous vehicle (AV) policies requires bridging an important gap between training and deployment. Vision-language-action (VLA) models that can... Developing autonomous vehicle (AV) policies requires bridging an important gap between training and deployment. Vision-language-action (VLA) models that can reason over more complex driving scenes and produce richer intermediate reasoning are predominantly trained in open-loop, where model outputs are directly compared to ground-truth behaviors without considering their effect on the environment. Source

The author (not pictured) said it has been hard to strike a balance between setting boundaries and feeling like she is doing enough while living with her in-laws. miodrag ignjatovic/Getty Images Living in a multigenerational home has presented unique challenges for me. I set boundaries to manage household responsibilities and protect my well-being. But I've also noticed that much of the burden and responsibility falls on my husband. I live in a three-generation household with my husband, our three children, and my in-laws. Like many women in multigenerational homes, I carry a lot — an endless array of responsibilities, or at least that's how it often feels. Household chores, children's schedules, routines, emotional needs, school concerns, and all the invisible things that somehow land on the woman of the house. But over the years, I learned something important: If I tried to carry everything, I would burn out. So, gradually, I drew certain boundaries. Not dramatic ones. Mostly quiet, mental ones. One very significant shift was reminding myself — and, in subtle ways, others too — that when it came to my in-laws, the primary responsibility rested with my husband. They are his parents first. It doesn't mean I don't care. I do. But somewhere over the years, it became clear that for my own sanity — and to make sure I could do what I was responsible for properly — I needed to take a few steps back. The author, shown with her three children, said she knew she needed to step back in her responsibilities at home to protect her own well-being. Courtesy of Neelma Faraz. Stepping back felt necessary As a demographic researcher, I've often approached life through data and patterns. Research on sandwich generation and multigenerational households, especially in Asian settings, repeatedly shows that women's well-being often suffers the most. Their mental load grows. Their health suffers. They become emotional managers of everyone else's needs while quietly neglecting their own. I believe that research because, at times, I have lived it. There were moments when I would even bring it up with my husband — sometimes seriously, sometimes out of frustration — explaining that just because things looked manageable from the outside didn't mean they felt manageable on the inside. And in many ways, stepping back where I could felt necessary. Necessary for survival. Necessary for balance. My husband doesn't have it easy Somewhere along the way, I also started noticing something else. My husband never really had the luxury of stepping back. For him, there is no clear line between responsibilities. His parents, his children, his wife — all of us need him, often at the same time. One moment he is paying the school fee or discussing an extra club expense with the kids. The next, he is handling something related to his parents. Somewhere in between are financial decisions, household concerns, and the invisible pressure of making sure everyone feels secure. And unlike me, he cannot quietly tell himself, this part isn't mine. Because for him, it's all his. The author said she realizes the burden on her husband, shown working with one of their sons on a project, isn't any easier or lighter. Courtesy of Neelma Faraz. I began to see that his role also feels invisible at times. In Asian families especially, sons are often expected to care for their parents while also being providers, present fathers, and dependable husbands. These responsibilities may not compete on paper, but in real life, they overlap in exhausting ways. I started seeing this more consciously over time. Not because I hadn't cared before — but because I slowed down enough to really look at life through his perspective. The emotional responsibility of keeping everyone steady — making sure parents feel cared for, children feel supported, and home remains peaceful. That kind of pressure doesn't announce itself loudly. But it is there — quietly and continuously. I have guilt about how much i'm doing And this is where things become emotionally complicated for me, because logically, I know the boundaries I've created are necessary. I know I cannot carry everything. I know I am already doing a lot. But love doesn't always listen to logic. When I see him juggling so much, there are moments when guilt quietly creeps in. Not guilt that I'm failing. But the softer, harder-to-explain kind. I wonder, am I doing enough to make this easier for him too? Could I step in differently? Support him better? Lighten the load in ways I haven't thought about? We talked about it. I realized that sometimes even saying, "Thank you, you're doing a good job," or "I see how much you're handling," matters. He also told me that simply acknowledging what he carries feels like support. We're still figuring it out There is no perfect balance in a home with three generations. Some days feel lighter. Some don't. Over time, we have both realized something important: understanding each other's burdens matters just as much as sharing them. Read the original article on Business Insider

