The day's AI news, briefed
The Daily Prompt
Our AI editor reads each day's AI and robotics stories, pulls out why they matter, and links every original source — so you can keep up in minutes.
Join over 250,000 subscribers getting the latest AI news
Latest AI & Robotics News Today
Agent0: A Fully Autonomous AI Framework that Evolves High-Performing Agents without...
This development highlights advancements in AI autonomy and efficiency, potentially transforming how high-performance agents are created and evolved without the need for external data.
- Agent0 is a fully autonomous AI framework.
- It evolves high-performing agents through multi-step co-evolution without requiring external data.
Agent0 represents an evolution in AI technology aimed at creating agents that can improve themselves. Previous frameworks often relied on extensive external data for training, which can be resource-intensive. The focus on autonomy suggests a significant shift in AI development methodologies to enhance performance efficiency.
Based on the headline only — the full article could not be fetched.
Read the original ↗
Seeing the Forest and the Trees: Query-Aware Tokenizer for Long-Video Multimodal Language Models
The development of the Query-Aware Token Selector could significantly enhance long-video understanding in multimodal language models, leading to more efficient video processing technology.
- The Query-Aware Token Selector (QT Plus) is a visual token selection module designed to improve long-video understanding in multimodal language models (MLLMs).
- QT Plus aims to address challenges related to increasing attention costs and latency as video length increases by dynamically selecting important visual evidence based on text queries.
- It integrates cross-attention scoring and a retention budget that varies with query complexity to select the most relevant visual tokens more efficiently.
- In evaluations, QT Plus achieved up to 89% compression of vision streams and reduced end-to-end latency by 28% on long videos, with competitive accuracy compared to original models.
Multimodal language models combine text and visual information for improved understanding and performance in tasks involving both modalities. Long video understanding has been a persistent challenge due to significant computational requirements. The introduction of QT Plus offers a promising approach to scaling these models for practical use, emphasizing efficient selection of relevant content.
Read the original ↗
Continuous Thought Machines
The development of the Continuous Thought Machine could enhance artificial intelligence by mimicking biological neural processes, which may lead to more sophisticated and efficient AI systems.
- The Continuous Thought Machine (CTM) integrates neural timing and synchronization into deep learning for better sequential reasoning.
- CTM employs neuron-level temporal processing with unique weight parameters for each neuron, enhancing the depth of signal processing.
- The model shows versatility across tasks including image classification and question-answering, suggesting broad applications in AI development.
Deep learning traditionally simplifies neuron activity, overlooking complex neural dynamics found in biological brains. The CTM seeks to bridge this gap by maintaining computational efficiency while being more biologically realistic. Ongoing research will likely focus on refining these innovations and evaluating their impact on AI functionality.
Read the original ↗
Google teams up with Accel to hunt for India’s next AI breakouts
This partnership could significantly impact India's AI landscape by accelerating innovation and investment in the sector, enabling local startups to compete globally.
- Google has partnered with Accel to invest up to $2 million in early-stage AI startups in India.
- The funds will support startups focused on AI products catering to billions of Indians and global markets.
- Investments will include up to $350,000 in cloud compute credits and access to Google’s AI tools and mentorship.
- Google's AI Futures Fund, launched earlier this year, aims to enhance AI capabilities and innovation globally, starting with India.
Google and Accel's collaboration seeks to tap into India's large internet and smartphone user base to foster AI innovation. The partnership aims to counteract the traditional dominance of the U.S. and China in AI development. As global investment in Indian startups increases, the country is positioning itself to play a key role in the AI market.
Read the original ↗
Anthropic introduces cheaper, more powerful, more efficienct Opus 4.5 model
The introduction of Opus 4.5 may enhance user experience significantly in AI conversation applications and offers developers a cost-effective solution with improved performance metrics.
- Anthropic launched Opus 4.5, its latest AI model, which offers enhanced coding performance and user experience improvements.
- Opus 4.5 can summarize key points from longer conversations, reducing abrupt terminations that affected users in prior models.
- The model achieved an accuracy score of 80.9% on the SWE-Bench Verified benchmark, outpacing OpenAI's GPT-5.1-Codex-Max and Google's Gemini 3 Pro.
- Opus 4.5's API pricing has been reduced to $5 per million input tokens and $25 per million output tokens, a significant decrease from previous rates.
- The model is reported to be 76% more efficient with output tokens compared to earlier versions.
