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
Anthropic launches Opus 5
The launch of Opus 5 indicates significant advancements in AI model performance while addressing privacy concerns, highlighting ongoing progress in competitive AI technology.
- Anthropic launched Opus 5, a new AI model, offering better performance at a lower cost compared to its predecessor, Fable 5.
- Opus 5 is designed to be less restrictive and is expected to engage safety classifiers 85% less often than Fable 5.
- The model retains meaningful safeguards, particularly around cybersecurity tasks, while allowing vulnerabilities in source code to be scanned.
- Anthropic introduced a feature called Automatic Fallbacks, which reroutes requests to a less powerful model to minimize disruptions.
- Opus 5 comes just two months after the release of Opus 4.8, reflecting Anthropic's rapid development cycle for AI models.
Anthropic's Opus series represents its ongoing effort to refine AI models particularly focused on performance and user accessibility. Previous versions have faced criticism over their restrictions and data retention policies. Future developments may focus on further refining these models and enhancing user experience with advanced features.
Read the original ↗Color Pass-Through via Camera-Display Coupling
Improving color reproduction for smartphone cameras and displays can enhance user experience and accuracy in digital photography, impacting both consumer and professional markets.
- A new framework called Color Pass-Through couples smartphone cameras and displays to minimize color reproduction errors.
- Current capture-to-display processes often lead to noticeable differences in color, brightness, and contrast.
- This end-to-end approach optimizes the color reproduction based on the observer, achieving over a 2-point improvement in user studies.
Color Pass-Through is designed to rectify the discrepancies between captured images and their displayed versions by optimizing the entire process rather than treating camera and display calibration separately. This approach is particularly relevant given the ongoing advancements in smartphone camera and display technologies and the demand for accurate image representation.
Read the original ↗
Team uses AlphaFold AI to redesign gene-editing proteins to make them safer
This research demonstrates a promising method for reducing off-target effects in gene editing, which may enhance the safety and efficacy of gene therapies.
- Researchers utilized AlphaFold AI to analyze gene-editing proteins and identify areas responsible for off-target effects.
- By modifying these proteins, they reduced off-target activity from 28 percent to 5 percent without compromising target efficacy.
- The approach may be applied to various gene-editing technologies, potentially improving safety across multiple therapies.
Gene editing utilizes systems to make precise changes to DNA, with CRISPR technology being among the most recognized. Previous gene-editing methods have faced challenges related to safety due to off-target effects, prompting ongoing research to improve specificity. The new method described could represent a significant advancement in developing safer gene therapies.
Read the original ↗
Anthropic’s Opus 5 Nears Frontier Intelligence at Opus Prices
Anthropic's Opus 5 provides a cost-effective alternative to higher-priced AI models, making advanced technology more accessible for everyday applications.
- Anthropic released Claude Opus 5 on July 24, 2026, positioned as a workhorse model at half the cost of its higher-end Fable 5.
- Opus 5 is available across Anthropic’s apps and cloud platforms, costing $5 per million input tokens and $25 per million output tokens.
- The model reportedly doubles the performance of Opus 4.8 at a lower price and approaches Fable 5's coding evaluation scores.
- Opus 5 is designed for efficiency, completing tasks in fewer steps than its predecessor, and offers a Fast mode at increased cost.
- The model is touted as the most aligned to Anthropic's safety criteria but shows limitations in certain specialized tasks compared to its Mythos counterpart.
Anthropic is a company specializing in AI development, known for its AI models including the Opus and Fable series. Prior to Opus 5, models like Sonnet 5 and Fable 5 filled gaps in its product lineup. The industry will closely monitor independent evaluations of Opus 5's reported capabilities.
Read the original ↗
Meta Turns Its AI Chatbot Into a Task-Running Assistant
Meta's new task-running capabilities for its AI assistant may reshape user engagement, marking a significant shift towards productivity in a competitive landscape dominated by established AI technology firms.
- Meta is rolling out an update to its AI assistant that enables task management features, such as access to users' calendars and automated daily briefings, starting July 24, 2026.
- Users can set recurring tasks, allowing Meta AI to manage them independently, condensing important information into summaries and flags for scheduling conflicts.
- The update shifts Meta AI's focus from entertainment and social connection to productivity, directly competing with established assistants from OpenAI and Google.
- The update uses Muse Spark 1.1, a model designed for task execution, and will initially be accessible in select markets within the Meta AI app and on meta.ai.
- Meta AI remains free while emphasizing its integration across major platforms like WhatsApp and Instagram, but faces challenges in user adoption and mindshare compared to competitors.