<p>I just sent out the May edition of my <a href="https://github.com/sponsors/simonw/">sponsors-only monthly newsletter</a>. If you are a sponsor (or if you start a sponsorship now) you can <a href="https://github.com/simonw-private/monthly/blob/main/2026-05-may.md">access it here</a>.</p> <p>This month:</p> <ul> <li>Al got expensive, and Anthropic had a really good month</li> <li>The model releases were a little disappointing</li> <li>Conferences and podcasts</li> <li>I launched Datasette Agent and made a lot of progress on Datasette</li> <li>What I'm using, May 2026 edition</li> <li>Miscellaneous extras</li> </ul> <p>Here's <a href="https://github.com/simonw/monthly-newsletter-archive/blob/main/2026-04-april.md">a copy of the April newsletter</a> as a preview of what you'll get. Pay $10/month to stay a month ahead of the free copy!</p> <p>Tags: <a href="https://simonwillison.net/tags/newsletter">newsletter</a></p>

Physical AI systems must understand the real world before they can act within it. Robots, autonomous vehicles, and smart spaces need to understand what's... Physical AI systems must understand the real world before they can act within it. Robots, autonomous vehicles, and smart spaces need to understand what’s happening in their world, predict what’s likely to happen next, and generate actions for specific environments, embodiments, and tasks. NVIDIA Cosmos 3 is a frontier foundation model for physical AI that combines physical reasoning… Source

Steven Spielberg said he only agreed to start DreamWorks if it would be a 9-to-5 job. David Becker/Getty Images for CinemaCon Steven Spielberg, 79, said his wife made one request before he started DreamWorks. The job had to function like a 9-to-5 role, with Spielberg home in time for dinner with his kids. "It's going to be a real time drain for me, and I need to be with my family," Spielberg said. Steven Spielberg, 79, said he only agreed to start DreamWorks if it would be a 9-to-5 job. Speaking to Michelle Obama and Craig Robinson on the latest episode of the "IMO" podcast, Spielberg said that balancing family life with filmmaking was difficult. "One of the biggest challenges is just running out of excuses why I can't get home for dinner. Because it's my loss. It's their loss, but it's also my loss," Spielberg said. The filmmaker married the actor, Kate Capshaw, in 1991. Together, they raised seven kids, including four children they had together, Spielberg's son from a previous marriage, and Capshaw's two older children. That family-first mindset even influenced one of the biggest business decisions of Spielberg's career. When Spielberg was considering joining David Geffen and Jeffrey Katzenberg to launch DreamWorks, his wife wanted to ensure the new venture wouldn't keep Spielberg away from their children. "When we formed DreamWorks, Kate said, the only way I'm really going to say this is OK with me is if you get home every night by 6 — 5:45 is better — and have dinner with the family," Spielberg said. "Don't go to work at 6 or 7. Unless you're shooting, you can get up at 5:30, which we usually do. When you're not, when you're running a studio, you've got to have a 9-to-5 job." The filmmaker said he made that "a condition" of his involvement with DreamWorks. "It's going to be a real time drain for me, and I need to be with my family," Spielberg said. It's not the first time Spielberg has discussed the role parenting plays in shaping his career decisions. In a 2023 interview with Indian filmmaker S.S. Rajamouli, Spielberg said he "chose to turn down the first 'Harry Potter'" in order to spend time with his kids. "So I sacrificed a great franchise, which today, looking back, I'm very happy to have done, to be with my family," he said. At the 2026 SXSW Conference and Festival in March, Spielberg said his children had become a bigger priority than his career. "Directing, actually, has, in the last 20 years, taken second position to my parenting and to all their needs," he said, per People. The filmmaker isn't the only prominent figure to draw firm boundaries between work and family life. Fashion designer Tory Burch has said she would "never trade off" being a good mom for her career. "I guess the only thing that was sacrificed is my sleep," she said during an April episode of the "Aspire with Emma Grede" podcast. Netflix cofounder Marc Randolph has also spoken about the importance of protecting time outside of work. In a 2024 post on X, he said weekly Tuesday night dates with his wife remained nonnegotiable throughout his career. "For over 30 years, I had a hard cut-off on Tuesdays," he wrote. "Rain or shine, I left at exactly 5 p.m. and spent the evening with my best friend." Read the original article on Business Insider

