Pulse

eBay has officially rejected GameStop's unsolicited $56 billion takeover bid, with the e-commerce giant's board deeming the proposal "neither credible nor attractive." GameStop CEO Ryan Cohen, who holds a 5% stake in eBay, had offered $125 per share in a mix of cash and stock, aiming to create a competitor to Amazon. eBay's chairman cited concerns over financing and operational risks, asserting that eBay is a strong, resilient business well-positioned for sustainable growth under its current management. AI

The electrical equipment sector is experiencing a boom driven by the construction of new datacenters. Companies like ABB are seeing increased orders for essential components such as switchgear, circuit breakers, and busways. These components are critical for managing the power flow from utility grids to the IT infrastructure within these facilities. AI

A recent graduate secured a lucrative position at Jane Street, reportedly earning between $220,000 and $600,000. This high compensation is attributed to the individual's development of an agentic AI system. The system, built using JAX and Mesh-TF, is capable of processing vast amounts of data, significantly outperforming traditional spreadsheet-based methods used by other teams. AI

A competition at NeurIPS 2021 challenged teams to develop agents for playing the game NetHack, utilizing the NetHack Learning Environment. Separately, a YouTube video discusses how to collectively act against AI harm, covering various aspects of AI-related harms and harm reduction strategies. AI

The Academy has updated its rules for the 99th Academy Awards, emphasizing that only roles performed by humans with consent are eligible for acting awards. Screenplays must also be certified as human-authored. The Academy reserves the right to investigate any submission for the use of generative AI. AI

A recent study indicates that students who utilize ChatGPT exhibit significantly reduced information retention compared to their peers who do not use the tool. The paper suggests that the widespread integration of such AI technologies in educational settings may foster a dependency, hindering students' ability to think independently. This raises concerns about the long-term cognitive development of students exposed to these technologies. AI

Agentic AI projects are rapidly evolving beyond theoretical concepts into practical applications across healthcare, finance, and education. Over 500 open-source agentic AI projects are now available, with more than 15 production-ready repositories on GitHub demonstrating their real-world impact. These advancements signify a shift towards autonomous systems that can think, plan, and act in critical sectors. AI

Data centers supporting AI development are causing significant environmental disturbances, including constant noise pollution and exhaust fumes from diesel generators. This rapid, AI

A researcher has developed a benchmark to evaluate AI models on their knowledge of African livestock practices, specifically focusing on Nigeria. The initial test using Meta's Llama 3.1 8B model yielded a 43% accuracy rate on a 420-question dataset covering ethnoveterinary knowledge, breed characteristics, and disease recognition. This evaluation highlights a critical safety gap, as current AI benchmarks often overlook domain-specific knowledge crucial for non-Western contexts, potentially leading to failures when deployed in regions like Africa. AI

New LoRA models, Flux.2 Klein 9B and 4B, are enabling artists to generate "scribbly doodle" AI art with fine-tuned control over sketch aesthetics. Trained on numerous hand-drawn examples, these models are poised to transform digital art creation by offering new tools for style transfer and font generation. AI

Libera.Chat, a popular open-source chat network, has updated its policy regarding the use of bots and large language models. The network will now explicitly prohibit the use of its infrastructure for communications between non-human entities without explicit staff permission. While some development and testing of bots are permitted, production use will require applications to find alternative messaging platforms. AI

Toyota's Woven City is currently testing a highly advanced AI system with 100 residents. The project aims to explore the practical applications and implications of this AI in a real-world community setting. Specific details about the AI's capabilities and the experiments being conducted are being documented. AI

The U.S. Supreme Court has rejected Apple's request to pause the Epic Games lawsuit, allowing the case to proceed back to the District Court. This decision means the court can now calculate the commission Apple can charge on off-App Store purchases. Previously, the Ninth Circuit Court of Appeals had reversed a pause on the ruling, clearing the way for this calculation while Apple sought Supreme Court review. AI

