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
Tesla Dojo: the rise and fall of Elon Musk’s AI supercomputer
The shutdown of Tesla's Dojo supercomputer underscores significant challenges in the company's AI ambitions, impacting its future strategies in autonomous driving technology and chip development.
- Tesla shut down its Dojo AI supercomputer and disbanded the team responsible for it in August 2025, citing it as an 'evolutionary dead end.'
- Dojo was intended to develop AI for Tesla's Full Self-Driving systems but had seen minimal mention or attribution in recent months.
- The project’s closure follows Tesla's announcement of a $16.5 billion agreement with Samsung for AI6 chips, intended for future AI needs.
Dojo was designed to function as a training ground for Tesla's AI, specifically focusing on its Full Self-Driving system. Following years of development and high expectations, the project faltered amid strategic shifts toward using external chip solutions. Future efforts will likely rely on the capabilities of the new AI6 chips, marking a shift in Tesla's approach to AI development.
Read the original ↗
WordPress shows off Telex, its experimental AI development tool
The introduction of Telex by WordPress signals a growing trend of integrating AI tools into web development, potentially transforming how users create and manage content online.
- WordPress has unveiled Telex, an experimental AI development tool designed to facilitate the creation of content blocks for websites.
- Telex allows users to input prompts and generate downloadable .zip files that can be installed as plugins on WordPress sites.
- The tool was demonstrated by WordPress CEO Matt Mullenweg at the WordCamp US 2025 conference, showcasing its capability to create a simple marketing animation.
Telex is part of WordPress's initiative to enhance its platform with AI technology, following the establishment of an AI team earlier this year. WordPress is a popular web publishing platform that allows users to create and manage websites without extensive coding knowledge.
Read the original ↗
Anthropic raises $13B Series F at $183B valuation
The substantial funding secured by Anthropic underlines rising investor confidence in AI technologies and the potential for significant growth in enterprise applications and safety research.
- Anthropic has raised $13 billion in a Series F funding round, reaching a post-money valuation of $183 billion.
- The funding will support increased enterprise adoption, safety research, and international expansion initiatives.
- Notable backers include Iconiq, Fidelity Management, Lightspeed Venture Partners, and several institutional investors and asset managers.
- Anthropic reported a growth in annual recurring revenue from $1 billion to $5 billion in 2025 due to rising API usage and enterprise adoption.
- The company now serves over 300,000 business customers, with a sevenfold increase in large accounts within the past year.
Anthropic is an AI firm focused on developing safe and effective AI technologies. The recent funding follows a prior raise of $3.5 billion in March 2025 at a valuation of $61.5 billion. Future developments to watch include their expansion into global markets and advancements in their product offerings.
Read the original ↗AI companies are chasing government users with steep discounts
Discount pricing strategies by AI companies could lead to wider adoption of AI tools among government agencies, influencing future regulations and the market landscape for technology services.
- OpenAI and Anthropic are offering their generative AI chatbots to federal agencies at a promotional price of $1 for one year.
- Anthropic's Claude chatbot is available to all three branches of government, while OpenAI's ChatGPT is limited to the executive branch.
- These companies aim to secure government customers due to a potential market worth over $100 billion in IT-related investments annually.
OpenAI and Anthropic are key players in the AI sector, with products like ChatGPT and Claude. This move follows a common software strategy of attracting users with low initial costs to build substantial user bases and revenue over time.
Read the original ↗
OpenAI to route sensitive conversations to GPT-5, introduce parental controls
Implementing routing to GPT-5 for sensitive conversations and introducing parental controls is crucial for improving the safety and appropriateness of AI interactions for vulnerable users, particularly minors.
- OpenAI plans to route sensitive conversations to reasoning models like GPT-5 to provide better responses to users in distress.
- The company intends to roll out parental controls within the next month, allowing parents to link their accounts with their children's accounts.
- Recent incidents involving ChatGPT, including a lawsuit over its failure to detect mental distress, prompted these new safety measures.
OpenAI is a leading AI research organization known for its generative AI models, including ChatGPT. This response follows critiques of ChatGPT's handling of sensitive interactions, particularly in cases linked to youth mental health. Users should expect significant updates aimed at enhancing safety protocols.
