Garry Tan, the CEO of Y Combinator, has humorously referred to himself as a victim of “cyber psychosis,” often finding it hard to catch sleep due to his enthusiasm for engaging with AI agents. In a recent episode of Walter Isaacson’s podcast, The Builders, which is set to air on Tuesday, Tan disclosed the financial investment he makes in his AI endeavors.

"I spend around $100,000 to $200,000 annually on tokens," he shared with Isaacson. Tan employs these AI agents for a variety of work-related tasks, including coding, and has made his collection of 23 AI skills, dubbed gstack, publicly accessible on Github. This initiative aims to educate others about the potential of AI agents.

He utilizes his AI the same way many do: for research and preparation related to Isaacson’s podcast, anticipating the questions he may face. However, his AI also plays a significant role in managing his everyday life.

“It has access to all my emails and Dropbox content. It has even gone through my Apple Notes, including therapy notes from the past two decades,” he explained.

Tan likens the experience of using AI agents to having a supportive companion that assists in nearly any daily activity, operating with a degree of autonomy. "Imagine it as a constant Jiminy Cricket on your shoulder, always aware of your preferences and there to lend a hand. It proactively suggests ideas and helps you think ahead,” he articulated.

As an enthusiastic early adopter of OpenClaw, Tan highlights an important lesson: treating AI agents as traditional software is a mistake. According to him, some individuals are so preoccupied with ensuring “repeatability and safety” that they overlook the extraordinary benefits of collaborating with AI, which can be remarkably human-like.

For example, Y Combinator has integrated an AI agent, aptly named Stray Light, which is transforming the company's operations. “This agent is designed with its own personality and possesses a comprehensive understanding of our history with founders, past communications, and applications,” Tan noted.

As a passionate tech advocate—fitting for someone at the helm of Silicon Valley's renowned startup accelerator—Tan is convinced that personal AI agents are on the verge of mainstream adoption, contingent on one key factor: reducing costs.

“It’s prohibitively expensive at present,” he remarked, acknowledging that his level of spending on AI agents represents a “luxury” not accessible to everyone.

However, he warned against relying on free AI agents managed by large corporations. Reflecting on the influence of companies like Meta on social media and user experience, he stated, “It’s a colossal entity that governs the algorithm, and on some level, we all wish that power didn’t rest with them. This paves the way for the idea of a personal AI that you can control and customize.”

Y Combinator's Garry Tan spends as much as $200K annually on tokens due to his heavy use of AI agents.

Generate a cinematic news thumbnail for title:Y Combinator’s Garry Tan uses AI agents so much, he spends up to $200K a year on tokens. No text in image.

Y Combinator’s Garry Tan uses AI agents so much, he spends up to $200K a year on tokens - TechCrunch

- Y Combinator CEO Garry Tan spends $100,000 to $200,000 annually on AI token subscriptions.  
- He uses AI agents for research, life management, and automation of daily tasks.  
- Tan advocates for affordable personal AI agents while cautioning against centralized control by big tech.  

Y Combinator CEO Garry Tan has, in the past, jokingly admitted that he’s suffered from “cyber psychosis” and sometimes barely sleeps because he’s so excited to be working with his AI agents. On a new episode of Walter Isaacson’s podcast The Builders, set to publish on Tuesday, Tan put a price tag on his passion.

“I spend about maybe $100[,000]-$200,000 a year on tokens,” he told Isaacson. He uses agents for countless work projects and coding tasks and has freely shared his set of 23 Claude Code skills, called gstack, on Github — an effort to help teach others what agents can do, he said.

On the one hand, he uses his agent the way many people use AI: to help him research Isaacson’s podcast and prep for the kinds of questions he might be asked. On the other, his agent also helps him run his life.

“It’s read all of my email and all of my Dropbox. It’s read, like, all of my Apple Notes, even my therapy notes going back 20 years,” he said.

He describes using AI agents as getting help with “almost anything that you do in your day-to-day life, and for it to do it somewhat more autonomously. Think of it as a Jiminy Cricket on your shoulder at all times. Like, almost anything you want to do, it knows you, and it’s there to help. And it will actually sort of run ahead and say, ‘What about this?’ Or, it’ll help you see around corners.”

As an early OpenClaw adopter who is now a power user, Tan says one thing he’s learned is not to treat AI agents like traditional software. Some people are so worried about “repeatability and safety that they actually trade off all of the beautiful parts of working with an AI agent, which basically can be so similar to a human,” he said.

For instance, Y Combinator is also using an agent that is “rebuilding the way we operate at YC,” he said. “We now have an internal AI agent, for instance, named Stray Light that kind of has a personality, and then it has a company brain. Like it actually has the total context of every founder we’ve ever backed, anyone who’s ever emailed us, like all the applications.”

Tan who is, naturally, an avid lover of technology (as anyone running Silicon Valley’s famed startup accelerator would be), believes personal AI agents will eventually become mainstream, if one major problem is solved: Costs have got to come down.

“It’s too expensive right now,” he said, adding that he knows the amount of money he spends on AI agents is “a crazy privilege that not everyone can do.”

But freebie agents controlled by big tech maybe aren’t the answer, either. When discussing the future he’d like tech to build, and how Meta as a social media company has impacted us, he said, “This is a huge company that controls the algorithm, and we all wish that they didn’t at some level,” he said. “So you could imagine a personal AI that you control, that you program.”

Y Combinator’s Garry Tan uses AI agents so much, he spends up to $200K a year on tokens

This week marks the debut of "Ramayana: Rise of a Legend," a significant new cinematic interpretation of the ancient Indian epic, the Ramayana. The film features prominent actors, including Ranbir Kapoor, Yash, and Sai Pallavi, who bring this classic tale to life for modern audiences.

In a notable move, the film will be available not only in English but also in various Indian languages, making it accessible to a diverse audience just in time for Diwali, the vibrant festival of lights.

BBC culture correspondent Noor Nanji had the opportunity to engage with the film's stars, during which Kapoor expressed his reflections on the impact of artificial intelligence in filmmaking. He also shared insights into his experience working on "Ramayana" as a new father, adding a personal touch to his portrayal in this epic story.

Ranbir Kapoor discusses AI, cinema, and being a father.

Generate a cinematic news thumbnail for title:Indian film star Ranbir Kapoor on AI, movies and fatherhood. No text in image.

Indian film star Ranbir Kapoor on AI, movies and fatherhood - BBC

- "Ramayana: Rise of a Legend" is a new adaptation of the Indian epic.  
- The film stars Ranbir Kapoor, Yash, and Sai Pallavi in leading roles.  
- It releases ahead of Diwali in English and multiple Indian languages.  

This week sees the release of Ramayana: Rise of a Legend, a major new adaptation of the Indian epic Ramayana.

The film features Ranbir Kapoor, alongside co-stars Yash, and Sai Pallavi.

It's being released in English as well as several Indian languages and is coming out ahead of Diwali - the festival of lights.

The BBC's culture correspondent Noor Nanji has been to meet the stars, with Kapoor sharing his thoughts on AI in film, and how he felt making Ramayana as a new father.

Indian film star Ranbir Kapoor on AI, movies and fatherhood

The United Kingdom is projected to confront artificial intelligence-driven cyberattacks targeting its essential infrastructure by the year 2028, as both organized crime and state actors, particularly China and Russia, increasingly incorporate AI into their cyber operations, a warning issued by the nation's cyber chief indicated on Tuesday.

This shift could lead to the deployment of AI-enhanced malware used in ransomware assaults, with the potential to unearth and exploit "vulnerabilities that have been tolerated for far too long" at an unprecedented scale, according to Richard Horne, the head of the U.K.'s National Cyber Security Center (NCSC). 

During a speech that coincided with the release of the NCSC's annual cyber report, Horne highlighted the responsibility of business leaders and organizations that have skimped on technology investments. He cautioned that they will need to justify to the public why crucial resources like food and electricity are unavailable, or why companies are failing, if they neglect updates to their cybersecurity measures or continue using outdated technology.