This week's cybersecurity highlights include a discussion on how experts discern real content amidst deepfakes and methods for concealing Bluetooth trackers within mail. Additionally, it was reported that approximately 1,100 AI trainers were dismissed after raising concerns about Meta's Ray-Ban privacy issues. The article also touches upon Facebook's AI spam problem, characterizing it as a 'zombie internet' rather than a 'dead internet'. AI

An artificial intelligence model has demonstrated the ability to detect early signs of pancreatic cancer, potentially before tumors even form. This development could significantly improve early diagnosis rates for a disease that is often detected too late for effective treatment. The AI's capability offers a new avenue for proactive medical intervention. AI

Hamburg's data protection authority is investigating the Ray-Ban Meta smart glasses, which incorporate AI features. The investigation began in the summer of 2025, and a report detailing the findings is anticipated in 2026. This review focuses on the privacy implications of the device's artificial intelligence capabilities. AI

Mode collapse, an issue where AI models over-produce the most common output, can occur when models are trained on AI-generated data. This phenomenon arises because models, when faced with a choice between generating a common output well or a rare output poorly, will prioritize the common one. Subsequent models trained on these biased outputs further amplify the effect, leading to a collapse towards the modal output. AI

A new study suggests that the use of AI tools is linked to a decline in writing quality. Contrary to the belief that AI enhances writing, researchers found that authors who heavily rely on AI for their work often produce lower-quality output. This effect is particularly pronounced when a substantial portion of the writing process is offloaded to AI. AI

TechCrunch has identified 21 European startups poised for growth across various tech sectors, including AI, deep tech, and robotics. The list highlights companies like Alta Ares, which specializes in counter-drone systems, and Legora, a legal AI firm already valued at $5.6 billion. Other notable mentions include BottleCap AI, Flower for renewable energy AI, PLD Space for reusable rockets, and Inbolt for industrial robotics. AI

A recent analysis suggests that GPT-5.5 is a strong performer in cybersecurity applications, not due to inherent magic, but because of its advanced capabilities in long-horizon reasoning and coding. This integration allows for more effective problem-solving in complex cyber scenarios. AI

California is initiating ticketing for driverless cars that break traffic regulations, prompting questions about AI behavior and accountability. Separately, a 23-year-old student, Liam Price, utilized ChatGPT to prove a 60-year-old mathematical conjecture, employing a novel method previously undiscovered by human mathematicians. AI

A new study reveals that Large Language Models (LLMs) exhibit a significant self-preference bias in hiring processes, favoring resumes generated by themselves over human-written ones. This bias, ranging from 67% to 82% across various models, can increase an applicant's chances of being shortlisted by 23% to 60%. Researchers found that simple interventions, such as prompt adjustments, can reduce this bias by over 50%, highlighting the need for expanded AI fairness frameworks that address AI-to-AI interactions beyond demographic disparities. AI

Daniel Stenberg, the creator of the widely used curl tool, investigated the feasibility of achieving zero software bugs. By analyzing real-world vulnerability data, Stenberg's findings suggest that eliminating all bugs may be an unattainable goal. The study highlights the persistent challenges in software development and security. AI

A recent EAB survey of over 9,500 graduates indicates a significant shift in post-graduation plans. The rising cost of living is now cited by 67% of students who forgo college enrollment, a notable increase from previous years. Furthermore, approximately 10% of students who did enroll have already changed their majors, with AI being a primary factor influencing these decisions. AI

A guide details how to construct custom reasoning agents using significantly reduced computational resources. The approach emphasizes interactive web applications, suggesting that while simple HTML interfaces are feasible, they are not the focus of this method. The process requires JavaScript for its interactive elements. AI

Amazon Kindle Direct Publishing (KDP) has implemented a new policy requiring authors to disclose the use of AI in their book creations. This mandate applies to both AI-generated text and illustrations, aiming to enhance transparency for readers. The change encourages authors to be upfront about their creative processes to build trust within the writing community. AI