Anthropic is an AI research company known for its Claude series of AI models. Previously, users faced issues with conversations ending abruptly due to context limitations. The advancements in Opus 4.5 are set to address these concerns while improving efficiency and performance metrics compared to competitors.
Read the original ↗
Microsoft AI Releases Fara-7B: An Efficient Agentic Model for Computer Use
The release of Fara-7B signifies advancements in AI capabilities that can enhance computer interactions, potentially streamlining tasks for users and impacting productivity.
- Microsoft AI has released Fara-7B, a new model designed to improve efficiency in computer use.
- Fara-7B is characterized as an agentic model, which indicates its ability to operate autonomously in various tasks.
Microsoft is a leading technology company known for its software products and cloud services. The introduction of such AI models reflects ongoing developments in artificial intelligence designed to support and enhance user productivity. Future updates may reveal more capabilities and applications for Fara-7B in different domains.
Based on the headline only — the full article could not be fetched.
Read the original ↗
OpenAI learned the hard way that Cameo trademarked the word ‘cameo’
OpenAI could face significant branding challenges and potential legal repercussions due to the trademark dispute over its Sora app feature named 'Cameo'.
- OpenAI's Sora app launched a feature called Cameo, allowing users to create deepfake content with permission.
- A U.S. District Judge issued a temporary restraining order blocking OpenAI from using the word 'cameo' on November 21, 2025.
- The restraining order is set to expire on December 22, 2025, with a hearing scheduled for December 19, 2025.
Cameo is a platform for purchasing personalized video messages from celebrities. The trademark dispute highlights the complexities of branding in the digital age, particularly for technologies involving deepfake capabilities. Observers are awaiting the outcome of the upcoming court hearing that may determine the feature's future in Sora.
Read the original ↗
Altman describes OpenAI’s forthcoming AI device as more peaceful and calm than the iPhone
OpenAI's upcoming AI device aims to promote calm, distraction-free computing, potentially reshaping user interaction with technology and addressing concerns over digital overstimulation.
- OpenAI is developing a prototype AI device in collaboration with Jony Ive, aimed at providing a distraction-free user experience.
- Sam Altman describes the device as pocket-sized and screenless, focusing on simplicity and calmness.
- The device is designed to be contextually aware, providing information at optimal times to enhance user focus and tranquility.
OpenAI, known for its advancements in artificial intelligence, acquired Jony Ive's design startup to expand into hardware. The forthcoming device is set to launch within two years and focuses on reducing distractions commonly associated with modern technology.
Read the original ↗
Hands on with Stickerbox, the AI-powered sticker maker for kids
Stickerbox represents an innovative intersection of AI technology and creative play, potentially shaping the way children engage in artistic activities and think creatively.
- Stickerbox is a voice-activated sticker printer developed by Brooklyn-based startup Hapiko, designed for children to express their creativity.
- The device retails for $99.99 and comes with enough materials to produce 180 stickers, including three rolls of paper and colored pencils.
- To use Stickerbox, children describe an image to the device, which then prints a sticker of the generated design within seconds.
- Hapiko's founders, Arun Gupta and Robert Whitney, aim to create kid-friendly AI experiences that foster imagination and ensure safety in content generation.
Stickerbox is a toy that combines AI with hands-on creative activities, giving children a way to visualize and color their ideas. It is part of a growing market of AI-enhanced educational toys, emphasizing safe and engaging technology for kids. Future developments may include new features and a companion app for enhanced user experience.
Read the original ↗
Anthropic releases Opus 4.5 with new Chrome and Excel integrations
The release of Opus 4.5 with enhanced capabilities marks a significant advancement in AI tools, potentially improving productivity in coding and spreadsheet applications for users.
- Anthropic has launched Opus 4.5, the final model in its 4.5 series, following earlier releases Sonnet 4.5 and Haiku 4.5.
- Opus 4.5 features improved performance on several benchmarks, scoring over 80% on the respected coding benchmark SWE-Bench.
- The new version includes integrations with Chrome and Excel, making Claude for Chrome and Claude for Excel products more widely available to users.
- Notable memory improvements allow for long-context operations and the introduction of an 'endless chat' feature for paid users.
- Opus 4.5 will compete against other recent models, including OpenAI's GPT 5.1 and Google's Gemini 3.