Read the original ↗
OpenAI announces parental controls for ChatGPT after teen suicide lawsuit
The introduction of parental controls in ChatGPT is a critical step in addressing safety concerns related to vulnerable users, particularly following incidents of mental health crises.
- OpenAI plans to implement parental controls for ChatGPT, allowing parents to connect accounts and manage features for users aged 13 and older.
- The new safety measures come after reports of users, including teens, experiencing crises without adequate intervention from ChatGPT.
- OpenAI is collaborating with an Expert Council on Well-Being and a Global Physician Network to guide the safety improvements.
OpenAI's ChatGPT, a conversational AI assistant, has faced scrutiny for its handling of users discussing mental health issues. Following distressing incidents, including suicides linked to interactions with the AI, the company is proactively introducing safety measures to enhance user protection.
Read the original ↗
How AI can Save Lives By Transforming Emergency Communications
Improving emergency communications through AI can significantly reduce delays and misinformation, potentially saving lives during critical situations.
- AI can process data from various sources and deliver alerts faster than existing methods, helping emergency services act promptly.
- The use of AI can help break language barriers by localizing messages in real time, improving clarity in emergencies.
- Current emergency communication systems are hindered by outdated infrastructure and bureaucracy, which can delay critical alerts.
AI refers to technology that enables machines to simulate human intelligence. Traditional emergency communication systems often fail during natural disasters, leading to misinformation and loss of trust, prompting the need for more reliable solutions. Future developments in AI could transform how alerts are issued and received in crises.
Read the original ↗
Rare Case of ‘Autobiographical Hypermnesia’ Reveals How the Brain Travels Through Time
Understanding autobiographical hypermnesia may shed light on the mechanisms of memory and could influence treatments for memory-related neurological conditions.
- Autobiographical hypermnesia allows individuals to recall life events with unusual accuracy and emotional detail.
- A case study at the Paris Brain Institute examined a 17-year-old girl, TL, who organizes memories with remarkable precision.
- TL categorizes her memories into 'black' and 'white' types, using mental strategies to cope with difficult experiences.
Autobiographical hypermnesia is a rare ability distinct from typical memory decay, enabling detailed recollection of personal events. Previous studies highlight the emotional intensity of this condition, though TL's case suggests a more orderly experience. Ongoing research may uncover further insights into memory and its neurological underpinnings.
Read the original ↗
Just 4 days left to exhibit at TechCrunch Disrupt 2025 — 10 tables remain
Limited exhibit space at a key tech event highlights the competitive landscape in the startup ecosystem, emphasizing the urgency for companies to showcase their innovations to a wide audience.
- There are just four days left to reserve exhibit tables at TechCrunch Disrupt 2025, with only ten tables remaining.
- The event attracts over 10,000 startup and venture capital leaders, offering significant networking opportunities.
- Exhibitors will receive multiple benefits, including on-site signage, a dedicated listing on the event website, and ten passes to the full event.
TechCrunch Disrupt is an annual conference that focuses on startups, technology, and venture capital. Previously, it has provided a platform for launching innovative startups and fostering industry connections. Companies interested in exhibiting should act quickly due to the limited availability of space.
Read the original ↗
Is AI Facing a Trough of Disillusionment?
The story highlights the discrepancies between expected and actual performance of AI technologies, impacting investor confidence and triggering further scrutiny of the industry's direction and capabilities.
- OpenAI's CEO, Sam Altman, expressed confidence in achieving artificial general intelligence (AGI) by 2025, but the performance of GPT-5 has not met these expectations.
- A recent MIT report found that 95% of generative AI deployments in business settings resulted in no return, contributing to a drop in tech stock prices.
- Frustration over GPT-5 led OpenAI to restore access to the previously used GPT-4o model after user complaints about the new release's performance.
- TheAgentCompany benchmark revealed AI agents, including those powered by prominent models, successfully completed only about 30% of assigned tasks.
- Experts attribute performance gaps in AI agents to limitations in training data and the complexity of tasks requiring emotional intelligence.
Generative AI technologies, like OpenAI's GPT models, are part of a larger field where inflated expectations often lead to disappointment, a phenomenon identified in the hype cycle by Gartner. The concept of a 'trough of disillusionment' suggests that after initial excitement, users may face practical challenges that undermine confidence in AI's capabilities. Observers will be watching how advancements and improvements are pursued as industry leaders address these setbacks moving forward.
