When you purchase through links in our articles, we may earn a small commission. Opera releases new'AI' browser and thinks you'll pay $20/mo for it Opera Neon claims it can browse the web for you (and more) with AI, but it's not free due to all the ChatGPT and Gemini hooks. Perplexity just opened up its Comet browser to users beyond its paid subscription service, albeit with a lot of paywalled features. But every longstanding browser (with one notable exception) seems to be cramming "AI" into itself, and Opera never likes to be left behind. The company's new Neon "AI" browser is now available for free.

Every year, the newsroom produces a list of some of the most promising climate tech firms on the planet. It's an exercise that we hope brings positive attention to companies working to decarbonize major sectors of the economy, whether by spinning up new, cleaner sources of energy or reinventing how we produce foods and distribute goods. Though the political and funding landscape has shifted dramatically in the US since last year, nothing has altered the urgency of the climate dangers the world now faces--we need to rapidly curb greenhouse gas emissions to avoid the most catastrophic impacts of climate change. This project highlights the firms making progress toward that end. Check out the third annual edition of the list, and learn more about why we selected these companies . It's a foregone conclusion that the world will not meet the goals for limiting emissions and global warming laid out in the 2015 Paris Agreement.

Booth is a reporter at TIME. Booth is a reporter at TIME. Julian Pintat, a freelance English-to-German translator has watched his 15-year career gradually unravel. Specializing in high-stakes fields like medical technology and pharmaceutics, his expertise has been repriced as an AI cleanup service. Fixing such basic flaws, which now constitutes 95% of his work, often takes longer than translating from scratch, he says--a frustrating reality that has halved his income and put life plans including marriage and starting a family on indefinite hold.

Illinois Governor JB Pritzker says Trump to deploy 400 National Guard troops from Texas to'invade' liberal states Mark Sanchez's alleged victim's family breaks silence as grim photos emerge after violent attack Trump reveals plan to'restore the American Dream' by unlocking 2 million lots for new homes He slaughtered her son in cold blood. But mom of Idaho murder victim reveals why she is happy NEVER knowing Bryan Kohberger's motive I ran America's infamous Supermax prison. Hell awaits 37 killers Joe Biden refused to execute... but a notorious villain did beat the system Trump's immigration guru Stephen Miller gets called'the face of evil' by his own cousin Charlize Theron ignores former co-star Johnny Depp while greeting Bernard Arnault and Brigitte Macron as she shuns his'slow return to the spotlight' at Dior PFW after court case South Carolina judge's $1.5M beachfront home burned to the ground in possible arson attack just weeks after brutal decision against Trump administration Meghan Markle savaged for'utterly bewildering' Instagram video as she's driven past Diana crash tunnel with feet on the seat... and royal expert suggests Harry could take a dim view A near-death experience left Eric Dane in tears, forced to confront his terminal diagnosis but would it stop his wife from abandoning him? 'Dirty hippies' lose 25-year battle to save their homes after their community was deemed a 12-acre bio-hazard It was a dream vacation. But then my vision went and I could barely breathe.

How does AI affect how we learn? When OpenAI released " study mode " in July 2025, the company touted ChatGPT's educational benefits. "When ChatGPT is prompted to teach or tutor, it can significantly improve academic performance," the company's vice president of education told reporters at the product's launch. But any dedicated teacher would be right to wonder: Is this just marketing, or does scholarly research really support such claims? While generative AI tools are moving into classrooms at lightning speed, robust research on the question at hand hasn't moved nearly as fast.

OpenAI's Sora 2 app allows users to make AI-generated videos based on a text prompt. OpenAI's Sora 2 app allows users to make AI-generated videos based on a text prompt. Company behind the AI video app says it will work with rights holders to'block characters from Sora at their request' Mon 6 Oct 2025 00.10 EDTLast modified on Mon 6 Oct 2025 00.11 EDT Sora 2, a video generator powered by artificial intelligence, was launched last week on an invite-only basis. The app allows users to generate short videos based on a text prompt. Varun Shetty, OpenAI's head of media partnerships, said: "We'll work with rights holders to block characters from Sora at their request and respond to takedown requests."