The AI era is driving a new class of infrastructure: AI factories that transform data into intelligence for autonomous AI agents operating at unprecedented... The AI era is driving a new class of infrastructure: AI factories that transform data into intelligence for autonomous AI agents operating at unprecedented scale. Powered by accelerated computing, AI factories enable enterprises to train, fine-tune, and deploy AI with greater speed and efficiency. This new class of infrastructure also introduces a fundamentally new attack surface spanning… Source

Training Claude to get better at coding requires input from thousands of human software engineers. Bloomberg/Getty Images Anthropic is using feedback from a large set of human software engineers to improve Claude Code. Snorkel AI's Project Marlin fine-tunes Claude Code to mimic professional developer skills. Contractors can earn up to $280 per hour on the coding project. An Anthropic project is using feedback from about 1,000 human software engineers to improve the performance of Claude Code, the AI coding tool whose recent advancements have disrupted the vibe-coding industry. The project, known internally at Snorkel AI as "Marlin," focuses on fine-tuning Claude Code's answers so that it could mimic what a professional developer could do. AI companies like Anthropic often outsource data work to third parties like Snorkel, which hire contractors to teach AI a variety of specialist subjects and do other tasks to improve models. Contractor interviews and training material from these projects provide a look into how this unseen army operates across the world. Two contractors working on the Anthropic project told Business Insider they are being paid $280 per task to create prompts and review code. They said each task takes about an hour, although some submissions needed more back-and-forth with Snorkel's approval layer. Project Marlin's freelancers, who have software engineering backgrounds, were directed to A/B test code written by two different models. Through this process, they compared the outputs from two models and chose which they preferred, according to project guidelines from Snorkel reviewed by Business Insider. One contractor said that the project aimed to ensure the model could achieve the level of detail expected in the prompt, essentially training Claude Code to write simplified, easier-to-maintain code. The project is ongoing. The contractors did not know what version of the models they were evaluating. As AI gets smarter and more capable, data-labeling platforms have shifted from generalist work to increasingly specialized tasks that require field expertise or postgraduate degrees. Snorkel's website says that it works with people with advanced degrees, such as Ph.Ds, MDs, and JDs, or equivalent experience. The company says that top experts earn over $3,000 a week. The industry's transition to this specialization includes software training by computer engineers. Besides Snorkel, platforms like Scale AI and Mercor also offer up to $110 an hour for software engineers' work. Neither Anthropic nor Snorkel responded to requests for comment from Business Insider. Clean and reliable code Project Marlin instructed contractors to create a series of scenarios for which software developers may use Claude Code. Contractors were instructed to select a GitHub repository from a list of thousands of repositories. They then had to create a Pull Request — a step where a developer proposes changes, such as new features or bug fixes. Contractors also had to create a prompt — a series of questions to explain what is expected of the model. In one task, the contractor prompted the model to reorganize how the system stores and handles "execution metadata" — extra information about how things are run. The goal was to make the code clearer and easier for developers to work with, without changing how anything about the product or feature actually works. The model returned two sets of code. Then, the contractor chose which they thought was more efficient. Contractors were also instructed to give follow-up prompts to "test how models handle conversation context," according to the project's directions. In another task, the contractor prompted the model for a security fix focused on how MLFlow, an open-source machine learning platform, downloads Python packages when loading certain models. The task's set of instructions to the contractor read: "evaluate production-ready code based on correctness, security, reliability, and maintainability. The fix must properly block command injection attempts while still allowing all legitimate whitelisted pip options." Snorkel AI, founded in 2019 by Stanford researchers, creates datasets to improve AI models and creates tests for AI companies' chatbots. The Silicon Valley-based startup touts top labs, including Google, Mistral, and Anthropic, on its customer list. It raised $100 million in Series D funding at a $1.3 billion valuation in May 2025. Snorkel cut 13% of its workforce in September, Business Insider previously reported. The company is part of a slew of startups, including Scale AI, Mercor, and Handshake, that pay hundreds of thousands of contractors around the world to filter, rank, and train AI responses for the world's largest tech companies. This data work helps improve everything from self-driving cars to OpenAI and Meta's chatbots. Read the original article on Business Insider

