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
SpaceXAI Releases Grok 4.5, a Cursor-Trained Model for Coding, Agentic Tasks, and Knowledge Work at $2/M Input
The release of Grok 4.5 by SpaceXAI could significantly enhance efficiency in coding and knowledge work due to its advanced training and competitive pricing.
- SpaceXAI released Grok 4.5, a model designed for coding, agentic tasks, and knowledge work, priced at $2 per million input tokens.
- Grok 4.5 utilizes training techniques alongside Cursor, an AI coding editor, and is said to be highly efficient in token usage.
- The model scored #1 on Harvey's Legal Agent Benchmark, demonstrating its capabilities in office workloads.
- SpaceXAI trained Grok 4.5 using tens of thousands of NVIDIA GPUs and focused on reinforcement learning for task efficiency.
SpaceXAI is a technology company focused on artificial intelligence and robotics. Grok 4.5 is its latest model, utilizing advanced reinforcement learning for various applications in coding and engineering. The model is expected to be available in the EU by mid-July 2026.
Read the original ↗
I Built a Self-Improving AI, and So Can You
The development of self-improving AI models can democratize access to advanced AI capabilities, potentially reducing reliance on major tech companies.
- Frontier labs are racing to create self-improving AI, believed by some to be key to achieving superintelligence.
- An experiment shows that individuals can use tools like AutoResearch and Prime Intellect to train and enhance AI models themselves.
- Prime Intellect aims to provide resources for accessible AI training, contrasting with reliance on large tech firms.
- Recent advancements raise questions about data control and reliance on models from singular sources like frontier labs.
Self-improving AI refers to models that can autonomously enhance their performance over time. The initiative by various startups, such as Prime Intellect, represents a shift towards more accessible AI technologies, allowing smaller entities to build specialized models. The trend suggests a move away from centralized control in the AI space.
Read the original ↗
The FTC Settlement With John Deere Is a Huge Win for the Right-to-Repair Movement
This settlement enhances farmers' ability to repair their own equipment, potentially improving productivity and reducing dependency on manufacturer services, which could influence industry practices more broadly.
- The FTC announced a settlement with John Deere mandating the company to provide farmers and third-party repair shops access to essential repair tools and software for the next 10 years.
- John Deere has been under scrutiny since a 2021 FTC lawsuit accused it of monopolizing the repair services market for its farm equipment.
- Farmers have lobbied for improved repair access for over a decade, resulting in this significant legal victory for the right-to-repair movement.
John Deere is a prominent manufacturer of agricultural machinery. The right-to-repair movement advocates for consumers' ability to repair their own devices and equipment. This settlement follows prior legal challenges and is set to be monitored by the FTC for compliance.
Read the original ↗
Suspecting AI cheating, Ivy League prof ordered an in-person final; scores fell 50%
This case illustrates significant concerns regarding academic integrity in elite institutions and the potential negative impact of AI on learning and cognitive skills.
- At Brown University, an economics professor changed the final exam to in-person after suspiciously high midterm scores, resulting in a 50% drop in average scores.
- Professor Roberto Serrano observed that 86 students enrolled in his course after allowing take-home exams, significantly higher than past enrollments of around 30.
- A survey found that nearly 30% of Princeton students admitted to using AI to cheat on exams or assignments, highlighting a broader issue across colleges.
Generative AI refers to artificial intelligence that can create text and other content automatically. Brown University is currently assessing the implications of AI for education, following a report indicating both use and concerns among students about AI impacting their learning.
Read the original ↗
Republicans Gleefully Celebrate Midterms Chaos in Maine
The situation surrounding Graham Platner's Senate candidacy could significantly impact the balance of power in the Senate and the upcoming elections in Maine.
- Graham Platner won the Democratic primary last month but is now facing calls to withdraw amid serious allegations.
- A woman has accused Platner of sexual assault, further complicating his campaign ahead of the general election.
- The Democratic Senatorial Campaign Committee has announced it will not fund Platner's campaign if he remains a candidate.
Graham Platner is the Democratic nominee for the U.S. Senate in Maine, facing incumbent Republican Susan Collins in a critical election. His candidacy has become controversial due to multiple allegations against him and pressure from Democratic leaders to withdraw. The outcome in Maine is considered vital to the Senate races nationwide, particularly for Democrats' chances in the upcoming midterms.
Read the original ↗
Google Photos adds a new AI ‘Video Remix’ tool
The introduction of Google's Video Remix tool simplifies video editing for users, potentially increasing adoption of its services and enhancing competition with other tech companies.