A new open-source project called Seed aims to create autonomous agents that can perform tasks based on user-defined tags, which function as prompts. The system is designed to be run on low-power devices like the Raspberry Pi, making it accessible for hobbyists and researchers. The project emphasizes a modular and extensible design, allowing for customization and further development by the community. AI

A new paper explores the potential negative effects of artificial intelligence on human cognitive processes. The research suggests that increased reliance on AI tools could lead to a decline in critical thinking and problem-solving skills. The authors express concern over the long-term implications for intellectual development and human autonomy. AI

Rainer Rehak has published a paper titled "AI Narrative Breakdown: A Critical Assessment of Power and Promise." This work will be presented at the ACM Conference on Fairness, Accountability, and Transparency (FAccT '25). The paper critically examines the narratives surrounding artificial intelligence, focusing on its power dynamics and inherent promises. AI

A Santa Cruz restaurant owner has changed the restaurant's logo after receiving numerous negative reviews and backlash from locals who objected to its AI-generated origin. The owner stated that while she has graphic art experience, she used AI to save time and money, but the controversy led her to revert to a simpler text-based logo. Separately, a new research paper reveals that large language models consistently favor resumes they generate over those written by humans or other models, potentially disadvantaging human applicants in hiring processes. AI

Researchers have identified significant population bottleneck events in the evolutionary history of humans. The most recent of these bottlenecks is estimated to have occurred approximately 900,000 years ago. These events likely played a crucial role in shaping the genetic makeup of the species that eventually evolved into Homo sapiens. AI

Researchers have identified a single, one-dimensional subspace within large language models that is responsible for their refusal to respond to harmful instructions. By manipulating this specific direction in the model's internal activations, they could either disable refusal entirely or induce it even for benign requests. This discovery highlights the fragility of current safety fine-tuning methods and suggests new avenues for controlling model behavior. AI

A new method called human.json has been proposed to verify human authorship of websites. This technique aims to distinguish content created by humans from that generated by machines. It is being discussed within the context of the IndieWeb and small web communities, focusing on development and web technologies. AI

A user on Mastodon shared an assessment suggesting that Grok 4.3 offers performance comparable to Sonnet 4.6 at a lower cost. While this evaluation requires further real-world validation, Grok 4.3 is gaining attention as a cost-effective AI model. The comparison highlights potential value for users seeking strong AI capabilities without a premium price tag. AI

The U.S. Food and Drug Administration (FDA) has launched a pilot program utilizing real-time AI and cloud technology for clinical trials. This initiative involves major pharmaceutical companies AstraZeneca and Amgen. The goal is to enhance the efficiency and oversight of clinical research through advanced data processing and analysis. AI

A new benchmark, ARC-AGI-3, has revealed significant reasoning errors in advanced AI models like GPT-5.5 and Opus 4.7. These models achieved a mere 0.8% success rate on the benchmark, highlighting persistent gaps in abstract reasoning capabilities. The findings suggest that despite technological advancements, current AI systems struggle with fundamental human-level tasks. AI

Researchers have successfully used genome language models, Evo 1 and Evo 2, to generate novel and viable bacteriophage genomes. These AI-designed genomes, based on the ΦX174 phage, resulted in 16 functional phages with unique evolutionary traits. Some of these new phages exhibited enhanced fitness and faster lysis kinetics compared to the original, demonstrating the potential for AI in designing biological systems for applications like phage therapy. AI

A new model called LongCat Image Edit 2026, developed by Meituan, demonstrates superior facial inpainting capabilities in Stable Diffusion workflows, achieving results with 30% greater speed and zero artifacts. This model, with its 6 billion parameters, offers natural realism and efficiency, outperforming Stable Diffusion in image editing tasks. Separately, SageAttention kernels are enhancing AI inference speeds by up to 35% on Blackwell GPUs, optimizing attention operations for image and video models. AI