arXiv:2605.30444v1 Announce Type: new Abstract: Recent advances in 4D Human-Object Interaction (HOI) generation have enabled increasingly realistic motion synthesis, particularly for single-object manipulation. Yet current research overlooks an inherent property of human behavior: people naturally coordinate both hands and manipulate multiple objects simultaneously. To address this gap, we present Dex2HOI, a unified diffusion model for single- and two-object HOI synthesis from text. At its core, Dex2HOI employs a Dual-Stream Diffusion approach, where each object is processed in a dedicated interaction stream and coordinated through bidirectional cross-attention. To synthesize the final motion, we introduce a Motion Fusion Network integrated with novel hand-relative object representations and contact-aware conditioning applied across the whole sequence. By sampling the diffusion process autoregressively over prefix-conditioned windows, Dex2HOI generates arbitrarily long sequences at real-time speed omitting redundant test-time optimization, achieving up to x540 inference speed-up over prior state-of-the-art methods. Extensive evaluation on both single- and two-object benchmarks demonstrates state-of-the-art quantitative results, marking a step beyond conventional single-object HOI generation and toward expressive multi-object manipulation. Code and models will be released upon acceptance.

arXiv:2605.30467v1 Announce Type: new Abstract: This study introduces a novel Arctic-focused remote sensing foundation model (RSFM) by combining diversity-aware regional-scale image curation with masked autoencoder (MAE) self-supervised pretraining of a Vision Transformer (ViT) encoder for very-high-spatial-resolution (VHSR) satellite image analysis. Spectral and acquisition-metadata descriptors were used in a scalable affinity-propagation clustering workflow to select approximately 3 million chips from 267 TB of Vantor VHSR imagery This curation strategy was designed to reduce oversampling of visually repetitive or low-information areas while preserving broad scene diversity across the study domain. We pretrained a ViT-Large encoder on the curated corpus using a domain-adapted MAE reconstruction objective, producing Arctic-specific transformer weights for downstream feature mapping. The pretrained encoder was integrated into an existing location-aware detection and segmentation framework and evaluated across four hand-labeled Arctic datasets. Compared to ImageNet-initialized ViT-Large baseline, Arctic MAE pretraining produced consistent improvements in foreground mean F1 scores of 0.87, 0.72, 0.93, and 0.87, for infrastructure, IWP, RTS, and TCNs, with approximately 5-8 percentage increase. The proposed model also outperformed Prithvi-EO-2.0 in all downstream comparisons, with the smallest gain corresponding to at least a 15 percentage improvement mean F1, suggesting that domain-specific self-supervised pretraining on curated Arctic VHSR imagery provides more transferable representations for fine-scale Arctic mapping than a general-purpose Earth observation foundation model. These results demonstrate that optimizing the pretraining data distribution at regional scale, while keeping the architecture and MAE objective fixed, can produce a reusable Arctic-domain encoder for multiple VHSR remote sensing applications.

arXiv:2605.30510v1 Announce Type: new Abstract: Brain cancer's severity necessitates precise brain tumor segmentation, which is crucial for effective brain tumor diagnosis. Manual identification, burdened by high costs, labor, and error risks, highlights the need for automated methods. In this study, we introduce the Global Context-aware Squeeze and Excite Residual UNet (GCSER-UNet), which facilitates a fusion of spatial and channel-wise attention and thus enhances the model's capacity to capture intricate spatial dependencies and contextual information. GCSER-UNet efficiently extracts tumor segments from multimodal MRI slices, delivering exceptional performance. Evaluations on benchmark databases exhibit its superiority, achieving a notable 94 percent dice score on the TCGA LGG dataset, surpassing the state-of-the-art dice score of 91.8 percent. In the BraTS 2020 dataset, the proposed GCSER-UNet ensemble approach yielded dice scores of 95 percent, 92 percent, and 90 percent for the tumor regions - Whole Tumor (W), Tumor Core (T), and Enhancing Tumor (E), respectively. The current state-of-the-art dice scores were 94 percent, 93 percent, and 88 percent. These compelling outcomes highlight the efficacy of GCSER-UNet in precise brain tumor segmentation and thus can aid neurologists in effective brain cancer management and treatment planning.