Meta's AI features have traditionally emphasized social interaction but are now pivoting to include productivity tools for users. This overhaul coincides with strategies of leading AI companies like OpenAI and Google, driving the need for Meta to provide compelling performance. Observers will be watching how well users adopt these new features over time as the rollout proceeds.
Read the original ↗
You Can Now Talk to ChatGPT While It Controls Your Computer
The integration of ChatGPT Voice into desktop applications marks a significant advancement in AI capabilities, allowing users to manage tasks hands-free, which could redefine user interactions with technology.
- OpenAI has launched ChatGPT Voice for its desktop app on macOS and Windows, enabling users to control their computers using voice commands.
- The desktop version can manage coding agents and coordinate tasks, distinguishing it from the mobile version which lacked such functionality.
- GPT-Live's full-duplex voice engine allows ChatGPT to listen and respond simultaneously, enhancing natural conversation flow during interactions.
- Users can access Codex features remotely via an iOS app, with Android support anticipated soon.
- The desktop app now includes a feature called Appshots, allowing ChatGPT to see the active window context on Mac devices.
ChatGPT is an AI-based conversational tool developed by OpenAI, which recently merged Codex, a programming AI, into its desktop app for improved functionalities. This new voice feature aims to streamline task management and coding interactions, signaling a shift towards more integrated AI tools in everyday computing.
Read the original ↗
Is A2E AI Really Uncensored? I Tested the Claims (Plus an Image-to-Video Review)
A2E AI's content policies provide a more flexible environment for creators, enhancing opportunities for artistic expression while still adhering to necessary legal guidelines.
- A2E AI offers a more permissive approach to content generation than mainstream platforms like Runway and Midjourney, allowing artistic nudity and experimental themes.
- Despite marketing itself as uncensored, A2E maintains some content restrictions, categorizing itself as low-filter rather than no-filter.
- The platform aggregates over 12 AI models for video and image generation, enabling users to switch models without multiple subscriptions.
- Image-to-video output requires 30 credits for 5-second clips, with an effective output rate often lower due to a 3- to 5-try failure rate for usability.
A2E AI (Avatars To Everyone) is a video and image generation platform that aggregates multiple AI models into one tool. This contrasts with competitors that impose stricter content moderation, offering artists and creators creative freedom without fully removing restrictions. Users should note the credit expiration policy and output quality compared to mainstream options.
Read the original ↗Silicon Valley Is Completely Divided Over Chinese AI
The division in Silicon Valley over Chinese AI reflects broader concerns about competition, innovation access, and potential security risks in the rapidly evolving AI landscape.
- A debate is ongoing in Silicon Valley over the impact of Chinese-made AI tools, especially open-weight systems.
- Concerns include the potential for Chinese firms to use techniques like distillation to replicate proprietary AI models.
- Over 200 startups signed a letter opposing government restrictions on open-weight AI, arguing it could harm U.S. innovation and competition.
Silicon Valley is facing a divide regarding the accessibility and regulation of open-weight AI models, which can be swiftly adapted for various uses. With concerns about intellectual property theft and safety, the debate includes major firms and newer startups advocating for different approaches to regulation and market access.
Read the original ↗
A New York High School Is Getting an AI-Powered Robot Assistant. Parents and Teachers Have Concerns
The introduction of AI-powered robots in schools raises important questions about educational methods and resources, particularly in underserved communities, impacting students' future learning experiences.
- The Salamanca City Central School District has contracted Realbotix to provide an AI-powered humanoid robot named Sally and an at-home AI tutor called Optio for $57,590.
- The AI system is set to assist students with personalized learning in select robotics classes starting this September, benefiting 500 students initially.
- Many parents and union members have expressed concerns about the appropriateness of AI technology in the classroom, advocating for more human interaction in education.
The Salamanca City Central School District is integrating AI technology into its curriculum, following a trend of increased interest in AI in education since the rise of tools like ChatGPT. There have been widespread debates about AI's role in classrooms, especially among teachers and parents in economically disadvantaged areas. Ongoing discussions examine the implications of this move and whether it adequately supports student needs and learning outcomes.
Read the original ↗My Life With Looki Is a Harbinger of Always-On Recording Glasses to Come
As technology evolves, the introduction of devices like the Looki raises critical discussions about privacy and the implications of always-on recording capabilities in everyday life.
- The Looki L1 is a pendant-shaped, AI-powered device designed for lifelogging, which means it captures audio and video of daily activities.
- Unlike smart glasses, the Looki L1 has a battery life of up to a day and uses proprietary AI technology for recording and memory generation.
- The device has generated mixed user experiences, with many clips deemed awkward or inappropriate for sharing due to unclear contexts.