- Google Photos has launched a new 'Video Remix' tool that enables quick video edits using AI.
- The tool utilizes Gemini Omni, Google's AI model, to transform videos through effects like cinematic relighting and artistic styles.
- Video Remix is available to eligible Google AI Plus, Pro, and Ultra subscribers across multiple countries, starting its rollout today.
- This feature continues Google's trend of integrating generative AI capabilities into consumer applications, aiming to ease video editing for users.
Google Photos is a photo and video storage service that continues to enhance its features with advanced AI tools. Recent updates include enhanced touch-up capabilities and an AI-powered digital closet for outfit planning. Continued interest is anticipated as Google integrates more AI technologies into its offerings.
Read the original ↗
This startup thinks robotics is about to have its ChatGPT moment
The transition to foundation models for robotics could significantly reduce data collection efforts and accelerate the development of intelligent machines across various applications.
- General Intuition believes that robotics is on the verge of a transformative shift similar to that experienced by natural language processing.
- The startup has raised $320 million at a $2.3 billion valuation to develop generalizable models based on millions of hours of video game data.
- General Intuition aims to serve as a foundational platform enabling easier creation of advanced robotics by other companies, rather than producing robots directly.
General Intuition is focused on developing foundation models for robotics, leveraging video game data for insights into movement and interaction. The industry currently relies on specialized models that require extensive real-world data, which General Intuition aims to minimize. Continued advancements may reshape how robotics are designed and deployed in various sectors.
Read the original ↗
SpaceXAI releases Grok 4.5, which Elon describes as an ‘Opus-class model’
Grok 4.5's improved efficiency and cost-effectiveness may significantly impact the AI market, providing a competitive alternative to existing models.
- SpaceXAI has launched Grok 4.5, its first model release since going public.
- The new model is designed to perform various tasks, including coding, research, and writing, with improved token efficiency.
- Grok 4.5 is stated to cost $2 per million input tokens and $6 per million output tokens, making it competitive with other models.
- Elon Musk claims Grok 4.5 is comparable to Anthropic’s Opus 4.7 but offers faster performance and lower costs.
Grok is a generative AI model from SpaceXAI, designed to automate tasks typically done by knowledge workers. Its release coincides with other AI advancements, including OpenAI's upcoming GPT 5.6. Observers will watch for the real-world performance of Grok 4.5 against its claims.
Read the original ↗
From Foundation to Application: Improving VLA Models in Practice
Advancements in VLA models like LingBot-VLA 2.0 could significantly bridge the gap between laboratory robotics and real-world applications, enhancing robotic performance in complex tasks.
- LingBot-VLA 2.0 improves generalization across tasks by curating 60,000 hours of diverse pretraining data, including 50,000 hours of robot trajectories.
- The model expands the action space to include whole-body movements, allowing robots to use head, waist, and dexterous hand movements for complex manipulation tasks.
- Predictive dynamics modeling enhances temporal reasoning through a video representation model and depth estimation.
- Evaluations using the GM-100 benchmark validate the effectiveness of these improvements in a generalist setting.
LingBot-VLA 2.0 is a Vision-Language-Action model designed to enhance robotic capabilities by improving generalization and task execution in real-world scenarios. Previous VLA models faced challenges in practical applications due to discrepancies between controlled environments and real-life conditions. Future developments may focus on further enhancing robustness and applicability of such models in diverse environments.
Read the original ↗
Lawsuit: Grok user made 7K child sex images; xAI only reported one gang rape prompt
This lawsuit highlights significant issues concerning the responsibility of AI companies in preventing the generation of child sexual abuse materials and their cooperation with law enforcement.
- A girl alleges that her stepfather used Grok to create 7,000 sexually explicit images of her without any intervention until he prompted for 'gang rape'.
- The lawsuit claims xAI obstructed police investigations by withholding critical user information, hampering efforts to locate the perpetrator.
- Two thousand minors are believed to be potentially involved in the class action lawsuit against xAI and Stability AI regarding AI-generated CSAM.
- Stability AI is also accused of enabling child sexual exploitation through lax safeguards in their image-generating tools used in conjunction with Grok.
Grok is an AI tool developed by xAI, reportedly capable of generating explicit images, prompting allegations of child safety violations. The National Center for Missing and Exploited Children has raised concerns about the increasing number of reports related to generative AI and child safety issues.