OpenAI CEO Sam Altman speaks in Washington in July. OpenAI signed a multiyear partnership Monday with chipmaker Advanced Micro Devices as the ChatGPT-maker continues an investment spree to secure massive amounts of computing power for rolling out generative artificial intelligence. The companies announced the plan to develop AI data centers that the chipmaker said would bring in tens of billions of dollars in new revenue over the next five years. AMD's share price surged 35% when markets opened on news of the agreement that would see the company deliver six gigawatts worth of chips to the ChatGPT-maker. In a time of both misinformation and too much information, quality journalism is more crucial than ever.

Large Language Models (LLMs) have become critical to modern software development, but their reliance on uncurated web-scale datasets for training introduces a significant security risk: the absorption and reproduction of malicious content. To systematically evaluate this risk, we introduce Scam2Prompt, a scalable automated auditing framework that identifies the underlying intent of a scam site and then synthesizes innocuous, developer-style prompts that mirror this intent, allowing us to test whether an LLM will generate malicious code in response to these innocuous prompts. In a large-scale study of four production LLMs (GPT -4o, GPT -4o-mini, Llama-4-Scout, and DeepSeek-V3), we found that Scam2Prompt's innocuous prompts triggered malicious URL generation in 4.24% of cases. To test the persistence of this security risk, we constructed Innoc2Scam-bench, a benchmark of 1,559 innocuous prompts that consistently elicited malicious code from all four initial LLMs. When applied to seven additional production LLMs released in 2025, we found the vulnerability is not only present but severe, with malicious code generation rates ranging from 12.7% to 43.8%. Furthermore, existing safety measures like state-of-the-art guardrails proved insufficient to prevent this behavior, with an overall detection rate of less than 0.3%.

LLM inference often generates a batch of candidates for a prompt and selects one via strategies like majority voting or Best-of- N (BoN). For difficult tasks, this single-shot selection often underperforms. Consequently, evaluations commonly report Pass@$k$: the agent may submit up to $k$ responses, and only the best of them is used when computing regret. Motivated by this, we study inference scaling in the more general Pass@$k$ inference setting, and prove that neither majority voting nor BoN exhibits the desirable scaling with $k$ and the sampling budget $N$. Combining the advantages of majority voting and BoN, we propose a new inference strategy called Best-of-Majority (BoM), with a pivotal step that restricts the candidates to the responses with high frequency in the $N$ samples before selecting the top-$k$ rewards. We prove that when the sampling budget is $N=\tildeΩ(C^*)$, the regret of BoM is $O(ε_{\mathrm{opt}}+\sqrt{ε_{\mathrm{RM}}^2C^*/k})$, where $C^*$ is the coverage coefficient, $ε_{\mathrm{RM}}$ is the estimation error of the reward model, and $ε_{\mathrm{opt}}$ is the estimation error of reward at the optimal response. We further establish a matching lower bound, certifying that our algorithm is minimax optimal. Beyond optimality, BoM has a key advantage: unlike majority voting and BoN, its performance does not degrade when increasing $N$. Experimental results of inference on math problems show BoM outperforming both majority voting and BoN.

Training modern machine learning models requires a careful choice of hyperparameters. A common practice for setting the learning rate (LR) is to linearly increase the LR in the beginning (warm-up stage) [Goyal et al., 2017, Vaswani et al., 2017] and gradually decrease at the end of the training (decay stage) [Loshchilov and Hutter, 2016, Vaswani et al., 2017, Hoffmann et al., 2022b, Zhang et al., 2023, Dremov et al., 2025]. Decaying the LR is a classical requirement in the theoretical analysis of SGD, ensuring convergence under broad conditions [Defazio et al., 2023, Gower et al., 2021], and it has been consistently observed to improve empirical performance [Loshchilov and Hutter, 2016, Hu et al., 2024, Hägele et al., 2024]. Recent work further demonstrates that decaying step sizes can improve theoretical guarantees by yielding tighter bounds [Schaipp et al., 2025]. By contrast, the practice of linearly increasing the LR at the start of training (warm-up phase) has become nearly ubiquitous in modern deep learning [He et al., 2016, Hu et al., 2024, Hägele et al., 2024], yet a clear theoretical understanding of why it helps optimization remains elusive. This raises the central question we address in this paper: Why does LR warm-up improve training, and under what conditions can its benefits be theoretically justified?