🚀 Zaprep: Your Socials on Steroids. 免费开始 — 每月自动发送 1,000 条私信，将互动转化为潜在客户。 ，每月 1,000 条自动私信。
最新 AI 资讯
Databricks co-founder wins prestigious ACM award, says ‘AGI is here already’
Databricks co-founder and CTO Matei Zaharia almost missed the email telling him that he was the 2026 recipient of the ACM Prize in Computing. “Yeah, it was a surprise,” he told TechCrunch. Back in 2009, the tech Zaharia developed for his PhD at UC Berkeley, under the tutelage of famed professor Ion Stoica, was launched into Databricks. Zaharia had created a way to dramatically speed the results of slow, clunky, big data projects and released it as an open source project called Spark. Big data was in those days what AI is today and Spark turnedthe tech industry on its ear. The 28-year-old Zaharia became a tech celeb. Since then, he has helmed the engineering at Databricks, growing it into a cloud storage giant and now a data foundation for AI and agents. Along the way the company has raised over $20 billion —valuing it at $134 billion— and hit $5.4 billion in revenue. The Silicon Valley dream. On Wednesday, the Association for Computing Machinery issued him the award for his collective contributions. The award comes with a $250,000 cash prize that he is donating to an as-yet-to-be-determined charity. Zaharia, who in addition to his CTO duties is also an associate professor at UC Berkeley, is looking forward, not back. Like everyone else in the Valley, the future he sees is filled with AI. “AGI is here already. It’s just not in a form that we appreciate,” he told TechCrunch. “I think the bigger point of it is: We should stop trying to apply human standards to these AI models.” A person, for instance, can only pass the bar exam to be a lawyer if they’ve integrated vast amounts of knowledge. But an AI can ingest vast amounts of facts easily. If it answers knowledge questions correctly, that doesn’t equate to general knowledge. This tendency to treat AI like a human can have some profoundly negative impacts. He offers the example of the popular AI agent OpenClaw. “On the one hand, it’s awesome. You can do so many things with it. It just does them automatically,” he said. But it’s also “a security nightmare” because it’s designed to mimic a human assistant that you trust with things like passwords. That leads to the risk of being hacked, or the agent spending unauthorized money from your bank because your browser is logged in. “Yeah, it’s not a little human there,” he says. As a professor and product engineer, Zaharia is most excited about how AI can help automate research on everything from biology experiments to data compilation. Just like how vibe coding made prototyping and programming accessible to anyone, he thinks that accurate, no-hallucinations AI-powered research will someday become universal. “Not that many people need to build applications, but lots of people need to understand information,” he said. Eventually we’ll make AI work better for us by having it lean into its strengths: telling us what every rattle in our car means, or scanning beyond text and images to include radio and microwaves, or, what he’s seeing students do now, simulate molecular-level changes and predict their effectiveness. “The thing that I’m most excited about is what I’d call AI for search, but specifically for research or engineering,” he said.
View
OpenAI releases a new safety blueprint to address the rise in child sexual exploitation
In response to escalating concernsabout child safety online,OpenAI has unveiled a blueprint to enhance U.S. child protection efforts amid the AI boom. TheChild Safety Blueprint, which was released Tuesday, is designed to help with faster detection, better reporting, and more efficient investigation into cases of AI-enabled child exploitation. The overall goal of the Child Safety Blueprint is to tackle the alarming rise in child sexual exploitation linked to advancements in AI. According to theInternet Watch Foundation(IWF), more than 8,000 reports of AI-generated child sexual abuse content were detected in the first half of 2025, a 14% increase from the year prior. This includes criminals using AI tools to generate fake explicit images of children for financial sextortion and to generate convincing messages for grooming. OpenAI’s blueprint also comes amid increased scrutiny from policymakers, educators, and child-safety advocates, especially in light oftroubling incidentswhere young individuals died by suicide after allegedly engaging with AI chatbots. Last November, the Social Media Victims Law Center and the Tech Justice Law Project filed seven lawsuits in California state courts, alleging that OpenAI released GPT-4o before it was ready. The suits claim the product’s psychologically manipulative nature contributed to wrongful deaths by suicide and assisted suicide. They cite four individuals who died by suicide and three others who experienced severe, life-threatening delusions after extended interactions with the chatbot. This blueprint was developed in collaboration with the National Center for Missing and Exploited Children (NCMEC) and the Attorney General Alliance, as well as with feedback from North Carolina Attorney General Jeff Jackson and Utah Attorney General Derek Brown. The company says that the blueprint focuses on three aspects: updating legislation to include AI-generated abuse material, refining reporting mechanisms to law enforcement, and integrating preventative safeguards directly into AI systems. By doing so, OpenAI aims not only to detect potential threats earlier but also to ensure actionable information reaches investigators promptly. OpenAI’s new child safety blueprint builds onprevious initiatives, including updated guidelines for interactions with users under 18, which prohibits the generation of inappropriate content, or encouraging self-harm, and avoiding advice that would help young people conceal unsafe behavior from caregivers. The company recently released a safety blueprint for teens in India.