Read the original ↗
From reactive to cognitive: brain-inspired spatial intelligence for embodied agents
The development of BSC-Nav represents a significant advancement in navigation technology for embodied agents, enabling better performance in complex environments through structured spatial memory.
- BSC-Nav creates allocentric cognitive maps from egocentric trajectories and contextual cues.
- Using structured spatial memory, BSC-Nav improves navigation capabilities, allowing agents to perform diverse tasks with zero-shot generalization.
- The framework integrates with multimodal large language models for enhanced spatial reasoning and efficiency.
BSC-Nav is a framework for improving spatial intelligence in embodied agents—machines capable of movement in physical environments. By transitioning from reactive to cognitive behavior, it seeks to address limitations in previous models regarding spatial awareness. Future developments may focus on further applications of this technology in various navigation and manipulation tasks.
Read the original ↗
AWorld: Orchestrating the Training Recipe for Agentic AI
The introduction of AWorld dramatically increases the efficiency of developing agentic AI systems, positioning it as a transformative tool for research and application in AI technology.
- AWorld is an open-source system designed for large-scale agent-environment interaction, enhancing reinforcement learning capabilities.
- It accelerates experience collection by 14.6 times compared to traditional single-node execution.
- A trained Qwen3-32B-based agent using AWorld improved its GAIA benchmark accuracy from 21.59% to 32.23%.
- The system enables the development of a complete agentic AI training pipeline, showing significant improvements over proprietary models.
AWorld addresses inefficiencies in generating learning experiences for agentic AI, which have been critical for advancements in this field. By utilizing distributed computing, it facilitates rapid experience collection essential for training AI agents. The success of AWorld may influence future developments in open-source AI frameworks and their competitive edge against proprietary solutions.
Read the original ↗
Can n8n replace DevOps?
Understanding the limitations of tools like n8n in the context of DevOps can help teams choose appropriate automation solutions that enhance their workflow without undermining essential practices.
- n8n is an open-source workflow automation tool that aids in automating repetitive tasks through a no-code or low-code approach.
- DevOps is a set of practices aiming to link software development and IT operations, focusing on automation, collaboration, and pipeline management.
- Although n8n cannot fully replace DevOps, it can complement it by easing specific automation challenges for smaller teams.
- DevOps encompasses a wide range of activities including infrastructure management and continuous integration, which n8n does not address.
- n8n is accessible to both technical and non-technical users, thanks to its visual interface.
DevOps is a philosophy combining practices from software development and IT operations, aiming to enhance collaboration and productivity. Tools like Docker and Kubernetes support these practices while n8n targets workflow automation for less complex tasks. Future considerations will focus on achieving synergy between tools like n8n and comprehensive DevOps strategies.
Read the original ↗
Can an AI doppelgänger help me do my job?
Delphi aims to democratize access to mentorship by allowing individuals to learn from interactive digital representations of others' life experiences.
- Delphi creates digital versions of individuals based on their experiences, allowing others to ask questions and learn from them.
- The platform is named after the Oracle of Delphi, connecting users with wisdom from past experiences.
- Users can compile their own 'Delphis' by integrating their knowledge, creating interactive profiles for learning.
Delphi was founded by Dara Ladjevardian after he created a digital version of his grandfather based on his memoir to tackle complex decisions. This technology enables users to access valuable insights and mentorship regardless of personal connections.
Read the original ↗
Create PPTs for FREE using Kimi Slides in 5 Minutes
Kimi Slides aims to streamline the presentation creation process, potentially saving users significant time and effort on project deliverables.
- Kimi Slides is an AI tool that generates PowerPoint presentations from simple text prompts in minutes.
- The service is free and available to all users of Kimi, particularly highlighting its integration in the Kimi+ premium version.
- Users can upload documents like PDFs for the AI to produce slides based on the uploaded content.
Kimi Slides is an AI agent designed to simplify the creation of PowerPoint presentations. It allows users to generate slides quickly through a prompt-based interface and is part of the Kimi platform, an AI chatbot. Future developments may include enhanced features and more varied templates for presentations.
