Baidu, Pony AI, and WeRide each have hundreds of robotaxis on the road in China operating commercial paid services without a human safety driver. Their technology and rider experience are generally similar to the driverless taxis operated by Waymo in the US. China's robotaxi fleet is projected to grow to tens of thousands of vehicles by the end of next year. The Chinese companies are testing autonomous vehicles in around half a dozen countries - Waymo is only testing its vehicles in one country outside of the US.
Preventive, a startup backed by OpenAI CEO Sam Altman and his husband, as well as Coinbase CEO Brian Armstrong, is working toward creating genetically engineered human children. Editing gene embryos with the intention of creating babies is banned in the US and many countries, so the company has been searching for places to experiment where the practice is allowed. Preventive is part of a growing number of startups funded by powerful people in Silicon Valley that are pushing the boundaries of fertility and working to commercialize reproductive genetic technologies.
AI spend is rising, but how are you measuring return on investment? We love this guide from You.com [9], which gives leaders a step-by-step framework to measure, model, and maximize AI impact. Get clear KPIs, four ROI formulas (and when to use each), and a battle-tested LLM prompt [9] for quickly creating your own interactive ROI calculator. Turn “we think” into “we know.”
Valdi is a cross-platform UI framework that compiles TypeScript directly to native views on iOS, Android, and macOS. It delivers native performance without sacrificing developer velocity with automatic view recycling, optimized component rendering, an optimized layout engine, and viewport-aware rendering. Valdi eliminates the traditional compile-test-debug cycle that slows native development with instant hot reload, full VSCode debugging, and a familiar syntax. It integrates easily into existing apps and can be scaled as needed.
Work is much messier today than when roles were somewhat rigid and handoffs were clean. Loops are now tighter, and the fast path from idea to impact crosses multiple disciplines. It is now better to have cross-domain fluency along with product sense and engineering craft. AI doesn't lower the cost of choosing the right thing, shaping it well, and operating it in the wild. Models can accelerate execution, but developers still own intent, architecture, and accountability.
There's a lot of mystique about Cursor. This article takes a look inside the company from the perspective of a team member who has been there for two months. The company's leaders are enthusiastic about establishing a new playbook for company building. The employees are focused on the mission - the company has made them wealthy, but there is zero chatter from employees about getting rich.
China has a million people working on AI 24/7, while the US has maybe 20,000 in Silicon Valley. US export controls are producing precisely the opposite of their intended effects and accelerating China's development. They have triggered a national mobilization. China is predicted to have more AI compute than the rest of the world combined by 2027.
UnisonDB is a database designed specifically for Edge AI and Edge Computing that enables near-instant fan-out replication across hundreds of nodes while preserving strong consistency and durability.
The External Secrets Operator is a Kubernetes operator that integrates external secret management systems, reads information from external APIs, and automatically injects the values into a Kubernetes Secret.
Tesla shareholders have voted to approve a plan to motivate Elon Musk with a bonus of as much as $1 trillion in additional stock. Musk had threatened on social media to leave the company if the measure was rejected. He claimed he wanted a big enough ownership stake in Tesla so that he couldn't be fired, so the 'robot army' he is developing doesn't fall into the wrong hands. The new pay package includes 12 chunks of stock that could give Musk control of as much as 25% of Tesla if he hits a series of milestones in the next decade.
Google's latest generation of Ironwood accelerators will become generally available in the coming weeks. The TPU v7 accelerators are a major leap in performance over prior generations. They offer performance close to Nvidia's Blackwell GPUs, when normalizing floating point perf to the same precision. The TPUs can be scaled into truly enormous compute domains - Google is offering the chips in pods of 256 up to 9,216.
The UAE's Emirate Genome Programme started six years ago with the goal of mapping the genetic code of all of the nation's around one million citizens. The country believes that gene sequencing is both a public-health project and a key part of the country's multitrillion-dollar effort to refashion itself from a petrostate into an AI powerhouse. The project has already collected DNA samples from more than 750,000 Emirates. Efforts to pair AI with genomic data have yet to yield profitable drugs, but many believe they will soon.
'Mind captioning' is a method developed at NTT Communication Science Laboratories in Japan that uses a blend of brain images and AI to generate text descriptions of what people are seeing (or even visualizing) based only on their neural patterns. In tests, the system was able to correctly describe scenes viewed by participants about half the time, even when presented with a hundred possibilities - and even when participants were just visualizing the scenes in their mind. The technology is still far from being a mind-reading device, but it could have serious ethical implications when it becomes more capable.
