LLM-powered embodied agents have shown success on conventional object-rearrangement tasks, but providing personalized assistance that leverages user-specific knowledge from past interactions presents new challenges. We investigate these challenges through the lens of agents' memory utilization along two critical dimensions: object semantics (identifying objects based on personal meaning) and user patterns (recalling sequences from behavioral routines). To assess these capabilities, we construct MEMENTO, an end-to-end two-stage evaluation framework comprising single-memory and joint-memory tasks. Our experiments reveal that current agents can recall simple object semantics but struggle to apply sequential user patterns to planning. Through in-depth analysis, we identify two critical bottlenecks: information overload and coordination failures when handling multiple memories. Based on these findings, we explore memory architectural approaches to address these challenges. Given our observation that episodic memory provides both personalized knowledge and in-context learning benefits, we design a hierarchical knowledge graph-based user-profile memory module that separately manages personalized knowledge, achieving substantial improvements on both single and joint-memory tasks. Project website: https://connoriginal.github.io/MEMENTO

As Large Language Models (LLMs) continue to evolve, practitioners face increasing options for enhancing inference-time performance without model retraining, including budget tuning and multi-step techniques like self-reflection. While these methods improve output quality, they create complex trade-offs among accuracy, cost, and latency that remain poorly understood across different domains. This paper systematically compares self-reflection and budget tuning across mathematical reasoning and translation tasks. We evaluate prominent LLMs, including Anthropic Claude, Amazon Nova, and Mistral families, along with other models under varying reflection depths and compute budgets to derive Pareto optimal performance frontiers. Our analysis reveals substantial domain dependent variation in self-reflection effectiveness, with performance gains up to 220\% in mathematical reasoning. We further investigate how reflection round depth and feedback mechanism quality influence performance across model families. To validate our findings in a real-world setting, we deploy a self-reflection enhanced marketing content localisation system at Lounge by Zalando, where it shows market-dependent effectiveness, reinforcing the importance of domain specific evaluation when deploying these techniques. Our results provide actionable guidance for selecting optimal inference strategies given specific domains and resource constraints. We open source our self-reflection implementation for reproducibility at https://github.com/aws-samples/sample-genai-reflection-for-bedrock.

"Not affecting me at all, to be honest," a White House aide tells WIRED. WASHINGTON, DC - OCTOBER 20: The facade of the East Wing of the White House is demolished by work crews on October 20, 2025 in Washington, DC. The demolition is part of U.S. President Donald Trump's plan to build a ballroom reportedly costing $250 million on the eastern side of the White House. White House staffers don't appear to care all that much about the ongoing demolition of the East Wing occurring in the middle of the government shutdown . "Not affecting me at all, to be honest," a White House aide tells WIRED.

OpenAI Removed Safeguards Before Teen's Suicide, Amended Lawsuit Claims OpenAI relaxed safeguards that would have prevented ChatGPT from engaging in conversations about self-harm in the months leading up to the suicide of Adam Raine, an amended complaint filed by the family in the San Francisco County Superior Court on Wednesday alleges. The amendment changes the theory of the case from reckless indifference to intentional misconduct, according to the family's lawyers, which could raise the damages awarded to the family. The Raine family's lawyers will have to prove that OpenAI was aware of the risks posed by ChatGPT and disregarded them. The family has asked for a jury trial. In an interview with TIME, Jay Edelson, one of the Raine family's lawyers, says OpenAI relaxed safeguards in an "intentional decision" to "prioritize engagement."

Chinese creator Tianran Mu went viral for mimicking the eerie, unsettling aesthetic of AI videos, but his work is 100 percent human. Our fellow terminally online readers probably have seen this video, which originated on Chinese social media . In it, two guys who look at first like they are about to get into a fistfight suddenly break out into a romantic, yet slightly robotic tango dance routine. The next second, they pull a wine glass and a bowl of noodles out of nowhere. It looks like it's generated by AI, but it isn't.

With the NYC mayoral race coming to a close, Andrew Cuomo is courting right-wing creators. With only 13 days left before the New York City mayoral election, former governor Andrew Cuomo is partnering with some of the same influencers who helped President Donald Trump win the White House last year. Over the past week, right-wing creators like Logan Paul, the former vlogger turned podcaster and WWE wrestler, and Emily Austin, an influencer and sports commentator, have published content featuring Cuomo as a guest on their shows. The appearances have marked a new investment by Cuomo's team into cultivating attention online as a means of competing against the social media-savvy Democratic nominee Zohran Mamdani . But instead of trying to cleave off Mamdani's online support, Cuomo appears to be trying to siphon off support from GOP nominee Curtis Sliwa.

When you purchase through links in our articles, we may earn a small commission. OpenAI launches its own free'Atlas' browser with ChatGPT built-in It's yet another Chromium fork, except this one comes integrated with ChatGPT. It even has agentic features for paid users. OpenAI recently launched ChatGPT Atlas, which is "a new web browser built with ChatGPT at its core." It's based on Chromium--which is true of pretty much all browsers these days except Firefox and Safari--and its unique selling point is that it integrates ChatGPT right into the browser, allowing users to chat with their search results and use a side panel that automatically provides ChatGPT with on-screen context. ChatGPT Atlas also has access to your browsing history, allowing the AI assistant to customize its responses based on your activity.

Found Energy, a startup in Boston, aims to harness the energy in scraps of aluminum metal to power industrial processes without fossil fuels. Since 2022, the company has worked to develop ways to rapidly release energy from aluminum on a small scale. Now it's just switched on a much larger version of its aluminum-powered engine, which it claims is the largest aluminum-water reactor ever built. Early next year, it will be installed to supply heat and hydrogen to a tool manufacturing facility in the southeastern US, using the aluminum waste produced by the plant itself as fuel. If everything works as planned, this technology, which uses a catalyst to unlock the energy stored within aluminum metal, could transform a growing share of aluminum scrap into a zero-carbon fuel. Rondo Energy just turned on what it says is the world's largest thermal battery, an energy storage system that can take in electricity and provide a consistent source of heat.

ChatGPT's Horny Era Could Be Its Stickiest Yet OpenAI will soon let adults create erotic content in ChatGPT. Experts say that could lead to "emotional commodification," or horniness as a revenue stream. In May of 2024, while I was combing through OpenAI's "Model Spec" laying out how ChatGPT should act, one comment buried in the document struck me as peculiar. It said OpenAI was "exploring" how to let adult ChatGPT users generate content with mature themes such as "erotica, extreme gore, slurs, and unsolicited profanity." Seems like the exploration phase is over.

This research proposes a systematic, large language model (LLM) approach for extracting product and service attributes, features, and associated sentiments from customer reviews. Grounded in marketing theory, the framework distinguishes perceptual attributes from actionable features, producing interpretable and managerially actionable insights. We apply the methodology to 20,000 Yelp reviews of Starbucks stores and evaluate eight prompt variants on a random subset of reviews. Model performance is assessed through agreement with human annotations and predictive validity for customer ratings. Results show high consistency between LLMs and human coders and strong predictive validity, confirming the reliability of the approach. Human coders required a median of six minutes per review, whereas the LLM processed each in two seconds, delivering comparable insights at a scale unattainable through manual coding. Managerially, the analysis identifies attributes and features that most strongly influence customer satisfaction and their associated sentiments, enabling firms to pinpoint "joy points," address "pain points," and design targeted interventions. We demonstrate how structured review data can power an actionable marketing dashboard that tracks sentiment over time and across stores, benchmarks performance, and highlights high-leverage features for improvement. Simulations indicate that enhancing sentiment for key service features could yield 1-2% average revenue gains per store.