The report identified Chinese cyber efforts as the "most sophisticated and sustained cyber challenge" facing the U.K., and suggested that both Russia and China would routinely leverage AI to fulfill their geopolitical ambitions. Notably, Russia appears to be conducting cyber operations that intentionally remain below the threshold of armed conflict, while Iran is directly targeting individuals considered threats to its regime.

As malicious actors gain access to AI capabilities, the report predicts a significant boost in the scale and effectiveness of their operations. However, it is suggested that the primary use of AI may involve enhancing existing strategies rather than uncovering entirely new weaknesses.

Recently, governments and private sector organizations scrambled to implement vital security updates after AI identified critical vulnerabilities in a programming language that supports roughly 70% of websites worldwide. These two major flaws have the potential to grant unauthorized access to websites, enabling cybercriminals to remotely control web servers without needing a password.

Horne noted that modern technology is integral to the automation of essential processes underpinning daily life, including electricity management, water facility operations, dam controls, and manufacturing processes. He cited recent cyber intrusions into refrigeration systems in French supermarkets as a stark example of how easily adversarial actors can compromise civilian infrastructure.

The report further referenced past cyber endeavors by Russia, detailing an attack attributed to its FSB security service against Poland's energy grid in December 2025, a breach that could have left 500,000 people without power during winter.

In the U.K., cybercrime has became a significant financial burden, costing organizations nearly $20 billion annually, according to the NCSC. The car manufacturer Jaguar Land Rover faced over $260 million in losses due to a cyberattack in 2025, which affected suppliers and had a negative impact on the U.K. economy.

Anne Keast-Butler, director of GCHQ—Britain's signals intelligence agency, underlining that the NCSC is part of this agency, emphasized the urgency of action from organizations across the U.K. to safeguard not only their customers and services but also to serve as a frontline defense for the nation. She advised the implementation of fundamental yet effective security protocols, such as ensuring that devices are updated, limiting administrative access to crucial systems, and regularly backing up important data.

UK cybersecurity leader warns of AI-driven cyberattacks on essential services by 2028.

Generate a cinematic news thumbnail for title:UK cyber chief says the UK will face AI cyberattacks on critical infrastructure by 2028. No text in image.

UK cyber chief says the UK will face AI cyberattacks on critical infrastructure by 2028 - Yahoo

- The U.K. faces potential AI cyberattacks on critical infrastructure by 2028, warns cyber chief.  
- China's cyber activity poses a significant threat, with both states using AI for warfare.  
- Malicious cyber activities cost U.K. organizations roughly $20 billion annually, necessitating urgent cybersecurity investments.  

LONDON (AP) — The U.K. will likely face artificial intelligence cyberattacks on its critical infrastructure by 2028 as criminals and states, including China and Russia, increasingly use AI in their cyber activity, Britain's cyber chief warned Tuesday.

That could include AI-powered malware for ransomware attacks as the technology will allow "vulnerabilities that have been tolerated for far too long" to be discovered and exploited "at a scale that was once unimaginable," Richard Horne, the head of the U.K.'s National Cyber Security Center said.

In a speech marking the publication of the NCSC's annual cyber report, Horne said the leaders of businesses and organizations of all sizes who have "got away with tech on the cheap" will have to explain why "food isn't on the shelves, lights aren't on, businesses have gone under," if they fail to invest in cybersecurity or update out-of-date technology.

Advertisement Advertisement

Advertisement Advertisement

Chinese cyber activity is the "most sophisticated and sustained cyber challenge facing the U.K." while both Russia and China will routinely use AI to achieve their geopolitical goals, the report said. Russia is intentionally conducting cyber campaigns which fall below the threshold of armed conflict while Iran is carrying out attacks against the U.K. and targeting individuals who are a threat to the Iranian regime, Horne said.

The use of AI by states and criminals is highly likely to increase the scale and efficiency of hostile operations, but AI is more likely to be used to enhance and scale existing tactics and techniques rather than find brand new vulnerabilities, the report said.

Last week, governments and corporations worked on urgent security updates after the AI-assisted discovery of vulnerabilities in a programming language vital to 70% of websites globally. The two critical flaws identified by AI could allow bad actors to get unauthenticated access to sites which, in theory, could allow them to control a web server remotely without needing a password.

Technology "underpins the automation of physical processes that are vital to modern life," Horne said. That includes, he said, electricity switching, sluice gates in water facilities, dam operations, manufacturing lines, heating and refrigeration controls. The recent cyberattacks on refrigeration systems in French supermarkets showed how easily hostile actors could exploit civilian infrastructure, Horne said.

Advertisement Advertisement

Russia has already attacked critical infrastructure in Europe, according to the U.K. and EU member states that said Russia's FSB security service was behind an attack on Poland's energy grid in December 2025. If the attack had been successful, 500,000 people could have lost electricity during winter.

In the U.K., malicious cyber activity is costing organizations almost $20 billion every year, the NCSC said. In 2025, the carmaker Jaguar Land Rover suffered more than $260 million in direct losses following a cyberattack which also hit suppliers and dented the U.K. economy.

"We need every organization in the UK to take action now. Not only to protect their customers, services and livelihoods, but because cybersecurity is now part of the front line defense of our nation," said Anne Keast-Butler, director of GCHQ, Britain's signals intelligence agency, which the NCSC is part of.

That includes implementing basic but effective cybersecurity measures such as making sure devices are up to date, limiting administrator access to systems and backing up important files and data, the NCSC said.

UK cyber chief says the UK will face AI cyberattacks on critical infrastructure by 2028

The National Cyber Security Centre (NCSC) has issued a stark forecast, suggesting that by 2028, AI-driven malware is highly likely to be utilized in cyberattacks targeting the UK's vital infrastructures, including power stations, airports, and hospitals. This insight comes from the NCSC’s annual report, which indicates an impending threat to critical national infrastructure (CNI) across multiple sectors such as energy, transportation, defense, food, healthcare, and finance.

The NCSC, an entity under the auspices of the GCHQ spy agency, warns that advancements in artificial intelligence will facilitate ransomware attacks aimed at these crucial facilities within the next five years. The report highlights that adversaries, particularly from nations like Russia and China, are expected to integrate AI more routinely into their cyber operations.

According to the NCSC, "It is … highly likely that between now and 2028, AI-powered malware will have been used in a ransomware attack on UK CNI." The rise of ransomware has been a prominent strategy among cybercriminals, particularly those operating from former Soviet states. In such attacks, hackers infiltrate a victim’s systems, encrypt sensitive data using tailored malicious software, and subsequently demand a ransom in cryptocurrency to restore access and data.

The application of AI is poised to elevate this ransomware threat by assisting malicious actors in identifying targets, pinpointing vulnerabilities within IT infrastructures, and crafting sophisticated malware, as noted by the NCSC.

Recently, the UK government advised citizens to stockpile emergency supplies, including up to nine liters of water and three days' worth of food, to prepare for various emergencies, including severe weather and cyberattacks. Defense Secretary Wes Streeting emphasized the importance of preparation against potential state-sponsored threats from Russia and other antagonistic forces aimed at civilian infrastructure.

In his forthcoming speech aligned with the annual review, NCSC chief Richard Horne is set to alert the public to the imminent surge of cyber incidents. He will declare, “In the coming months, we face a real prospect of an unprecedented rise in the number of cyber incidents,” attributing this risk to the ability of AI to reveal long-ignored vulnerabilities in IT frameworks.

A recent instance showcasing AI's capability to uncover security weaknesses was the identification of a critical flaw in PHP, a programming language employed by approximately 70% of global websites. Alan Woodward, a cybersecurity expert from Surrey University, emphasized the significance of the NCSC's warning, advocating for nations and businesses to utilize state-of-the-art AI technologies to mitigate these risks.