- With a price tag of $250, the Looki L1 does not require a subscription, but its AI capabilities are limited to its own platform and cannot connect to other services.
- Potential advancements in AI wearables could significantly enhance battery life and on-device processing for future iterations of devices like the Looki.
Looki L1 is part of a growing trend in AI wearables aimed at capturing everyday moments without user intervention. This device follows a history of lifelogging technologies, but faces challenges regarding privacy and user comfort. Future developments in AI wearables may lead to improved functionality and user acceptance.
Read the original ↗
Bluesky’s AI assistant Attie expands into an open social research tool
The expansion of Attie into an open social research tool could enhance user engagement on Bluesky, addressing slowing growth and potential monetization challenges.
- Bluesky's AI assistant Attie has introduced a feature called Quests, enabling users to ask about news and trends across the Bluesky network and apps using the AT Protocol.
- User growth for Bluesky has slowed, with around 45.6 million registered accounts, prompting the development of new tools like Attie to boost engagement.
- Attie, launched in March, allows users to create custom social feeds and may start charging for its services in the future.
Bluesky is a social network designed to prioritize user control and decentralized interaction via the AT Protocol, a framework for building social media apps. Despite early rapid growth, Bluesky faces a need to innovate and monetize, thus the introduction of features like Attie's Quests. Future developments may include accessibility upgrades and premium services.
Read the original ↗
As US weighs response to Chinese AI, industry urges against broad open-weight restrictions
Broad restrictions on open-weight AI models could hinder innovation and competition in the AI industry amid rising geopolitical tensions and concerns over intellectual property theft.
- Several AI companies, including Nvidia and Microsoft, signed a letter opposing broad restrictions on open-weight AI models as the U.S. considers its response to Chinese AI.
- The letter aims to separate legitimate model-development techniques, like distillation, from unlawful intellectual property theft.
- Signatories assert that open-weight models are vital for competitive innovation and cybersecurity capabilities.
- The letter encourages policymakers to expand access to computational resources and investments in shared training assets.
Open-weight AI models allow developers to access and improve existing technology, playing a critical role in the AI innovation landscape. This demand for open models comes amid U.S. concerns about Chinese firms allegedly stealing American AI technologies. Future discussions will likely focus on balancing innovation with national security.
Read the original ↗
Google Signs EU Code on Labeling AI-Generated Content
Google's commitment to the EU's Code of Practice on AI-Generated Content marks a significant step toward ensuring transparency in synthetic media, impacting compliance for the entire industry.
- Google announced on July 24, 2026, that it will sign the EU’s voluntary Code of Practice on Transparency of AI-Generated Content.
- The EU’s AI Act, effective August 2, 2026, requires companies to label synthetic media outputs in a machine-readable format.
- Violations of the transparency rules can result in fines of up to 15 million euros or 3% of a company's annual worldwide turnover.
- Google's signature signifies a legal commitment to comply with the code's obligations, providing predictability across the 27 EU member states.
- The code does not mandate specific technologies but requires a combination of methods for labeling AI-generated content.
The EU's Code of Practice aims to ensure transparency in AI-generated content, offering guidance for compliance under the forthcoming EU AI Act. As part of its strategy, Google has previously signed a code for general-purpose AI models and has been enhancing watermarking technologies to support labeling efforts. Observers will be watching how effectively the regulations are enforced after the initial deadline in August 2026.
Read the original ↗
Nvidia and Microsoft Back Open-Weight AI in Joint Letter
The coalition's call for an open model ecosystem highlights a significant debate over AI leadership and regulatory approaches in the U.S., especially in light of competitive pressures from international players.
- A coalition of 25 American tech companies published a joint letter advocating for open-weight AI models to maintain U.S. leadership in the field.
- Signatories argue that open-weight models enhance safety, increase access, and foster competition, contrasting with closed models that limit user control.
- The letter counters U.S. government efforts to restrict access to open models, particularly those from China, amid ongoing cybersecurity concerns.
Open-weight AI models are systems with publicly available trained parameters, allowing users to modify and run them independently. This debate arises amidst a context of increasing competition in AI, especially from Chinese models such as Kimi K3, which is already challenging leading American models. Upcoming developments will reveal the real-world implications of this advocacy as more open models are released and analyzed.
Read the original ↗
U.S. Space Development Agency Orders 36 Space Vehicles for Defense Against Missile Attacks
The U.S. Space Development Agency's order for 36 space vehicles enhances the country's capacity for missile defense against evolving threats, which is critical for national security.
- The U.S. Space Development Agency has contracted L3Harris Technologies and Sierra Space Corporation to produce 36 Accelerated Missile Defense Tranche 3 space vehicles.
