Hollywood icon Robert Redford has died at the age of 89, his publicist has said. The actor was known for his roles in blockbuster movies such as Butch Cassidy and the Sundance Kid, and went on to win an Oscar for Ordinary People when he turned his hand to directing. David Sillito looks back at his life. 'No doubt' Russia will cross Nato border if Ukraine falls, former US VP says Former US Vice-President Mike Pence calls for security guarantees in Ukraine to help deliver "just and lasting peace". The US House Oversight Committee has released new surveillance footage recorded hours before the convicted paedophile's death.

Contrary to popular belief, most property crimes--including burglaries and package theft--happen during the day, not under cover of darkness. But night still brings unique challenges: fewer people around, limited visibility, and more opportunity for intruders to move unseen. If your security camera can't see clearly after dark, you're missing protection when you might need it most. Night vision lets security cameras capture what the human eye can't see in the dark. Some cameras shine invisible infrared light to illuminate a scene, while others rely on light-sensitive sensors to amplify what little light is already there.

Dual lenses give Reolink's latest floodlight camera an incredibly wide field of view, while its bright and capable floodlights ensure the scene is effectively lit. No matter how wide a viewing angle a given fixed-lens security camera might have, it will invariably suffer from this flaw: It won't be able to see everything in front of it. And while some cameras try to solve this problem by employing a fish-eye lens to widen that viewing angle, the resulting image usually suffers from a degree of barrel distortion. Reolink's Elite Floodlight WiFi mitigates that problem via a clever hack. It uses two camera lenses instead of just one to deliver a combined (and relatively distortion-free) 180-degree field of view.

In this work, we explore the structure of the embedding space of a transformer model trained for playing a particular reinforcement learning (RL) game. Specifically, we investigate how a transformer-based Proximal Policy Optimization (PPO) model embeds visual inputs in a simple environment where an agent must collect "coins" while avoiding dynamic obstacles consisting of "spotlights. " By adapting Robinson et al. 's [15] study of the volume growth transform for LLMs to the RL setting, we find that the token embedding space for our visual coin collecting game is also not a manifold, and is better modeled as a stratified space, where local dimension can vary from point to point. We further strengthen Robinson's method by proving that fairly general volume growth curves can be realized by stratified spaces. Finally, we carry out an analysis that suggests that as an RL agent acts, its latent representation alternates between periods of low local dimension, while following a fixed sub-strategy, and bursts of high local dimension, where the agent achieves a sub-goal (e.g., collecting an object) or where the environmental complexity increases (e.g., more obstacles appear). Consequently, our work suggests that the distribution of dimensions in a stratified latent space may provide a new geometric indicator of complexity for RL games. 1

Home security cameras are easy to install, easy to use, and incredibly affordable these days. They let you keep tabs on your home–inside and out–from wherever you have internet access. They can respond to motion, creating a visual record of everything that's happened within their field of view, and high-end models can distinguish between people, pets, and even cars. The latest security cams require minimal installation and offer flexible setups and a range of security features--so many features, in fact, that it can be difficult to decide what you need and which model you should buy. Should you get an outdoor camera with a space-illuminating floodlight and a weatherized shell, an indoor cam with AI-powered pet detection and a motorized lens that patrols the room, or something in between? We've tested dozes of the top home security cameras available in real-world conditions, and we've distilled a list of the very best models. Whether you're looking to check on your kids and pets, or need a full-service sentinel with humans watching for intruders in real time, we'll help find the right security camera for your needs. Easy to set up yet packed with cutting-edge features, the Arlo Pro 5S 2K is the best choice for a battery-powered 2K security camera that can track moving subjects, see in the dark, and connect to speedy 5GHz Wi-Fi networks–everything you need to keep an eye on your homestead.

Electroencephalography (EEG) is a non-invasive technique widely used in brain-computer interfaces and clinical applications, yet existing EEG foundation models face limitations in modeling spatio-temporal brain dynamics and lack channel permutation equivariance, preventing robust generalization across diverse electrode configurations. To address these challenges, we propose DIVER-0, a novel EEG foundation model that demonstrates how full spatio-temporal attention-rather than segregated spatial or temporal processing-achieves superior performance when properly designed with Rotary Position Embedding (RoPE) for temporal relationships and binary attention biases for channel differentiation. We also introduce Sliding Temporal Conditional Positional Encoding (STCPE), which improves upon existing conditional positional encoding approaches by maintaining both temporal translation equivariance and channel permutation equivariance, enabling robust adaptation to arbitrary electrode configurations unseen during pretraining. Experimental results demonstrate that DIVER-0 achieves competitive performance with only 10% of pretraining data while maintaining consistent results across all channel permutation conditions, validating its effectiveness for cross-dataset generalization and establishing key design principles for handling the inherent heterogeneity of neural recording setups.

Breakthroughs, discoveries, and DIY tips sent every weekday. When you think of koalas (Phascolarctos cinereus), chances are that words like cute or fluffy come to mind--not cryptic or stealthy. And yet, researchers in southeastern Australia have just discovered hundreds of previously undocumented koalas living surprisingly close to the city of Newcastle. The team conducted what they claim to be the largest and most accurate peer-reviewed koala survey to date. As detailed in a study published this month in the journal Biological Conversation, the survey estimates that a population of 4,357 koalas across 166,302 acres of land is living in the state of New South Wales.

Human planning is efficient -- it frugally deploys limited cognitive resources to accomplish difficult tasks -- and flexible -- adapting to novel problems and environments. Computational approaches suggest that people construct simplified mental representations of their environment, balancing the complexity of a task representation with its utility. These models imply a nested optimisation in which planning shapes perception, and perception shapes planning -- but the perceptual and attentional mechanisms governing how this interaction unfolds remain unknown. Here, we harness virtual maze navigation to characterise how spatial attention controls which aspects of a task representation enter subjective awareness and are available for planning. We find that spatial proximity governs which aspects of a maze are available for planning, and that when task-relevant information follows natural (lateralised) contours of attention, people can more easily construct simplified and useful maze representations. This influence of attention varies considerably across individuals, explaining differences in people's task representations and behaviour. Inspired by the 'spotlight of attention' analogy, we incorporate the effects of visuospatial attention into existing computational accounts of value-guided construal. Together, our work bridges computational perspectives on perception and decision-making to better understand how individuals represent their environments in aid of planning.

Prompt engineering for large language models is challenging, as even small prompt perturbations or model changes can significantly impact the generated output texts. Existing evaluation methods of LLM outputs, either automated metrics or human evaluation, have limitations, such as providing limited insights or being labor-intensive. We propose Spotlight, a new approach that combines both automation and human analysis. Based on data mining techniques, we automatically distinguish between random (decoding) variations and systematic differences in language model outputs. This process provides token patterns that describe the systematic differences and guide the user in manually analyzing the effects of their prompts and changes in models efficiently. We create three benchmarks to quantitatively test the reliability of token pattern extraction methods and demonstrate that our approach provides new insights into established prompt data. From a human-centric perspective, through demonstration studies and a user study, we show that our token pattern approach helps users understand the systematic differences of language model outputs. We are further able to discover relevant differences caused by prompt and model changes (e.g. related to gender or culture), thus supporting the prompt engineering process and human-centric model behavior research.

Last night, X's "For You" algorithm offered me up what felt like a dispatch from an alternate universe. It was a post from Elon Musk, originally published hours earlier. "This is the first time humans have been in orbit around the poles of the Earth!" he wrote. Underneath his post was a video shared by SpaceX--footage of craggy ice caps, taken by the company's Dragon spacecraft during a private mission. Taken on its own, the video is genuinely captivating.