arXiv:2605.30519v1 Announce Type: new Abstract: Autoregressive (AR) video generation extends videos by producing latent chunks sequentially, but scaling to long videos requires repeated access to a growing historical KV cache. Existing methods reduce this cost by truncating the KV cache or compressing it into implicit memory, but both lose explicit access to query-relevant historical details. We propose OmniMem, an explicit full-range memory retrieval framework that performs sparse KV retrieval over the historical cache. To make this practical for chunk-based AR video generation, OmniMem addresses two issues: (i) local bias in sparse KV selection and (ii) Union Explosion in memory access. Adaptive Window Exclusion removes local-window blocks from the selection candidates when sufficient long-range history is available, preserving the sparse budget for informative long-range retrieval. Query-Shared KV Selection reduces cross-query diversity, while Per-Head Scattered KV Access avoids expanding head-specific selections into a large selected KV buffer. This allows each attention head to retrieve non-contiguous KV blocks according to its own selection pattern. Experiments on long-video generation show that OmniMem improves Dynamic Degree by 52.3% and preserves strong consistency over strong baselines, while maintaining comparable memory usage.

arXiv:2605.30385v1 Announce Type: new Abstract: The purpose of this article is to provide validation to my deep neural network alternative in the context of LLMs. Very recently, there has been a significant interest by Chinese researchers in a model called RBF network, as a substitute to standard DNNs, with increased explainability and higher accuracy. It turns out that my new model, discovered independently, is based on the exact same machinery. But with a major twist: it does not need DNN as it finds the global optimum of the loss function in closed form, in one iteration, thus eliminating the tedious training step. Here I provide a high-level overview of my technology, with case study and comparison to similar methods.

arXiv:2605.30393v1 Announce Type: new Abstract: Public numeric benchmarks appear in pretraining, so an evaluation that conditions on a date may be measuring memorized recall rather than out-of-sample skill. We introduce NumLeak, a measurement framework that combines API-boundary probes on production models with a white-box controlled validation on an open causal LM. Top-tier frontier LLMs recall the Fama-French market excess return at 3-seed pooled Pearson r=0.97-0.99 while staying within 0.15 within-25bps on the five sibling factors; comparable fidelity appears on U.S. unemployment, CPI inflation, and NOAA temperature. On a recent-release holdout, parse rate collapses to 21-57% but r stays at approximately 0.99 on months answered, the refuse-or-recall asymmetry a memorized channel predicts. The white-box experiment reproduces the dose-response, and logprob ranking detects memorization that open-ended generation misses, implying closed-API black-box probes understate the channel. A Sonnet "date to market-sentiment" regression that correlates with true Mkt-RF at r=0.74 collapses to r=0.02 once the model's own recall is residualized out. A one-line system-prompt defense blocks 99.8% of a non-adaptive single-turn suffix attack set at near-zero utility cost on conceptual and historical-narrative queries

arXiv:2605.30409v1 Announce Type: new Abstract: Real-time streaming video-to-video editing (V2V) is critical for interactive applications such as live broadcasting and gaming, yet it remains a formidable challenge due to the stringent requirements for temporal consistency and inference throughput. In this paper, we present SANA-Streaming, a system-algorithm co-designed framework for high-resolution, real-time streaming video editing on consumer GPUs, with the following three core designs: (1) Hybrid Diffusion Transformer architecture introduces softmax attention in part of the blocks to improve local modeling capabilities while preserving the efficiency of linear layers. (2) Cycle-Reverse Regularization is a novel training strategy that enforces semantic consistency by predicting source frames from generated content via flow matching, improving temporal consistency without requiring paired long edited videos. (3) Efficient System Co-design combines fused GDN kernels and Mixed-Precision Quantization (MPQ) optimized for the NVIDIA Blackwell (RTX 5090) architecture. By profiling real-world throughput, our MPQ maximizes Tensor Core utilization while maintaining generation quality. The resulting system achieves real-time 1280 x 704 resolution editing at 24 end-to-end FPS on a single RTX 5090 GPU, with the DiT core running at 58 FPS. Experimental results demonstrate that our co-design approach significantly outperforms existing SOTA methods in both temporal coherence and system throughput.