Some concepts are easy to grasp in the abstract, but others you really need to try before you understand them. Regardless of whether or not LLM agents are snake oil or not, it's worth learning how to make one - startups are raising millions building agents at a time when nobody really knows anything yet. It's easy to get started. This post walks readers through how to create an agent.
As large language models become more dominant, new frameworks will find it hard to gain traction because the models won't have enough training data about them. The industry needs to innovate and build new frameworks, libraries, and platform features to push the web forward and create competition. Developers need to have clear strategies to get their work into the LLM training corpus, system prompts, and developer minds. If the industry continues its current focus on maintainability and developer experience, we'll end up in a world where the web is built by LLMs using only what is entrenched in the training data.
This post presents a deconstructive view on how games are designed. It presents twelve aspects of game design to help readers get better at making games of any type. Each of the topics is deep, and there are plenty of links for further reading. The fun of making games is always right outside the edge of what the designers know.
61% of customers leave after one bad service moment. Discover how leading brands pair agent empathy with AI efficiency to deliver smarter, more trusted customer experiences. Read the ungated insight
Microsoft's new MAI Superintelligence Team, led by Mustafa Suleyman, will put human interests and guardrails first, creating systems that are aligned to human values by default.
Advice on making Android notifications perfectly reliable from a developer with years of real-world experience building apps with zero notification failure reports.
Enterprise customers expect more than a login screen. They demand SSO, directory sync, granular permissions, and detailed audit logs, all built to strict compliance standards.
Apple plans to pay Google $1 billion a year for a 1.2 trillion-parameter AI model to help run its Siri voice assistant. The model will run on Apple's own Private Cloud Compute servers, ensuring that user data remains secure. The companies are still finalizing the agreement, which will be an interim solution until Apple's models are powerful enough. The new Siri is on track to launch in Spring.
Blue Origin is getting ready for the second launch of New Glenn, which is scheduled to lift off from Cape Canaveral, Florida, no earlier than 2:51 PM ET on Sunday. The heavy-lift launch vehicle is now in the final stages of preparation for the flight. The mission will include another boost recovery attempt and the launch of NASA's ESCAPADE mission. The success of the launch will determine whether New Glenn could become a viable option for critical NASA science missions and national security contracts.
Quantinuum's Helios machine contains 98 physical qubits, and from those, it can deliver 48 logical error-corrected qubits. The 2:1 ratio is unique - other machines require anything from dozens to hundreds of physical qubits to create one logical qubit. Quantinuum expects to create a large-scale fault-tolerant machine by 2029 with its upcoming Apollo machine, which will have thousands of physical qubits and hundreds of logical ones. Its customers include JPMorgan Chase, which has been using Helios in private preview to run complex algorithms.
More than 10,000 infants in the US are born prematurely each year. Premature birth remains the nation's second-leading cause of infant death, and those who survive it may face crippling lifelong complications. Artificial wombs promise to save more babies, but the technology raises ethical questions that need to be answered before human trials begin. If perfected, the technology could rewrite the limits of viability.
Gartner's annual report looks at 11 popular collaboration tools. Atlassian was named a Leader for its Teamwork Collection — Jira, Confluence, Loom, and Rovo — which connects work, knowledge, and goals in one AI-powered system across every team. Find out how your teams can benefit from this foundation [9] on the Atlassian blog, where you can also get a copy of the report [9].
There are currently many mature options for compiler targets. This post presents a short survey of some of the popular and interesting options. Compiler targets can include machine code, other high-level languages, virtual machines, and more. Some compiler developers pick unconventional targets for academic curiosity, artistic expression, or to test the boundaries of viable compilation targets.
This post contains a list of markup/configuration languages along with commentary about what's wrong with them. Some specifications are too big and contain too many features, some are no longer maintained, and others are too ambiguous. All languages have trade-offs.
Andy Jassy recently laid out a vision for how he wants Amazon to operate like the world's largest startup. The company has been working hard to flatten its organization, getting rid of bureaucracy in order to move faster and stay competitive. Amazon announced last week it would lay off about 14,000 corporate employees and said more cuts are expected soon. The next big wave of cuts is expected to start in January. Jassy has been searching for Amazon's next opportunity - the company has made big bets on satellite internet, health care, grocery, entertainment, and self-driving vehicles.