“AI can assist in identifying and fixing these vulnerabilities, but without access to these tools, the bad actors will outpace us,” he cautioned. Top AI firms, including OpenAI and Anthropic, have started granting select organizations early access to their advanced models to address serious IT vulnerabilities.

In his address, Horne will also point out that the communications systems utilized by the British military are subjected to ongoing, aggressive cyber intrusions. While Russia has been implicated in these attacks, they have deliberately stopped short of crossing into open conflict, opting instead for targeted campaigns that fall below the threshold of warfare as part of their hybrid warfare strategy.

Horne will reference recent cyber incidents affecting supermarket refrigeration systems in Europe, where hackers manipulated temperatures to exceed 50°C, serving as a stark reminder for businesses to prioritize cybersecurity resilience. “For decades, we have collectively operated with technology that is too inexpensive,” he will remark, underscoring the urgent need for better cybersecurity measures.

AI-driven malware expected to likely target UK infrastructure by 2028.

Generate a cinematic news thumbnail for title:AI-powered malware ‘highly likely’ to be used to hack UK infrastructure by 2028. No text in image.

AI-powered malware ‘highly likely’ to be used to hack UK infrastructure by 2028 - The Guardian

- AI-powered malware is likely to be used in ransomware attacks on UK infrastructure by 2028.  
- NCSC warns that Russia and China will increasingly deploy AI for cyber-attacks.  
- Government urges citizens to prepare for emergencies amid rising cyber-threats and vulnerabilities.

AI-powered malicious software is “highly likely” to be used in hacking attacks against the UK’s power stations, airports and hospitals by 2028, according to the state cybersecurity body.

The National Cyber Security Centre (NCSC) said AI will probably enable a ransomware attack against critical national infrastructure [CNI] between now and 2028. The UK’s definition of CNI covers a range of facilities and services across sectors including energy, transport, defence, food, health and finance.

The warning from the NCSC, part of the GCHQ spy agency, was published in the organisation’s annual review, which also warned that Russia and China will start using AI “routinely” in cyber-attacks.

“It is … highly likely that between now and 2028, AI-powered malware will have been used in a ransomware attack on UK CNI,” said NCSC.

Ransomware has become a cornerstone of the criminal hacker’s repertoire and is a signature ploy of gangs that typically operate out of former Soviet Union states.

After gaining access to a victim’s IT systems, the attackers then encrypt files with bespoke malicious software – or malware – and then demand a payment in cryptocurrency to unlock the crippled systems and return any stolen data. UK ransomware victims include Royal Mail, outsourcing firm Capita and various public bodies, including councils.

AI is set to enhance the ransomware threat by helping attackers with tasks such as scouting targets, locating weaknesses in IT networks and writing malicious code, said the NCSC.

Last month, the government urged Britons to stock up to nine litres of bottled water and three days’ worth of food in order to ensure that households can survive a weather, cyber or other serious emergency. Wes Streeting, the defence secretary, said as the Be Ready campaign was launched that Britain needed to prepare for the threat of “state-sponsored terrorism” by Russia or other hostile powers targeting everyday infrastructure.

NCSC’s chief executive, Richard Horne, will warn the public in a speech accompanying the review on Tuesday that AI is going to unleash a wave of cyber-attacks imminently.

“In the coming months, we face a real prospect of an unprecedented rise in the number of cyber incidents,” he will say.

Horne will attribute the threat to the technology’s ability to uncover weaknesses in IT systems that have been “tolerated for far too long”.

An example of AI’s capacity to root out previously undetected vulnerabilities came last month, when the technology helped discover a critical weakness in PHP, a computer language used by almost 70% of the world’s websites.

Alan Woodward, a professor of cybersecurity at Surrey University, said NCSC’s warning underlined the need for nation states and companies to access cutting-edge AI technology.

skip past newsletter promotion Free newsletter | Every weekday Sign up to First Edition Our morning email breaks down the key stories of the day, telling you what’s happening and why it matters Enter your email Sign up after newsletter promotion

“AI can help people find these flaws and patch them, but if we don’t get the access the bad guys will beat them to it,” he said. So far, leading AI companies like OpenAI and Anthropic have given some companies advanced access to their most powerful models in order to help them identify serious IT glitches.

Horne will also say in his speech that communications networks used by the British military are subjected to daily attacks that are “persistent” and “sustained”.

Horne will say Russia was one of the perpetrators, but the attacks have deliberately not crossed the line into warfare, with Moscow carrying out “notable campaigns” that “intentionally fall below the threshold of armed conflict as they escalate their attempts at hybrid warfare”.

Horne will add that recent cyber-attacks on supermarket fridges in Europe, which have seen fridge temperatures raised by hackers to more than 50C, are a warning to companies and organisations to invest strongly in cyber resilience.

“For decades we have collectively got away with tech on the cheap,” he will say.

AI-powered malware ‘highly likely’ to be used to hack UK infrastructure by 2028

In the future, businesses might be able to simulate robot and AI training in artificially created environments before introducing them to real-world factory settings. However, limitations in Google DeepMind's Genie 4 currently constrain the extent to which companies can apply this innovative approach.

On October 27, Google released the latest iteration of its world model, providing an interactive research preview that allows users to alter generated scenes in real time as narratives unfold. 

Genie 4 enhances the capabilities of its predecessor, Genie 3, by introducing features such as integrated audio and dialogue along with improved physics and more lifelike human movements. Users are empowered to shape their environments using natural language commands to modify elements like weather conditions, characters, or camera angles as scenarios develop.

Researchers can leverage these simulations to investigate the dynamics of natural disasters, such as floods and wildfires, and create outdoor environments to evaluate the responses of autonomous vehicles, drones, and other robotic systems to varying weather patterns and environmental shifts.

A step forward in simulation technology

The world model functions by generating a dynamic interactive environment that reacts frame by frame, corresponding to actions taken by users or AI agents. In partnership with Gemini Robotics, Google integrated Genie 4 with Gemini models that break tasks into smaller steps, monitor the simulated environments, and adapt strategies based on changing conditions.

Chris Selland, the founder and CEO of the AI research and strategy firm Differential Factor, highlighted the significant benefit of Genie 4: it can create a responsive interactive space rather than merely producing a static video from user prompts. 

"This innovation could enable businesses to recreate and evaluate various scenarios without the need to replicate them in reality," he noted.

The value for businesses

Google's most apparent enterprise application lies in the field of robotics. In one demonstration, Genie 4 generated a virtual environment in which a simulated Apptronik’s Apollo humanoid robot played checkers. Such demonstrations showcase how businesses might eventually harness similar environments to simulate robots navigating a range of situations before actual deployment in factories or warehouses. Developers could expose robotic software to diverse layouts, obstacles, and lighting conditions without necessitating the physical creation of each scenario.

“Companies could ultimately utilize generated environments to prototype experiences, craft training modules, and test AI agents in situations that are otherwise costly or hard to recreate in the physical world,” said Manish Jain, CEO of Strategic Horizon Research, an AI advisory firm.

Selland also sees possibilities for rapid prototyping that includes constructing interactive sets while evaluating variables like camera angles, lighting, and spatial arrangements. 

"World models could address a limitation of agentic AI, which has largely functioned through text and APIs: they enable testing the physical repercussions of actions in a simulated space," he explained.

Simulation accuracy challenges

Despite its promise, the existing technology remains in the experimental stage, with Google acknowledging certain limitations. There can be discrepancies in visual details and object placement, especially after intricate prompt modifications or user navigation. The company also highlights difficulties with intricate physical interactions, spatial accuracy, audio fidelity, text legibility, and long-term consistency.

These inaccuracies are critical because consistent training is essential for developers to compare outcomes and understand why systems succeed or fail. If the simulated scenarios don’t accurately mirror the real world, it could become challenging for developers to ascertain whether a robot's performance could be replicated outside of a simulation.