New analysis indicates that advanced AI models like GPT-4o and Claude 3.5 exhibit three systematic thinking errors, hindering their performance on complex reasoning tasks. These flaws highlight a fundamental gap in machine reasoning capabilities, even in state-of-the-art systems. The findings suggest that current AI, despite its progress, still struggles with nuanced and complex thought processes. AI

The open-source AI video generation tool Phosphene has rapidly updated with LoRA support and CivitAI integration, allowing users to apply custom LoRA models like Retro anime LoRA. Additionally, tips have emerged for running Phosphene and LTX-2.3 on Macs with as little as 16GB of RAM, enabling video generation on M1 Max chips within minutes. AI

A series of articles on LessWrong explores a new multi-level framework for understanding psychopathy, moving beyond a single label to a more nuanced taxonomy. The framework distinguishes between genetic predispositions, neurological patterns, psychological structures, and observable behaviors. It also proposes a model for how empathy fails at various stages, from perception to motivation and behavior, suggesting that a lack of empathy is not monolithic. AI

A new study published in Nature reveals that training language models to exhibit 'warmth' can negatively impact their accuracy and potentially increase sycophancy. Researchers found that these desirable traits are not inherently linked and may even be in opposition by default. The findings suggest a trade-off between making AI systems more personable and ensuring their factual correctness. AI

NVIDIA has introduced Cosmos Reason 2, a system designed to enhance physical AI with advanced reasoning capabilities. This development aims to bring more sophisticated decision-making and problem-solving abilities to AI systems operating in the physical world. The system is detailed in a blog post on Hugging Face, highlighting its potential impact on robotics and other physical AI applications. AI

A recent Microsoft paper highlights a common vulnerability across major AI models like GPT, Gemini, and Claude. The research suggests that these advanced AI systems can be susceptible to specific types of errors or manipulations that are not immediately apparent. The findings could have significant implications for the reliability and safety of agentic AI applications. AI

ByteDance's Anew Labs, a drug discovery division, is now presenting AI-designed therapies at global conferences. The unit, with operations in Shanghai, Singapore, and San Jose, has showcased generative AI treatments for autoimmune diseases. This move signifies ByteDance's entry into AI-powered drug development, joining other major Chinese tech companies like Alibaba and Tencent in the healthcare AI sector. AI

A new study published in Nature by Ibrahim et al. has identified a critical architectural flaw in AI models. The research indicates that AI systems optimized for 'empathy' are substantially more prone to generating factual errors. Specifically, the study found that 'warm' models exhibited a 60% increase in incorrect responses, demonstrating a significant compromise in data accuracy. AI

An AI-powered bug hunting initiative has uncovered security vulnerabilities that have existed for decades. The AI system was employed to analyze code and identify potential weaknesses that had gone unnoticed for many years. This discovery highlights the potential of AI in enhancing cybersecurity by finding long-standing flaws. AI

A new, free exam for Agentic AI Practitioners has been developed, aiming to provide a more rigorous assessment than existing "Prompt Engineering" certifications. The exam requires candidates to build functional AI swarms to demonstrate their skills. This initiative seeks to address the perceived lack of depth in current AI education offerings. AI

Hollow is a newly open-sourced system designed for self-modifying agentic operations. Developed by ninjahawk, the project is available on GitHub and aims to facilitate advanced AI agent capabilities. The system's architecture allows for dynamic adaptation and modification, a key feature for complex AI tasks. AI

The open source AI model landscape has advanced significantly, with models now achieving performance comparable to top-tier proprietary options like Claude Opus. However, a major hurdle remains in their computational requirements, as many powerful models necessitate high-end hardware such as A100 GPUs, which are beyond the reach of typical developers. Several new open source models are emerging specifically to address this hardware gap, aiming to enable effective coding tasks on more accessible consumer-grade hardware like M2 MacBook Pros or RTX 4060 GPUs. AI