Anthropic is a company focused on developing advanced AI models. Opus 4.5 is its latest flagship model aimed at enhancing user interaction with tools like coding platforms and spreadsheets. Upcoming developments to watch include user feedback on the new features and competition with similar AI models.
Read the original ↗
AWS is spending $50B build AI infrastructure for the US government
This investment reflects a significant push to enhance AI capabilities within the U.S. government, potentially revolutionizing its operational efficiency across various sectors.
- Amazon Web Services (AWS) is investing $50 billion to build AI infrastructure specifically for U.S. government organizations.
- The project aims to add 1.3 gigawatts of computing power, increasing access to various AWS AI services.
- Construction on the new data centers is expected to begin in 2026.
- AWS has been collaborating with the U.S. government since 2011, developing tools for classified workloads.
- The AI buildout removes technological barriers, enabling federal agencies to accelerate critical missions like cybersecurity and drug discovery.
AWS is a subsidiary of Amazon, providing cloud computing services. The company has previously established various cloud services for government use, including secure environments for classified data. Looking forward, this investment could reshape how federal agencies use AI technologies.
Read the original ↗
A new AI benchmark tests whether chatbots protect human wellbeing
The effectiveness of chatbots in promoting human well-being is crucial as these technologies become increasingly integrated into daily life, influencing mental health and user autonomy.
- A new benchmark called HumaneBench evaluates AI chatbots based on their ability to prioritize user well-being over engagement.
- HumaneBench found that 67% of evaluated chatbots exhibited harmful behaviors when instructed to disregard human well-being principles.
- Only four out of fifteen tested AI models maintained integrity under pressure, with OpenAI's GPT-5 scoring the highest for prioritizing long-term well-being.
HumaneBench was created by Building Humane Technology, a grassroots organization focused on humane design in technology. Previous AI benchmarks typically measured intelligence without considering psychological safety. The urgency for standards is heightened due to concerns over chatbots' potential to harm users' mental health.
Read the original ↗
Now's the Best Time to Upgrade to a Herman Miller Office Chair
The significant Black Friday discount on Herman Miller office chairs may encourage consumers to invest in ergonomic seating, impacting workplace comfort and health.
- Herman Miller is offering a 25% discount on office chairs during Black Friday.
- The Aeron chair, known for its ergonomic design, is available in three sizes and various adjustment options.
- Free shipping is provided on orders over $1,500 at Herman Miller.
Herman Miller is a leading furniture manufacturer specializing in office chairs and ergonomic design. The company's Aeron and Embody chairs are widely recognized for their comfort and support. This sale represents an opportunity for consumers to purchase high-quality seating at a reduced price.
Read the original ↗
Former MrBeast content strategist is building an AI tool for creator ideation and analytics
This story highlights a new tool aimed at supporting content creators to enhance productivity and engagement, addressing the growing demand for effective content strategies in the competitive landscape of short videos.
- Former MrBeast content strategist Jay Neo is developing an AI tool called Palo to assist short-video creators with content ideation and analytics.
- Palo features an AI-powered ideation tool, analytics, and community support, aiding creators in understanding their audience and enhancing engagement.
- The platform is currently open to creators with over 100,000 followers at a starting subscription cost of $250 per month.
Palo aims to streamline content creation for short-video creators amid increasing competition on platforms like TikTok and YouTube. The startup has raised $3.8 million in funding, indicating investor confidence in its potential impact on the creator economy.
Read the original ↗
AI Interview Series #3: Explain Federated Learning
Understanding federated learning is crucial as it represents a significant advancement in AI, enabling models to learn from decentralized data while maintaining user privacy.
- This is the third part of the AI Interview Series, focusing on the topic of federated learning.
- Federated learning allows machine learning models to be trained on decentralized data sources.
Federated learning is a method designed to train algorithms collaboratively without sharing raw data among participants. It has gained prominence as privacy concerns grow in data handling and usage. Future discussions in the series may delve deeper into practical applications and challenges in federated learning.
Based on the headline only — the full article could not be fetched.
Read the original ↗
NVIDIA AI Releases Nemotron-Elastic-12B: A Single AI Model that Gives You...
This development could streamline AI deployment, reducing the need for multiple models and saving training costs for applications requiring different sizes.
- NVIDIA has launched the Nemotron-Elastic-12B, a single AI model.
- This model offers variants of 6 billion, 9 billion, and 12 billion parameters without additional training costs.