This phenomenon is referred to as the "reality gap" in robotics. A forthcoming review in 2026 will address how variances between simulated and actual environments can adversely affect robot performance when they transition out of simulation.

Selland cautioned that Genie 4 is still an early-stage research preview. "During sessions, details can drift, and the physics involved remain approximative, which makes it unsuitable for precision-oriented engineering applications today."

Jain echoed this sentiment, emphasizing that the crucial concern is whether the results can transfer effectively beyond the simulation environment. "Before enterprises commit to using generated environments for robot training or physical process validation, they need validation that the simulated conditions and outcomes align closely with real-world tasks," he said.

Companies should thoroughly test world models in controlled environments, or "sandboxes," to build trust in their viability for critical functions, according to Selland.

Exploring immediate uses versus long-term potential

Google plans to expand Genie 4's features to Google AI Ultra subscribers through Project Genie and to select early-access partners via the Gemini API and Google AI Studio. This limited rollout currently restricts access for enterprise developers, with Google yet to confirm pricing or specific release dates.

Other organizations are also in the race to develop world models. For instance, Runway launched its interactive world model, GWM Worlds 2, complete with audio in September, while World Labs, founded by AI pioneer Fei-Fei Li, is making strides in the sector as well. The growing interest in simulation-based training and testing is further fueled by advancements in physical AI and robotics.

As multiple vendors vie for attention, enterprises need to discern which developments are ready to act on now.

Jain suggests businesses begin utilizing world models to fabricate virtual scenarios for experimenting and training AI systems. However, applying these environments to train robots or validate their performance in the physical world presents a higher-level challenge that requires more rigorous simulation accuracy.

"Firms should clearly differentiate between utilizing world models for scenario experimentation and depending on them for predicting how robots or physical processes will behave in reality," he advised.

Kinza Yasar covers AI and emerging technologies for TechTarget, focusing on the ethics, governance, and business strategies surrounding enterprise adoption. Her background in IT and network support provides her with a comprehensive perspective on the development, deployment, and management of enterprise technologies.

Google Genie 4 advances world models for robot and AI evaluation.

Generate a cinematic news thumbnail for title:Google Genie 4 brings world models closer to robot and AI testing. No text in image.

Google Genie 4 brings world models closer to robot and AI testing - TechTarget

- Google DeepMind's Genie 4 enables real-time, interactive environment generation for AI testing.  
- Simulation accuracy issues present challenges for effective robotics training and deployment.  
- Enterprises should cautiously experiment with world models before relying on them for real-world applications.

Enterprises could someday rehearse robots and AI agents in generated worlds before risking them on a factory floor, but accuracy gaps in Google DeepMind's Genie 4 limit how far companies can take that approach today.

Released on Oct. 27 as a limited research preview, the updated world model generates interactive settings in real time, letting users change them as a scene unfolds.

Genie 4 builds on its predecessor, Genie 3, adding native audio and dialogue, more realistic physics and more natural human movement. Users can modify environments using natural-language prompts by changing the weather, adding objects or characters and adjusting camera angles as scenes unfold.

Researchers can use these generated settings to explore how natural disasters, including flash floods and wildfires, unfold. They could also create outdoor scenarios to test how autonomous vehicles, drones and other robotic systems respond to changing weather and environmental conditions.

Beyond video generation

A world model generates an interactive world frame by frame and reacts to what users or AI agents do. In a research setup tied to Gemini Robotics, Google paired Genie 4 with Gemini models that break tasks into steps, monitor the simulated scene and adjust plans as conditions change.

Chris Selland, founder and CEO of Differential Factor, an AI research and strategy company, said the important difference is that Genie 4 can create an interactive space that responds to what a person or AI system does, rather than simply generating a video from a prompt.

"It could give businesses a way to create and test different scenarios without having to recreate them in the real world," he said.

Why enterprises should pay attention

Google's clearest enterprise application is robotics. In a demonstration, Google used Genie 4 to generate a setting in which a simulated Apptronik’s Apollo humanoid robot played checkers. The demonstration illustrates how companies could eventually use similar situations to run robots through varied scenarios before deploying them on a factory floor or in a warehouse. Developers could expose robot software to different layouts, obstacles or lighting conditions without having to physically build each scenario.

"Enterprise teams could eventually use generated environments to prototype experiences, create training scenarios and test AI agents in situations that are costly or difficult to reproduce in the real world," said Manish Jain, founder and CEO of Strategic Horizon Research, an AI strategy and advisory firm.

Selland also sees potential in rapid prototyping, including building interactive sets and testing camera angles, lighting and layouts.

"World models could help address a limitation of agentic AI, which has largely operated through text and APIs: the ability to test the physical consequences of an action in a simulated environment," he said.

Simulation accuracy remains a hurdle

The technology is still experimental, and Google acknowledges limitations. Objects and visual details can drift, especially after complex prompt changes or as users navigate an environment. The company also cites challenges with complex physical interactions, spatial accuracy, audio accuracy, legible text and long-term consistency.

Those gaps matter because robot training must be sufficiently consistent for developers to compare results and understand why a system succeeds or fails. If simulated conditions don't accurately reflect the physical world, developers could have difficulty determining whether a robot's performance will transfer to a real-world setting.

This is known as the reality gap in robotics. A 2026 review of the reality gap examines how differences between simulated and physical settings can prevent robots from performing as expected outside a simulation.

Selland cautioned that Genie 4 is still an early research preview. "Sessions can drift, and its physics remain approximate, making it unsuitable today for engineering environments that require precision."

Jain similarly argued that the critical question is whether results transfer beyond the simulation. "Before relying on a generated environment to train robots or validate a physical process, enterprises would need evidence that its conditions and outcomes sufficiently match the real task," he said.

Enterprises should run world models extensively in sandboxes before relying on them for demanding tasks, adding that they are unlikely to trust the models without fully evaluating them internally, Selland said.

Near-term uses vs. longer-term claims

Google plans to make Genie 4 available to Google AI Ultra subscribers through Project Genie, its consumer prototype, and to early-access partners in the Gemini API and Google AI Studio.

The limited rollout still restricts access for enterprise developers. Google has not published pricing or specific release dates.

Other companies are also developing world models. Runway released its playable world model, GWM Worlds 2, with audio, in September, while World Labs, founded by AI pioneer Fei-Fei Li, is also working in the field. The broader push toward physical AI and robotics is also driving interest in simulation-based training and testing.

With several vendors racing to market, the question for buyers is which claims to act on now.

Jain said enterprises could begin by using world models to create virtual scenes for prototyping and training AI systems. But using those environments to train robots or verify how they will perform in the physical world is a more demanding application that requires much greater simulation accuracy.

"Businesses should distinguish between using world models to experiment with scenarios and relying on them to predict how robots or physical processes will behave in the real world," he said.

Kinza Yasar covers AI and emerging technology for TechTarget, with a focus on ethics, enterprise adoption, governance and business strategy. Before moving into journalism, she worked in IT and network support roles, giving her a systems-level perspective on how enterprise technologies are built, deployed and managed.

Google Genie 4 brings world models closer to robot and AI testing

On Monday, major media outlets including the Chicago Tribune, the New York Times, and other associated newspapers initiated federal legal action in Manhattan against Brave Software. They accuse the tech firm of employing advanced scraping technology to illegally extract news content from their websites, which it allegedly then resells to artificial intelligence companies, such as Anthropic and Perplexity.

The complaint, lodged in a Federal Court in Manhattan by the listed news organizations and their affiliated publications like the Daily News and the Orlando Sentinel, characterizes Brave as “the supplier, dealer, and distributor of stolen copyrighted content.” It claims the company provides users—and notable generative AI firms—with access to premium content without authorization, effectively bypassing paywalls to exploit their proprietary news.

The lawsuit states, “Brave operates like a burglar who appropriates publishers' content while also acting as an AI ‘fence’ that sells this stolen material for profit.” It further alleges that these companies knowingly accept the purloined content as they construct their lucrative businesses.