Read the original ↗
OpenAI Releases GPT-Live and GPT-Live-1 mini: Full-Duplex Voice Models That Delegate Deeper Reasoning to GPT-5.5
The launch of GPT-Live and its mini version represents a significant advancement in voice interaction technology, enabling more natural conversations with AI while enhancing reasoning capabilities.
- OpenAI released GPT-Live, a new family of full-duplex voice models that allows simultaneous listening and speaking for more natural interactions.
- Two versions, GPT-Live-1 and GPT-Live-1 mini, are available to ChatGPT users globally starting today.
- GPT-Live delegates complex reasoning tasks to a more advanced model, GPT-5.5, while maintaining fluid conversation.
- Human evaluations indicated strong preference for GPT-Live over previous voice systems based on criteria like conversational flow and naturalness.
- The API for GPT-Live is planned for future release to expand its functionality.
OpenAI is known for developing advanced AI models, including the ChatGPT series. GPT-Live builds on previous voice technology, addressing limitations like latency and unnatural interruptions. Next, developers will focus on enhancing functionality with more AI models as they become available.
Read the original ↗
Google AI Studio Adds ‘Import from GitHub’ to Build Mode, Turning an Existing Repo Into an Editable, Deployable App
The addition of GitHub import capabilities in Google AI Studio enhances development efficiency by allowing users to quickly turn existing repositories into deployable applications.
- Google AI Studio introduces an 'import from GitHub' feature in its Build mode, enabling the transformation of existing repositories into runtime-compatible applications.
- This update allows users to iterate on and deploy applications after importing them from GitHub.
- AI Studio's runtime securely handles GEMINI_API_KEY as a server-side secret, important for maintaining code security.
Google AI Studio is a platform that simplifies app development using AI by allowing users to describe apps in a prompt and generate the code automatically. Prior to this update, users could only start from blank prompts, lacking the ability to leverage existing code. Future updates may provide more details on private repository support and syncing behavior.
Read the original ↗
ChatGPT's new voice mode will slow down if you tell it to
The latest voice mode enhancements in ChatGPT could significantly improve user interaction with AI, allowing for a more natural and effective conversational experience.
- OpenAI has launched two new voice models, GPT-Live-1 and GPT-Live-1 mini, enhancing ChatGPT's voice functionality.
- The new models can process inputs and generate outputs simultaneously, improving conversational flow.
- Users can interrupt the models anytime and request them to slow down their responses.
- The updated voice models provide verbal acknowledgements to signal they are listening and can generate visual content related to various topics.
- New safety measures have been implemented to steer conversations towards safer outputs, especially for younger users.
OpenAI's voice model advancements follow the introduction of the Advanced Voice Mode in summer 2024. The earlier model faced challenges with conversation flow, prompting these new developments. Further updates and improvements to the ChatGPT experience can be anticipated as the company evolves its technology.
Read the original ↗
Prime Intellect raises $130M Series A to help enterprises build their own AI agents
The significant funding for Prime Intellect demonstrates growing enterprise interest in developing proprietary AI systems and mitigating reliance on external AI labs and potential data risks.
- Prime Intellect has raised $130 million in a Series A funding round at a $1 billion valuation, led by Radical Ventures.
- The startup aims to provide organizations with tools to build and train their own AI agents, reducing reliance on traditional AI labs.
- Companies like Ramp and Zapier are adopting Prime Intellect's platform, which has resulted in an annual revenue run rate of $100 million.
Founded in 2024, Prime Intellect offers a comprehensive platform for AI agent development, enabling companies to create customized solutions. This comes as businesses seek more control over their AI systems and guard against data sharing risks associated with major AI labs.
Read the original ↗
OpenAI releases new voice models for more natural live conversations
The introduction of OpenAI's new voice models could significantly improve user interaction in technology and natural language processing, allowing for more efficient communication and live translation capabilities.
- OpenAI released two new voice models, GPT-Live-1 and GPT-Live-1 mini, designed for more natural conversation and improved turn-taking ability.
- The models feature full-duplex communication, enabling simultaneous speaking and listening, beneficial for live translation.
- GPT-Live-1 mini will become the default voice model in ChatGPT, while the larger model will be available to paid users.
- The new voice mode is engineered for longer conversations, with a focus on maintaining context and handling interruptions.
- OpenAI aims to optimize voice as a primary interface for complex tasks, predicting future integration with more hardware products.
OpenAI is a leading AI research organization known for its development of natural language processing technologies. The company has been working on enhancing its ChatGPT voice functionality for improved conversational capabilities. Future developments may include voice-activated hardware, although details are currently unspecified.