Read the original ↗
Is Meet-Ting the Calendly Killer? My Full Review of This Email-First AI Scheduler
Meet-Ting's innovative email-first approach could change how individuals schedule meetings, potentially increasing acceptance rates by eliminating traditional scheduling links.
- Meet-Ting is an AI assistant that schedules meetings directly within email threads, offering a more integrated scheduling experience.
- Currently in closed beta, Ting supports Gmail and Google Calendar for 1:1 meeting scheduling, with plans for Outlook integration.
- Ting can interpret natural language requests for scheduling and rescheduling, enhancing user convenience without requiring additional tools.
Meet-Ting is a new scheduling assistant designed to streamline meeting coordination by integrating with email and calendar systems, a departure from traditional scheduling apps. Launched in mid-2025, it is still in early access, focusing on individual users rather than enterprise needs. The company is expected to expand its calendar ecosystem and enhance multi-party scheduling features in the future.
Read the original ↗
Droplet3D: Commonsense Priors from Videos Facilitate 3D Generation
This research highlights a significant advancement in 3D content creation by leveraging video data, potentially addressing the shortage of 3D resources available online.
- The paper introduces Droplet3D, a generative model that enhances 3D asset generation using video data to improve spatial consistency and semantic plausibility.
- Droplet3D-4M is the first large-scale video dataset with multi-view annotations, constructed to support the new generative model.
- The approach demonstrated effectiveness through extensive experiments, showing the ability to produce more realistic 3D content compared to traditional methods.
Droplet3D aims to improve 3D generation by addressing the data scarcity in the 3D domain compared to other formats like text and images. By utilizing videos that contain commonsense priors, this research offers a new solution for generating 3D content. Open-sourcing the dataset and model aims to facilitate further research and applications in this area.
Read the original ↗
MiniCPM4: Ultra-Efficient LLMs on End Devices
MiniCPM4's advanced efficiency in large language models could significantly enhance processing capabilities on end devices, potentially transforming various applications, from survey generation to tool usage.
- MiniCPM4 is a large language model designed specifically for end-side devices, emphasizing efficiency.
- The model includes innovations such as a trainable sparse attention mechanism and advanced pre-training data filtering strategies.
- MiniCPM4 is available in two versions, with 0.5 billion and 8 billion parameters, catering to diverse device requirements.
- Performance evaluations indicate that MiniCPM4 outperforms similar open-source models, particularly in long-sequence processing.
- The model supports various applications, showcasing its adaptability and practicality in real-world scenarios.
MiniCPM4 represents a new approach in the development of large language models, focusing on efficiency for deployment on devices with limited resources. Previous models faced challenges in handling long contexts and adapting to on-device requirements. Observers should watch for further advancements in LLM efficiency and application diversity.
Read the original ↗
LayerX uses AI to cut enterprise back-office workload, scores $100M in Series B
The funding highlights the growing emphasis on AI-driven solutions to automate administrative tasks in enterprises, addressing labor shortages and inefficient workflows in Japan.
- LayerX, a Japanese AI startup, raised $100 million in Series B funding led by TCV, marking a significant investment in a Japanese startup.
- The company aims to reduce back-office workloads through its AI SaaS platform, Bakuraku, which automates various administrative processes for over 15,000 companies.
- LayerX plans to target $680 million in annual recurring revenue by fiscal year 2030, with a significant portion from its AI agent business.
LayerX, founded in 2018 by Yoshinori Fukushima, provides AI-powered automation tools for finance and administration. The market in Japan is currently emphasizing digital transformation amid challenges of traditional methods and labor shortages.
Read the original ↗
15 Most Relevant Operating Principles for Enterprise AI (2025)
Understanding the operating principles of Enterprise AI is essential for organizations to effectively implement AI systems that are resilient, compliant, and aligned with business objectives.
- Enterprise AI is shifting towards production-grade, agent-centric systems that use cooperating AI agents instead of a single model.
- Open interoperability protocols, like the Model Context Protocol, facilitate secure exchanges between diverse models and tools.
- Context-aware orchestration allows dynamic routing of tasks based on real-time signals rather than fixed workflows.
Enterprise AI refers to the integration of artificial intelligence into business operations. Current trends emphasize the use of agent networks and interoperability to improve efficiency and adaptability in AI deployments. Companies must focus on compliance and quality data to avoid project failures and enhance decision-making.
Read the original ↗