According to the lawsuit, Brave has indexed and replicated content from over 30 billion webpages without notifying website owners or offering a means to block its crawling activities. The accusations extend to Brave’s web browser, which allegedly enables users to access paywalled copyrighted content, a significant revenue source for news outlets striving to produce high-quality content amid a rapidly changing media landscape.

This legal action adds to a series of emerging lawsuits filed nationwide by various news organizations against tech giants, claiming they are unlawfully using copyrighted material to train AI technologies. The plaintiffs, including The News and The Times, are also pursuing legal claims against OpenAI and Microsoft for similar alleged misconduct.

In one specific instance cited in the new complaint, users of Brave’s AI function, called “Ask Brave,” were prompted to access articles published that day by The News. The feature provided titles of 35 paywalled articles and asked whether users wanted access to the full texts, suggesting that it could retrieve complete versions upon request.

“The desperation of AI companies to evade accountability for stealing news content is evident, and Brave Software is providing them with a backdoor to journalism that they have no legal claim to. Purchasing stolen content doesn’t legitimize it. Brave must be held accountable for facilitating this piracy in the AI sector,” asserted Steven Lieberman, the attorney representing The News and its Tribune Publishing affiliates.

The lawsuit outlines five legal allegations, claiming that Brave has engaged in direct, contributory, and vicarious copyright infringement.

Frank Pine, Executive Editor at Tribune Publishing and MediaNews Group, criticized Brave for undermining intellectual property laws. He stated, “Brave, ostensibly advocating for its users and other well-known AI firms, exploits the internet and markets its ill-gotten content to anyone willing to pay. They boast about protecting privacy—though it may actually reflect their commitment to piracy—while attempting to erase evidence of their wrongdoing through their supposed ‘Zero Data Retention’ policy.”

Efforts to obtain commentary from Brave Software, Inc., were made, although Anthropic and other AI firms implicated in acquiring purportedly misappropriated content from Brave are not named as defendants in this latest lawsuit.

Brave accused of stealing Chicago Tribune's content and selling it to A.I. companies in copyright lawsuit.

Generate a cinematic news thumbnail for title:Tech company Brave plundered Chicago Tribune’s content and resold it to A.I. giants: copyright lawsuit. No text in image.

Tech company Brave plundered Chicago Tribune’s content and resold it to A.I. giants: copyright lawsuit - Chicago Tribune

- Chicago Tribune and others sued Brave Software for illegally scraping news content online.  
- Brave allegedly provides access to paywalled articles and sells content to A.I. companies.  
- Lawsuit claims Brave committed copyright infringement and undermined news organizations' revenue sources.  

The Chicago Tribune, the New York Times, and affiliated newspapers filed federal lawsuits in Manhattan Monday accusing the technology company Brave Software of using high-tech scraping tools to illegally plunder their websites for exclusive news stories to resell to A.I. giants, including Anthropic and Perplexity.

A Manhattan Federal Court complaint, filed by the News and affiliated Tribune Publishing newspapers the Daily News and the Orlando Sentinel, says Brave acts as “the supplier, dealer, and distributor of stolen copyrighted content” and provides a back door circumventing paywalls to its own users as well as prominent generative A.I. companies to help build their businesses.

“Brave acts both like a burglar that misappropriates the content of publishers, and an AI ‘fence’ that sells stolen content to third parties,” the lawsuit alleges. “And the third parties are serving as the knowing receivers of that stolen content, which they have used to build their trillion dollar businesses.”

The company sells copies of the publishers’ content to third parties for training their A.I. models and products and informing responses to users, the suit alleges, in addition to making unauthorized copies of copyrighted works through its own browser, providing free access to paywalled content.

The lawsuit says Brave crawled and copied content from more than 30 billion webpages in its index without identifying itself or providing the owners of websites any mechanism to block its crawler, according to The News’s complaint.

In addition to selling stolen stories to other companies, Brave also has a web browser that provides paywalled copyrighted content to its users, according to the suits. Paywalls provide a critical source of revenue to news organizations as they seek to compete in an evolving news environment by producing quality, original content.

The suits add to a growing pile of cases playing out across the country brought by news organizations and various owners of copyrighted content against tech giants alleging the wholesale theft of content used to train A.I. models. The News, The Times and other papers that filed suit Monday are also suing OpenAI and Microsoft, alleging similar conduct.

In one example detailed in the new lawsuit, a user prompted Brave’s A.I. feature known as “Ask Brave” to show them articles published that day in The News. The chatbot provided the titles of 35 articles that appeared in The News that day behind a paywall and asked, “would you like me to pull up the full text of any of these articles? If so, just tell me which one(s) and I’ll fetch the complete version.”

“AI giants are now desperate to avoid getting caught red-handed stealing news from publishers, so Brave Software is selling them a back door to the journalism they know they have no right to. Buying stolen goods secondhand doesn’t make them legitimate. Brave must be held responsible for laundering pirated reporting for the AI industry,” said Steven Lieberman, an attorney representing The News and its sister papers with Tribune Publishing in the suit.

The suit lists five legal claims, alleging Brave has committed direct, contributory, and vicarious copyright infringement.

In a statement, Frank Pine, the executive editor at Tribune Publishing and MediaNews Group, said the software company had made “a mockery of our intellectual property laws.”

“Brave, purportedly on behalf of its users but also on behalf of other more well-known AI companies, pillages the internet and resells its ill-gotten gains to anyone who will buy,” Pine said. “Then, proudly touting how it protects the privacy – read that as piracy – of its customers, it attempts to delete evidence of its misdeeds under its ‘Zero Data Retention’ policy.”

The News reached out to Brave Software, Inc., for comment. Anthropic and other A.I. companies accused of buying stolen content from Brave are not named as defendants in Monday’s lawsuits.

Tech company Brave plundered Chicago Tribune’s content and resold it to A.I. giants: copyright lawsuit

You are about to exit Barron's website.

By clicking the “Proceed” button below, you will be taken to a third-party site managed by Hong Kong Tiimoot Information Technology Co., Limited (HKT), based in Hong Kong. This external site operates independently of Barron's, and we do not oversee its content or operations. It's important to note that HKT's privacy procedures are governed by its own Privacy Statement, which we recommend you review carefully. Barron's assumes no responsibility for HKT's data privacy or other related practices.

Tech Downturn Intensifies: Increased AI Investments by Major Tech Firms and Higher Capital Costs Impact Stock Market

Generate a cinematic news thumbnail for title:Tech Correction Deepens: Big Tech AI Spending and Rising Capital Costs Shake Stock Market. No text in image.

Tech Correction Deepens: Big Tech AI Spending and Rising Capital Costs Shake Stock Market - Barron's

- Users will be redirected to a third-party website operated by HKT in Hong Kong.  
- Barron's does not control the content or operations of the third-party website.  
- Users should review HKT's Privacy Statement for its data privacy practices.  

You are now leaving Barron's website

By clicking on the “Proceed” button below, you will be redirected to a third-party website owned and operated by Hong Kong Tiimoot Information Technology Co., Limited. (“HKT”), which is located in Hong Kong. That website operates independently from Barron's and Barron's does not control the website. The privacy practices of HKT are subject to its Privacy Statement, so please read it closely. We are not responsible for HKT's privacy or other data-related practices.

Tech Correction Deepens: Big Tech AI Spending and Rising Capital Costs Shake Stock Market

The entertainment industry is known for its specific hotspots, places that have consistently served as the preferred gathering spots for those in the know. An A-list agent recently remarked during lunch, “There are five main places to be in this town,” identifying San Vicente Bungalows, Tower Bar, the Chateau, Polo Lounge, and Toscana as the go-to locations. Classic names like the Ivy, the Grill, Craig’s, and La Scala also come to mind as past or present staples.