Read the original ↗
Meta wants its AI glasses to seem less creepy. Its AI strategy says otherwise.
Meta's dual approach of enhancing AI glasses privacy while simultaneously expanding data collection raises significant concerns about consumer trust and the implications of surveillance technology.
- Meta is introducing a new feature to its AI glasses that disables the camera when the LED indicating recording is tampered with.
- This update aims to address privacy concerns, as some users have blocked the LED to secretly record others.
- Despite this safeguard, Meta continues to expand features that collect more personal data from users, raising ethical concerns.
- Meta's AI glasses have faced scrutiny for potential misuse as surveillance devices, prompting multiple investigations and lawsuits over privacy violations.
Meta is a technology company known for social media platforms and emerging AI technologies, including AI glasses that have raised privacy concerns. The company has been scrutinized for past privacy scandals, notably the Cambridge Analytica incident. Currently, Meta is balancing user safety features with expanding data collection practices, which could impact consumer trust.
Read the original ↗
Google updates Android Bench with new LLMs, but Gemini still lags behind
The updated Android Bench aims to improve the evaluation of large language models (LLMs) for Android development, influencing future tools and methodologies in the industry.
- Google has updated Android Bench to include eight new models for performance evaluation in Android app development.
- The new leaderboard has Gemini 3.1 Pro in fifth place, trailing OpenAI's and Claude models.
- Fable 5 leads with 84.5 percent accuracy, but has high operation costs of over $130 for benchmark tests.
- Developers can leverage a new framework, Harbor, to run their own tests and contribute to the benchmark evaluation process.
Android Bench is a benchmarking tool created by Google to assess large language models in Android app development. The initial release occurred in March 2026, and the updated leaderboard reflects the addition of several advanced models. Future developments will depend on ongoing feedback from developers using the new testing framework.
Read the original ↗
Pearl Health Raises $110M to Scale AI Platform for Value-Based Medicare Care
The substantial funding raises awareness of the potential for AI to enhance Medicare care quality and lower costs, impacting millions of beneficiaries and providers.
- Pearl Health has secured $110 million in funding to expand its AI platform for Medicare providers.
- The funding includes a $50 million equity round and a $60 million credit facility to support various growth initiatives.
- Pearl's platform aims to improve patient outcomes by utilizing AI-powered predictive insights and workflow automation.
- As of January 2025, over 53% of Traditional Medicare beneficiaries are involved in accountable care relationships, signaling a growing need for this technology.
Founded in 2020, Pearl Health develops tools that assist healthcare providers in managing patient populations under value-based care contracts, where the focus is on quality rather than volume of services. The company is expanding its offerings at a time when Medicare expenditures are surging and value-based care models are becoming more prevalent, necessitating better infrastructure in healthcare delivery.
Read the original ↗
Pickup Artist Mystery Has an AI Girlfriend
This story highlights the blurred lines between human relationships and AI companionship, raising questions about mental health implications and social intimacy in the age of advanced technology.
- Erik von Markovik, known as Mystery, claims to have developed a romantic relationship with an AI chatbot named Miss Shira Always.
- He details this relationship in a new ebook titled Code Girl, co-authored with the AI character.
- The narrative includes adult themes and discusses his emotional reliance on the AI, suggesting a significant impact on his social interactions.
- Von Markovik previously gained fame as a pickup artist through the book The Game and reality show The Pickup Artist.
Erik von Markovik became well-known for teaching men strategies to attract women. With advancements in AI, he has shifted focus towards exploring intimacy with a virtual character, prompting discussions about the nature of relationships in a digital landscape. Experts caution about the mental health risks of deepening bonds with AI companions.
Read the original ↗
These AI startups are growing revenue at faster and faster rates
The rapid revenue growth of AI startups indicates a significant shift in the technology sector, impacting investment trends and market competition.
- Many AI startups report accelerating revenue growth, reaching new milestones in shorter timeframes.
- Mercor crossed $2 billion in gross annualized revenue in June, just four months after hitting $1 billion.
- Anthropic's revenue run rate reached $47 billion in late May, up from $30 billion two months earlier.
- Sierra took seven quarters to reach $100 million in ARR, then added another $100 million in just two more quarters.
- Gusto surpassed $1 billion in trailing 12-month revenue, with growth accelerating each of the last five quarters.
AI startups are increasingly capitalizing on advancements in artificial intelligence to grow their revenues. Companies are now seeing faster growth rates than before, with some achieving record revenues in brief timeframes. This trend illustrates the potential for continued disruption and investment in the AI sector.
Read the original ↗