View
Astropad’s Workbench reimagines remote desktop for AI agents, not IT support
Demand for Apple’s Mac Mini hasskyrocketed,particularly in China, as the small computer has become an ideal platform for experimenting with autonomous AI agents likeOpenClawand others. Now, a company called Astropad is building out a remote desktop solution specifically for this use case. On Tuesday, Astropad CEO Matt Ronge introducedAstropad Workbench, a remote desktop solution for Apple devices that hepitchesas made “for the AI era.” While an AI agent running on a Mac Mini may not need a screen, its operator (the human) will want to log in at times to see what’s happening in order to check logs, monitor outputs, or restart stuck tasks, he says. The new remote desktop solution offers a variety of features, including high-fidelity streaming; the ability to dictate prompts and commands with your voice; plus support for other input methods like the keyboard, Apple Pencil, or touch; and clients for both the iPad and iPhone — the latter essentially putting the remote desktop solution into your pocket for on-the-go access. If you’re running AI agents across multiple Macs, Workbench offers a device chooser so you can move between them. The idea came about because it was something the team at Astropad had wanted for themselves, as had their friends. “We have heavily adopted AI at Astropad, and we’ve been using agents. And sometimes, you have an agent running on a long task, and you want to check on it,” says Ronge. “There’s not a great way to do this…there were existing remote desktop tools, but nothing built specifically for this,” he continues. “There have also been ways where you can use a terminal, or there are things like Telegram chats, but they’re limited. I mean, there are times you’ve got to see what’s happening on your Mac. You’ve got to approve a dialog or save something, or just visually see what’s happening.” Workbench also leverages the company’s proprietary, low-latency display protocol, which it calls LIQUID, which supports the workflows creative professionals use. It retains full fidelity, even at Retina resolutions, Astropad claims, and doesn’t blur lines or pixelate data. The protocol already powers Astropad’s other products, likeLuna Display, which turns your iPad into a second display, andAstropad Studio, which lets you use an iPad as a professional drawing tablet. While monitoring an AI agent may not always need a high-fidelity solution, Ronge points out that it’s something that’s nice to have — especially if you’re approving designs or mock-ups your AI agent made. Of course, remote desktop software has existed for some time, meaning Astropad has well-established rivals like Jump Desktop, RustDesk, AnyDesk, Parsec, VNC-based solutions, and many more. But Ronge suggests that those weren’t designed for the specific needs of using remote desktop software to keep tabs on AI agents. With Workbench, it’s easy to check on the status of logs to see your AI agents’ progress in order to spot issues, restart stalled jobs, and make other changes, but what’s more, you can do this from your iPhone or iPad. “We’ve been doing iPad stuff for years — it’s been, like, our whole company for the past 10 years. So we have a lot of experience in making good iPad apps,” Ronge says. “We know how to make good iOS apps…so we did that, and then we also added a voice model.” The tech uses Apple’s voice model so you can talk to your phone and direct your AI agent to do something with a press of the microphone button. “It’s a very natural way to work with agents. That’s the kind of feature that existing remote desktop [apps] just don’t have — they’re built for more traditional, enterprise-style remote desktop.” As a new release, there will still be some bugs and polishing needed, but the team is continuing to work on the product. Next up, they plan to launch Windows and Linux support and refine the iPhone app. The new software runs on macOS 15 and up and iOS 26,and is available as a free downloadoffering 20 minutes of access per day. For unlimited access, the cost is $10 per month, or $50 per year. Astropad, a bootstrapped and profitable small tech business, has over 100,000 customers, including those who have bought its iPad hardware accessories and its software. With Workbench, Ronge believes the company has the potential to reach both AI enthusiasts and businesses as remote support for AI agents becomes more common. “I totally think businesses are gonna buy it. I mean, just the productivity gains I’m seeing from it myself — this is totally headed to businesses. It’s just too powerful,” he notes.
View
Anthropic Taps Former Google Advisor to Lead India Policy Efforts
Anthropic plans to deepen engagement with policymakers, enterprises, and civil society as it builds a local presence.
View
Rupee Depreciation May Lift Indian IT Earnings Optics, Not Growth
Currency-led gains offer short-term relief, not a signal of recovery in IT demand.
View
Project Glasswing: Anthropic Debuts Claude Mythos Preview, Its First Cybersecurity-Focused AI Model
Anthropic, on Tuesday, announced Project Glasswing, a new initiative focused on artificial intelligence (AI)-powered cybersecurity. Powering the project for the San Francisco-based AI firm is a general-purpose model dubbed Claude Mythos Preview. The company introduces the model as its most capable when it comes to coding and agentic tasks. For now, it is only available to the partners of Project Glasswing and will enable them to detect security flaws and vulnerabilities across their tech stack. The AI firm has also allocated financial resources for the initiative.
View
Cyient Semiconductors Completes $85 Mn Acquisition of US-Based Kinetic Technologies
The deal aims to expand Cyient’s global power semiconductor capabilities and strengthen India’s role in chip innovation.
View
The Delve Controversy Reminds One of Builder.ai and the Pitfalls of AI Hype
The recent split between Y Combinator and the AI startup Delve has ignited a firestorm of scrutiny over its compliance practices.
View
BonV Aero Breaks Ground on ₹300 Cr Drone Facility in Odisha
Drone tech company BonV Aero’s project in Khordha aims to create 1,000 jobs in two years.
View
This Startup Wants to Make Voice Programmable for Developers
Murf is building a low-latency voice infrastructure layer, taking on global players with faster, cheaper, multilingual AI.
View
‘Campuses Not Career-Ready’: Karnataka Minister Flags Urgent Reset in Higher Education
Karnataka Minister MC Sudhakar has called for in-degree skilling and a focus on emerging areas.
View
Actress Milla Jovovich Launches AI-Powered Memory Tool Based on Memory Palace Technique
By integrating AI, MemPalace opens up the technique to a wider audience seeking to preserve cognitive sharpness.
View