While Culver City hosts major players like Amazon MGM, Sony, Apple, HBO, and TikTok, it hasn't typically been recognized as a prime dining destination. However, it is swiftly evolving into a focal point for AI startups eager to carve out their niche in Hollywood; the Laurel Grill has risen as the area’s premier spot for power lunches among local studio and tech executives.

Among those taking office space in Culver City is Sean Parker, the former president of Facebook, highlighting Silicon Valley's relentless march into the entertainment sector. 

In the past year, several AI startups have planted their roots in this compact 1.2-square-mile city, aiming to leverage what Dan O'Brien, a council member and former mayor, describes as a commercial real estate market that has just begun to recover.

George Strompolos, founder of Promise AI, notes, “Culver City is becoming a nexus for artists looking to collaborate and connect with studios.” He, along with former YouTube executive Jamie Byrne and filmmaker Dave Clark, has set up their startup right beneath the iconic Sony water tower, enjoying backing from Peter Chernin's North Road and Andreessen Horowitz.

Strompolos frequently mentions that the Laurel Grill is his preferred venue for meetings, stating, “Whenever I have a meeting with an Amazon executive, it’s at the Laurel Grill.”

Situated atop the Culver Steps and overlooking the historic Culver Studios and the Roaring '20s Culver Hotel, the Laurel Grill has been operating for two years, founded by Dean McKillen and Phil Howard, known for popular venues like Laurel Hardware and Ysabel.

A local studio executive mentions the lack of suitable business lunch venues in downtown Culver, which the Laurel Grill has successfully addressed with its inviting atmosphere and delightful offerings, including highly praised cheesy bread. A bartender confirmed that the restaurant often sees a crowd of executives from Amazon, Apple, and Sony.

While not yet on par with the legendary status of The Ivy — a favored lunch locale for countless Hollywood legends — the Laurel Grill has certainly made its mark. Notably, breakout actor Connor Storrie worked as a waiter there just prior to landing a role in the HBO Max hit “Heated Rivalry.”

Recently, power player Sean Parker opened a new office for his San Francisco-based Cantina Labs in the Hayden Tract, expanding his influence into the area. He also played a significant role in a recent $80 million investment in AI firm Stability AI.

Cantina Labs is described as a social AI startup boasting an advanced AI character creator capable of conversation and performance. Additionally, Luma Labs, co-founded by former Apple engineer Amit Jain and valued at $4 billion, established a substantial facility called Dream Lab in Culver City last year, with backing from major players like Andreessen Horowitz and NVIDIA.

Luma is making waves in the entertainment sector with its innovative video and image generation models, attracting interest from firms including Amazon and Jon Erwin’s Wonder Project.

Other startups dipping their toes into the Culver City waters this year include TrueShort, a microseries studio; PhantomX, an AI production house; and Native Foreign, a cutting-edge AI advertising agency.

The drive behind this AI movement in Culver City is often credited to Mike Gioia and Todd Terrazas, founders of AI on the Lot, a conference that has recently made its home at Amazon MGM’s Culver Studios. The conference has also held job fairs and includes facilities for AI filmmakers located strategically near the Culver Theater that Amazon acquired in 2022.

During LA Tech Week, the group organized an AI job fair at the Culver Steps. “To attract them, you need to create an open environment,” O’Brien explains. “And that’s what we’ve been working toward.”

Yet, not all members of the local council share the same enthusiasm. Some officials navigate the delicate balance of embracing the burgeoning AI community while remaining sensitive to the creative sector, which often harbors reservations about these new technologies.

Culver City's AI surge draws Silicon Valley companies and transforms a two-year-old restaurant into Hollywood's new dining hotspot.

Generate a cinematic news thumbnail for title:Culver City’s AI boom attracting Silicon Valley firms and making 2-year-old restaurant Hollywood’s latest power lunch spot. No text in image.

Culver City's AI boom attracting Silicon Valley firms and making 2-year-old restaurant Hollywood's latest power lunch spot - Page Six

- The Laurel Grill has quickly become the main power lunch spot in Culver City.  
- AI startups are flocking to Culver City, taking advantage of affordable commercial real estate.  
- Local officials recognize the need to balance AI growth with creative community concerns.  

Trends may come and go, but the entertainment business has anointed just a few particular power spots as acceptable watering holes over the years.

“There’s five places to go in this town,” an A-list agent recently exclaimed over lunch — naming San Vicente Bungalows, Tower Bar, the Chateau, Polo Lounge and Toscana as acceptable options. (Once iconic or of-the-moment names also, of course, include the Ivy, the Grill, Craig’s and La Scala.)

While Culver City is home to Amazon MGM, Sony, Apple, HBO and TikTok — it hasn’t exactly been known as a power dining destination. But these days, the area is quickly becoming a hub for AI startups looking to break into Hollywood — and the more down-to-earth Laurel Grill has emerged as the de facto power lunch spot for the neighborhood’s studio and AI execs.

5 The Laurel Grill has become the de facto power lunch spot in the area. laurelgrill/Instagram

Some of Silicon Valley’s biggest names — including billionaire former Facebook president Sean Parker — are gobbling up office space in Culver City as part of tech’s ongoing push to colonize the entertainment industry.

Over the past year, more than a half-dozen AI startups have moved into the 1.2-square-mile city — taking advantage of a commercial real estate market that, as council member and former mayor Dan O’Brien, tells Page Six Hollywood, “bottomed out last year.”

“Culver City is becoming a hotbed of artists who are finding work, connecting with each other and pitching themselves to studios,” says George Strompolos, who set-up the startup Promise AI in a building which is in the shadow of the Sony water tower.

He launched his company alongside former YouTube exec Jamie Byrne and filmmaker Dave Clark with backing from Peter Chernin’s North Road and Andreessen Horowitz. (The exec previously ran the digital media and entertainment company Fullscreen — also from offices in Culver Studios — but sold that company to AT&T’s Otter Media.)

Strompolos tells us the Laurel Grill has become one of his go-to working lunch spots. “Every time I’m having a meeting with an Amazon exec it’s at the Laurel Grill.”

The restaurant sits atop the Culver Steps, overlooking the Culver Studios and the Roaring ‘20s-era Culver Hotel. It was opened two years ago by Dean McKillen and Phil Howard, the owners of West Hollywood haunts Laurel Hardware and Ysabel.

A studio exec in the area points out, “There’s not a lot in downtown Culver in terms of places to have business lunches, so [Laurel Grill] filled that gap — plus it’s a good vibe and they have a nice patio. A+ cheesy bread.” A bartender at the spot told us it’s frequently packed with Amazon, Apple and Sony execs.

5 Former Fullscreen founder George Strompolos is back in Culver with a new AI startup. Jordan Strauss / Shutterstock

Local pol O’Brien likens the place to Culver City’s version of The Ivy, adding, “There’s all these deals being made between agents.”

(The Ivy is the Robertson Blvd. lunch spot frequented over the years by studio execs and agents plus Tom Cruise, Brad Pitt, Nicole Kidman, Kevin Costner, Kate Moss, Madonna and various Kardashians.)

And while the Laurel Grill hasn’t hit anywhere near that level of stardom quite yet, breakout actor Connor Storrie worked as a waiter at the eatery right up until he was cast to star in the HBO Max smash, “Heated Rivalry.”

5 The Culver Steps, with the Laurel Grill at the top, has become a center of gravity in the area. UCG/Universal Images Group via Getty Images

Among the power players in the area, Parker, a co-founder of Napster, opened an office in the Hayden Tract last month as part of the expansion of his San Francisco-based Cantina Labs which he founded in 2023. Parker also took part in an $80 million rescue of troubled AI image firm Stability AI.

(Not much is known about Cantina Labs, except that it’s a social AI startup that boasts having “the most advanced AI character creator” that can “talk, perform and interact with anyone,” per the company’s description.)

There’s also Luma Labs, a $4 billion Palo Alto-based startup co-founded by Amit Jain, which opened up a 7,000-square-foot facility, the Dream Lab, in Culver City last year. Jain is a former Apple engineer and is backed by Silicon Valley heavyweights Andreessen Horowitz, NVIDIA and Amazon.