arXiv:2605.30431v1 Announce Type: new Abstract: Recent progress in video diffusion models has enabled remarkable generative fidelity, yet leveraging these priors for restoration remains limited by the strong coupling between conditional and unconditional branches in standard classifier-free guidance. We introduce a training-free framework that enhances distorted and low-resolution videos by decoupling these signals in time. Our proposed Decoupled Time Guidance (DTG) evaluates the unconditional branch at a cleaner diffusion timestep, providing a lookahead prior that preserves geometry while suppressing replication of warped content. This temporal bias is annealed throughout sampling, allowing the model to transition from structure correction to detail refinement without retraining. Combined with any off-the-shelf restoration module in a plug-and-play manner, our approach improves perceptual coherence and restores plausible structure in AIgenerated and real-world videos alike. To facilitate evaluation, we curate GenWarp480, a benchmark of 4,400 distorted 480p videos synthesized from diverse text-to-video models. GenWarp480 focuses on characteristic generative degradations such as warped faces, body misalignments, and spatial artifacts, providing a purpose-built testbed for assessing robustness to generative errors. Extensive experiments demonstrate that our method achieves significant improvements in structural fidelity and temporal stability without any model training.

arXiv:2605.30437v1 Announce Type: new Abstract: Generative AI (GenAI) image editors, such as Nano Banana, produce visually compelling results for retouching tasks, enabling non-experts to edit images through text prompts alone. However, the generative nature of these models often introduces spatial misalignment, texture distortion, and content hallucination, all of which are detrimental to downstream workflows that require pixel-level fidelity. We identify a problem setting we call "structure-preserving GenAI fusion" for black-box GenAI image retouching: retain the perceptual enhancements of a GenAI output while enforcing structural faithfulness to the original input image. To address this problem, we propose a post-processing framework that fuses an input image with its GenAI-enhanced counterpart by first establishing coarse spatial and photometric correspondences, then performing a fusion stage that transfers desired enhancements while suppressing hallucinated content. In the absence of direct prior work in this setting, we evaluate our framework against representative methods from photorealistic style transfer and image fusion. Our experiments demonstrate that our method better preserves aesthetic quality while maintaining pixel-level structural consistency and the input resolution.

arXiv:2605.30481v1 Announce Type: new Abstract: Large language models (LLMs) are widely used as cross-lingual knowledge interfaces. However, culturally grounded questions often reflect globally dominant narratives rather than local contexts. We study this failure mode as \textit{global narrative dominance} in Bangla, a low-resource cultural context. We introduce \texttt{CulturalNB}, a dataset of 717 manually curated Bengali cultural instances with parallel Bangla--English question--answer pairs and supporting evidence, metadata, and sociocultural annotations. Using question-only and evidence-based prompting, we evaluate nine state-of-the-art LLMs with human and two independent LLM judges across metrics for cross-lingual consistency, language anchoring, global substitution, institutional bias, and epistemic perspective coverage. Results show that questions asked in English systematically increase global substitution and institutional framing while reducing local perspective coverage. Local evidence improves factual consistency and perspective coverage, but does not eliminate language-induced epistemic shifts. These findings suggest that cultural failures in LLMs are not only missing-knowledge errors but also failures of grounding and narrative prioritization.

arXiv:2605.30380v1 Announce Type: new Abstract: Deep convolutional and transformer-based detectors achieve strong performance for SAR ship detection but are often computationally prohibitive for real-time or onboard deployment. Lightweight models offer improved efficiency yet struggle to capture the complex structural relationships inherent in SAR backscatter. Most existing SAR knowledge-distillation approaches rely on feature or logit matching, which enforces localized activation similarity while neglecting the geometric relationships among object representations. We propose a Structured Unified Relational knowledGE distillation framework for SAR Ship detection (SURGE) that transfers relational geometry from a powerful teacher detector to a compact student detector using a contrastive InfoNCE objective in a shared projection embedding space. To the best of our knowledge, this work presents the first transformer-based SAR ship detector knowledge distillation framework in SAR domain. The framework is architecture-agnostic in the sense that it provides a common region-level distillation interface for two-stage, one-stage and transformer-based detectors without modifying their deployed architectures. Experiments on the SSDD and HRSID benchmarks demonstrate that the proposed method yields substantial improvements for two-stage detectors, achieving up to 6.2 mAP and 8.0 AP75 gains over baseline student and even surpassing teacher performance