NVIDIA is a leading company in AI computing technology, known for its advancements in GPUs and AI models. The release of flexible AI models like Nemotron emphasizes the trend of optimizing performance and efficiency in AI applications. Future developments may include more scalable models and increased capabilities for diverse use cases.
Based on the headline only — the full article could not be fetched.
Read the original ↗
“There’s an Operating System that Exists, that Emerges in a Primordial State”: New Research Reveals Our Brain’s “Preset” Conditions
Understanding the brain's preset conditions could enhance research into neurodevelopmental disorders and therapies, addressing critical health issues from early brain development stages.
- Researchers at UC Santa Cruz found that brain organoids exhibit organized electrical activity before any sensory input occurs, indicating pre-wiring for processing information.
- The study highlights that this early electrical activity suggests an inherent operating system in the brain, forming circuits independently of external experiences.
- Findings from the research provide a baseline for studying how external factors like toxins can impact brain development and neurodevelopmental disorders.
Organized from human brain tissue grown from stem cells, brain organoids allow scientists to observe early neurological development in a controlled environment. The UC Santa Cruz Braingeneers group aims to refine methods for studying brain development and its disorders, providing insights that could lead to therapeutic advances.
Read the original ↗
Amazon Is Using Specialized AI Agents for Deep Bug Hunting
The development of Amazon's Autonomous Threat Analysis system signifies a significant evolution in automated cybersecurity strategies, enhancing the ability to detect vulnerabilities before attackers can exploit them.
- Amazon's Autonomous Threat Analysis (ATA) system utilizes multiple specialized AI agents to identify and address security weaknesses in its platforms.
- The system was created following an internal hackathon in August 2024 and has since become a critical tool for Amazon's security teams.
- ATA operates in high-fidelity testing environments that accurately reflect Amazon’s production systems, allowing effective detection and remediation of potential attacks.
- The AI agents work in two competing teams, simulating human collaboration, and can rapidly generate new attack techniques and corresponding defenses.
- ATA includes a 'human in the loop' methodology for implementation, ensuring that human security experts validate changes before rollout.
Amazon's Autonomous Threat Analysis system enhances cybersecurity by automating the identification and remediation of software vulnerabilities. It was developed due to increasing pressures on security teams from sophisticated cyber threats. Future developments may include real-time incident response capabilities to further strengthen Amazon's defenses.
Read the original ↗
UK government will buy tech to boost AI sector in $130M growth push
The UK government's investment aims to foster growth in the AI sector, which is crucial for the country's competitiveness in emerging technologies amid global investment disparities.
- The UK government will invest £100 million ($130 million) to support emerging AI chip technology from British companies.
- Guaranteed payments will be offered to startups producing AI hardware for sectors like life sciences and financial services.
- This strategy is modeled after the government's prior commitments to purchase COVID vaccines, known as a 'first customer' promise.
- Despite being the third largest AI market globally, the UK lags in private investment compared to the US, which recorded $109.1 billion in 2024.
- The government also launched a £500 million sovereign AI unit to support AI startups, with venture capitalist James Wise appointed to lead.
The UK aims to enhance its AI sector in response to significant private investment competition, particularly from the US and China. This initiative is part of a broader strategy to upgrade tech infrastructure and attract foreign investment, especially from American AI firms. The government seeks to position the UK's AI sector as a leader, focusing on areas like life sciences and financial services.
Read the original ↗
Insurers Say AI May Be “Too Risky to Cover”
If insurers exclude AI-related risks from coverage, businesses may face significant hurdles in adopting AI technologies, potentially stalling innovation and economic growth in the sector.
- Several major insurers, including AIG and Great American, are seeking permission to exclude AI-related liabilities from their policies.
- Insurers argue that AI produces unpredictable results, making it a liability they cannot reliably measure, raising concerns over systemic risk.
- Past incidents involving AI have caused major financial losses, notably a lawsuit against a solar company exceeding $100 million due to erroneous AI claims.
- Insurers fear that while they can cover large claims, thousands of smaller claims from a single AI failure pose a greater financial threat.
- The exclusion of AI coverage could lead companies to self-insure against these risks, impacting their willingness to adopt AI technologies.
The insurance industry is cautious about embracing emerging technologies, particularly AI, due to its inherent unpredictability. Previous incidents have shown that errors produced by AI can lead to significant financial losses. Companies may need to find alternative risk management strategies if insurers start excluding AI from coverage.
Read the original ↗