5 Sony Pictures is a longtime Culver City resident. Getty Images

Luma has a suite of video and image generation models that have been gaining traction within Hollywood, including at Amazon and Jon Erwin’s Wonder Project, which is behind series including “House of David” and “Moses.”

Other AI startups that have set up shop in Culver City this year include microseries studio TrueShort, PhantomX, an AI production film studio, and Native Foreign, an AI ad agency.

Most people credit Mike Gioia and Todd Terrazas, who founded AI on the Lot — the four-year-old conference that for the past two years has taken over Amazon MGM’s Culver Studios — for spearheading the AI push into Culver City. (The company also has its own office space sandwiched between the studio gate and the Culver Theater, which Amazon took over in 2022.)

5 The AI on the Lot conference was held at Amazon MGM’s Culver Studios in the spring. Irina Logra

Last month during LA Tech Week, the group even held an AI job fair on the Culver Steps and routinely offers up its office space to AI filmmakers. Earlier this year, the Thunderbird School of Global Management at Arizona State University held its LA chapter meetup at Laurel Grill.

“If you want them to come to you, you need to tell them that they’re welcome,” said O’Brien of the AI boom in the area. “And we’ve been doing that.”

However, another city insider tells us that the feeling is not totally unanimous among council members — as local pols walk a tightrope of welcoming the AI crowd and being sensitive to the creative community that’s wary of the technology.

Culver City’s AI boom attracting Silicon Valley firms and making 2-year-old restaurant Hollywood’s latest power lunch spot

In an unexpected turn within the advertising landscape, a distinctive victor has emerged amidst the industry's push for efficiency driven by artificial intelligence: meticulously crafted creative content. 

As brands strive to differentiate themselves and respond to the growing backlash against AI-generated visuals, marketers are gravitating towards more tactile methods, including claymation and hand-drawn animation, as reported by Katie Deighton from the WSJ Leadership Institute. 

Today will see the launch of two holiday campaigns that embrace this analog aesthetic. One features e.l.f. Beauty, showcasing puppets adorned with actual cosmetics, while the other from Levi's draws inspiration from horror films and revisits the brand's memorable 1995 “Clayman” advertisement through stop-motion techniques. 

Yet, the trend extends beyond festive cheer. Apple has also been integrating traditional effects and human skill into its marketing strategies this year. Notably, they have created a fresh visual identity for Apple Music by passing sound waves through mineral oil, exemplifying a commitment to authentic, artisanal creativity.

AI is speeding ahead, prompting the ad industry to embrace stop-motion techniques.

Generate a cinematic news thumbnail for title:AI Is Stuck on Fast-Forward. The Ad Industry Is Turning to Stop Motion.. No text in image.

AI Is Stuck on Fast-Forward. The Ad Industry Is Turning to Stop Motion. - WSJ

- Brands are prioritizing handcrafted creativity in response to AI-generated imagery backlash.  
- New holiday campaigns feature claymation and puppetry for a unique marketing approach.  
- Apple employs practical effects in advertising to emphasize human craftsmanship and authenticity.  

Welcome back. A surprising winner has emerged amid the ad industry’s AI-fueled drive for efficiency: almost ostentatiously handcrafted creative.

Brands’ desire to stand out and backlash against AI imagery are driving marketers toward high-touch techniques like claymation and hand-drawn animation, filmmakers tell the WSJ Leadership Institute’s Katie Deighton.

Today alone will bring two new holiday campaigns with an analog approach: an e.l.f. Beauty push starring puppets wearing real makeup and a Levi’s stop-motion spot inspired by horror movies and the brand’s 1995 “Clayman” ad.

But it’s not all holiday whimsy. Apple has been using practical effects and human craft in its marketing this year, like running sound waves through mineral oil to create a new visual identity for Apple Music.

AI Is Stuck on Fast-Forward. The Ad Industry Is Turning to Stop Motion.

Marvell Technology announced on Monday its intention to acquire Lumentum, a manufacturer of optical components, to cater to the increasing connectivity needs stemming from the expansion of artificial intelligence infrastructure.

Matt Murphy, Chairman and CEO of Marvell, emphasized the significance of this acquisition, stating that "AI infrastructure is undergoing a fundamental architectural shift towards optical connectivity, creating a once-in-a-generation opportunity." He further noted that those who take the initiative in this transition will establish enduring standards for the future.

The acquisition agreement is structured as a fixed ratio deal with no cash involved, meaning its valuation hinges on Marvell's stock price. Marvell indicated that the deal carries an implied enterprise value of $115 billion for Lumentum.

With this merger, Marvell aims to integrate electrical and optical technologies to respond to the rapidly evolving AI connectivity requirements across various regions and platforms. As companies increasingly leverage optical connections for the interlinking of chips responsible for managing and training large language models, this acquisition becomes critical.

Under the terms of the agreement, Lumentum shareholders will receive 4.5 shares of Marvell for each share they hold in Lumentum. The acquisition is anticipated to finalize in the first half of 2028.

This move represents one of the most substantial investments in AI for Marvell this year, occurring despite growing concerns regarding the long-term viability of infrastructure expenditures. 

Moreover, Marvell has previously invested heavily in the AI sector, recently finalizing a multibillion-dollar acquisition of Celestial AI, a startup that specializes in optical interconnectivity technology.

The company has experienced impressive growth, with its market value surging more than threefold over the past three years, fueled by strong interest in AI. Similarly, Lumentum's stock has seen a more than twofold increase this year alone.

Prior to the opening bell, Marvell's shares fell by 3%, while Lumentum’s stock was down approximately 4%.

Marvell plans to acquire Lumentum in an all-stock transaction, significantly investing in AI connectivity.

Generate a cinematic news thumbnail for title:Marvell to buy Lumentum in all-stock deal, a massive bet on AI connectivity. No text in image.

Marvell to buy Lumentum in all-stock deal, a massive bet on AI connectivity - CNBC

- Marvell Technology will acquire Lumentum to enhance AI infrastructure connectivity and optical technology.  
- The acquisition has an implied enterprise value of $115 billion for Lumentum shares.  
- Marvell's deal reflects confidence in AI growth despite infrastructure spending concerns.  

Marvell Technology on Monday said it will buy optical components maker Lumentum to address growing connectivity demands from the artificial intelligence infrastructure buildout.

"AI infrastructure is undergoing a fundamental architectural shift toward optical connectivity, creating an opportunity that comes along once in a generation," said Marvell Chairman and CEO Matt Murphy in a release. "The companies that move first and help define this transition will set the standard for decades to come."

Since the tie-up is a fixed ratio deal with no cash, the value will depend on the price of Marvell shares. The company said the transaction has an implied enterprise value of $115 billion for Lumentum, in a release.

Through the acquisition, Marvell said it can combine electrical and optical technologies to meet accelerating and shifting AI connectivity demands across regions and systems. Companies are increasingly relying on optical connections to piece together chips used to run and train large language models.

Lumentum shareholders will receive 4.5 Marvell shares for each Lumentum share under the agreement. The deal is expected to close during the first half of 2028.

Marvell's takeover marks one of the most significant bets on AI in 2026 and comes despite recent concerns over the long-term sustainability of infrastructure spending.

This isn't the company's first bet on AI growth. This year, the company closed a multibillion-dollar deal for Celestial AI, a startup focused on optical interconnectivity technology.

Marvell has seen its market value more than triple over the last three years, buoyed by unrelenting AI enthusiasm. Lumentum's stock has more than doubled in 2026 alone.

Marvell shares slumped 3%, while Lumentum's stock declined about 4% before the bell.

Marvell to buy Lumentum in all-stock deal, a massive bet on AI connectivity

AI is increasingly becoming an integral part of our lives, raising important questions about the future of work and how we can navigate this new landscape.