- These vehicles will operate in four distinct orbital planes, expanding the U.S. missile tracking and defense network by 2028.
- Half of the vehicles will be constructed by Sierra Space Corporation, while L3Harris Technologies will create the other half, focusing on missile defense variants.
- The new vehicles will integrate infrared sensors and a low-latency communication network for improved interoperability with existing satellite systems.
The U.S. Space Development Agency supports the U.S. Space Force by managing space-based systems for missile defense. The new order builds on previous satellite tranches designed for enhanced missile warning and tracking capabilities, highlighting ongoing advancements in military satellite technology.
Read the original ↗
The 5,700-Year-Old Psychedelic Drug Mescaline Finally Got a Modern Brain Scan, And The Results Were Unexpected
The new findings on mescaline's brain activity offer crucial insights into its distinct neural pathways, which could reshape understanding of psychedelic substances and their effects on the brain.
- Mescaline's recent brain scan revealed it predominantly affects the cerebellum rather than traditional psychedelic areas of the brain.
- The study conducted by Northeastern University scanned 24 rats and observed significant changes in cerebellar activity after a single dose of mescaline.
- Previous studies have primarily investigated LSD and psilocybin, suggesting that mescaline may produce its effects through different neural mechanisms than these other psychedelics.
Mescaline, derived from the peyote cactus, has been used for thousands of years but was classified as a Schedule 1 substance in 1970, limiting research. Recent studies are finally beginning to explore its effects on the brain more deeply, especially after decades of regulatory restrictions.
Read the original ↗Artificial Intelligence (484)
Understanding the core concepts of agentic AI is essential for engineers to ensure effective and safe development of AI technologies that can act autonomously.
- The article discusses five key concepts related to agentic AI, which refers to AI systems that can make decisions and act on their own.
- It emphasizes the importance of these concepts for engineers working in the field of artificial intelligence.
Agentic AI is a concept in artificial intelligence where AI systems can operate independently and make decisions. Familiarity with foundational concepts in this area is crucial for developers to mitigate risks associated with autonomous AI actions.
Based on the headline only — the full article could not be fetched.
Read the original ↗Samsung’s Galaxy Unpacked Lineup, Netflix to Buy AI Startup, and the Latest on the Paramount Skydance Merger With Warner Bros.
The acquisition of AI startup by Netflix signals a growing trend of mainstream companies incorporating artificial intelligence into their services and offerings.
- Netflix is acquiring an AI startup founded by Ben Affleck, indicating its commitment to integrating AI technologies.
- Samsung's Galaxy Unpacked event showcases the latest advancements in technology, highlighting the company's position in the competitive tech market.
- Updates on the merger between Paramount and Skydance with Warner Bros. reflect ongoing changes in the media landscape.
Samsung is a leading technology company known for its smartphones and consumer electronics. Netflix continues to expand its content and technology offerings, and the merger between Paramount and Warner Bros. reflects a trend towards consolidation in the media industry. Future developments in AI integration and media mergers should be monitored closely.
Read the original ↗
OpenAI’s new voice mode makes it to the ChatGPT desktop app
The update enhances user interaction with AI, streamlining complex tasks through voice commands, which could significantly improve productivity in various applications.
- OpenAI has updated its ChatGPT desktop app to include support for ChatGPT Voice, allowing voice interaction with AI agents.
- The new voice mode utilizes OpenAI's ChatGPT-Live voice models, launched earlier this month.
- ChatGPT Voice can execute complex commands involving multiple steps and access on-screen content on macOS via Appshots.
OpenAI's ChatGPT is an AI language model designed for communication and task execution. This new feature builds on previous voice capabilities, particularly for mobile, and may influence further advancements in desktop AI interactions. Users should watch for ongoing developments in AI voice technology across various platforms.
Read the original ↗
AI firms want more data centers; Trump’s EPA may give neighbors less say
Proposed EPA rule changes could limit public involvement in permitting processes for air polluting facilities, impacting local communities near new data centers that rely on fossil fuel energy sources.
- The Trump administration is considering a rule change to allow states to control public participation in permitting for new air pollution sources.
- Data centers are increasingly relying on minor source permits with less public oversight, facilitating easier construction of polluting facilities.
- Communities across the U.S. are facing challenges as public opposition slows down data center developments, leading to this proposed regulatory change.
Data centers are facilities housing computer systems and associated components, increasingly reliant on fossil fuel-based energy. Currently, federal regulations require some public input for pollution permits, but variances exist among states. Future monitoring will focus on how these regulatory changes affect public engagement and environmental protections.
Read the original ↗