Many people facing uncertainty in their jobs—or concerning their children’s futures—are often advised to become AI specialists or pursue careers that seem immune to automation, like trades or healthcare. However, Ethan Mollick, a renowned professor at Wharton, believes this perspective is overly simplistic.

In his recently published book, "Co-Existence," Mollick emphasizes that advancements in AI are progressing at a pace that many individuals fail to comprehend. He argues that the best approach to thrive in an AI-driven world is to learn to coexist with these technologies rather than resist them.

A crucial part of this adjustment is recognizing that certain skills may become less relevant, while highlighting unique human traits that AI cannot replicate. I had a chance to discuss these themes with Mollick during a recent conversation on my podcast.

When I asked Mollick if acquiring AI expertise or seeking "AI-proof" jobs was good advice, he expressed skepticism. He likened it to a temporary solution, suggesting that even manual jobs could eventually be replaced by advanced automation. The evolution of coding illustrates this shift—while the nature of coding work is changing drastically, it remains uncertain whether this will lead to increased or decreased job opportunities in the field.

In his book, Mollick identifies four essential traits that can provide a competitive edge: deep knowledge, wide knowledge, discernment (or taste), and personal agency. For instance, today's coding landscape demands not just the ability to write code, but a vision of the product one aims to create. This shift means that job satisfaction and workplace dynamics for coders are changing, with some embracing the transformation while others struggle.

Mollick highlights two significant challenges: the individual’s tendency to offload cognitive tasks and the existing educational frameworks meant to prepare individuals for work. While offloading mundane tasks is a human instinct—just think of how we value conveniences like dishwashers or calculators—it becomes problematic when it leads to reliance on AI for tasks that require critical thinking and creativity. 

As for education, Mollick argues that while traditional schools are currently flawed, AI could enhance learning experiences by providing tailored tutoring and promoting active participation in classrooms. However, he expresses greater concern over apprenticeship systems. With the reliance on AI, there’s a risk that junior employees may underutilize their opportunities for growth, opting to depend on AI to get things done rather than honing their skills and taste.

This reliance creates a cycle where middle managers may bypass less experienced employees altogether, preferring AI solutions that seem more efficient. As a result, the classic mentoring relationships that cultivate talent and ensure a skilled workforce are in jeopardy.

To counter this shift, Mollick suggests that companies must either evolve by making education more practical, akin to vocational training, or commit to mentorship programs that genuinely cultivate young talent. Without these structures in place, businesses may face a critical shortage of skilled workers in the future as they opt for seasoned professionals over nurturing the next generation.

Wharton Professor and AI Specialist Ethan Mollick Discusses the Future of Work

Generate a cinematic news thumbnail for title:AI Expert and Wharton Professor Ethan Mollick on the Future of Work. No text in image.

You can't escape AI, but you can learn to live with it. Wharton Professor Ethan Mollick explains how. - Business Insider

- AI's rapid development requires adapting current skills rather than seeking AI-proof careers.  
- Deep knowledge, wide knowledge, taste, and agency can help workers excel alongside AI.  
- The mentorship model in workplaces is weakening due to reliance on AI over interns.  

You can't make AI go away. How can you live with it?

Or more precisely: How can you make a living in an AI world?

Here are a couple standard answers: "Learn to use AI." Or figure out an AI-proof career, one that can't be automated away. Maybe something with your hands.

Ethan Mollick doesn't think that will work.

Mollick is a Wharton management professor who has become one of the more influential guides to what AI can already do — and what it will mean for people who have to work alongside it. His new book, "Co-Existence," argues that the models are improving much faster than most people understand, and the only way to deal with their rapid acceleration is to learn how to live with them.

Related video How AI could destroy — or save — humanity, according to AI experts

That means acknowledging that a lot of what you can do today will be less useful in the future, and concentrating on the attributes you already have that AI can't replicate.

I talked to Mollick about what AI means for your job, your kids' future, and what happens when the systems that train workers start to break down on my Channels podcast. The following is an edited excerpt of our conversation.

Peter Kafka: People worried about their jobs, or their kids' jobs, are being told to become AI experts — or to find lines of work that are AI-proof, like plumbing or nursing. Is that useful?

Ethan Mollick: I don't think it's very useful. First of all, that's great until PlumBot 6000 comes along. I'm not a robotics expert, but the robotics experts I know went from, "LLMs will never … this AI boom is completely unrelated … robots are hard" to, "Actually there's a thing here."

I don't know what the future holds. I don't think anyone does, which is either comforting or very upsetting to hear from somebody who's an AI expert.

Look at the most changed field, which is coding. We already know coding is changing. [But] we aren't even clear whether there are going to be more or less jobs in coding. We know what you do as a coder is going to change and keep changing.

Business Insider has Games! Want more Business Insider in your news feed? Add BI in Google so our reporting is easier to find when you’re searching for what matters. Add to Preferred Sources

Writing error-free code is now less important than conceiving of what you want to have built. A lot of coders are happier than they've ever been before. Some are sadder. These are evolving jobs.

In the book, I talk about four things that give you at least some advantage. Deep knowledge: knowing an area really well, because that expertise lets you outcompete the AI. Wide knowledge: knowing about lots of different things.

One thing I see in companies is they say, "Our coders are doing design work, our designers are doing coder work, and the marketers are doing business plans." The more you know about a lot of little things, the more AI can help you.

Then there's taste. AI produces a lot. What prevents it from being slop? Your taste about which things to pick.

And finally, agency. I don't think "learning to use AI" means very much right now. We've been testing all the methods we use to teach people how to use AI. None of those tricks work anymore.

Prompting has started to look a lot more like managing. You don't have to do magic words anymore to make the AI do something. Instead, you specify: Here's the product I want, here's how I know it's good, here's how you should test it, here's what you show me partway. Give me five different mockups that I can pick among.

That seems like good news for someone mid-career, who has experience, taste, and has gotten good at their job. But if you're starting out, how do you develop that stuff in a world where AI says, "I'll do the research. I'll write it for you."

There are two problems. One is people as individuals, and the other is the institutions we have for teaching people.

As individuals, we love to cognitively offload stuff. That's what we're built to do. Technology has always been about replacing annoying labor. No one pines for washing dishes by hand once they have a dishwasher.

No one wants to churn their own butter.

My kids don't learn cursive. I don't know how to use a slide rule.

But we're bad at picking which things we want to offload. That's a problem with everyone cheating on homework.

And then there are the institutions. There are two sets of learning institutions for learning how to do a job. One is college and high school education. They're screwed up right now, but we'll figure that out. AI is actually a really good tutor, and we know how to test people without AI: You put them in a classroom with a blue book or have a conversation. We'll have better education as a result, with AI tutoring outside of class and active learning inside of class.

What I'm worried about is the other method, which is apprenticeship.

I send people to a bank, consulting company, audit firm, or car dealership, and they learn the same way everyone always has.

They do grunt work, poorly paid. They might get abused a little bit. The trade is that they learn skills and contacts, and they learn how to do stuff because we force them to do it.

And middle managers get somebody to do their crap work. Even if you're a bad mentor, people still learn the ropes. If they show they can do a good job, they advance. You get the next generation of good employees. It has worked out wonderfully for a few thousand years.

That all broke because the incentives are no longer aligned. As a junior employee, your best bet is to use AI for everything because it's better than you. Junior people are much worse at using AI because they dump everything into Claude and give you Claude's results. How could they tell you whether it's a good report or bad? They haven't built taste.

And managers realize: "Why would I go to an intern when the AI will solve the problem for me, at a better rate?" And company leaders say: "Why would I want to slow this train down? Why would I want worse work from interns?"

The training mechanism was this virtuous cycle, and it is completely shattered.

Either we will have to make more formal education happen — a lot of light, white-collar work will have to look more like welding or plumbing, with a trade-school aspect—or companies are going to have to get serious about mentorship as a real thing they do. It takes time and costs productivity, but you gain the benefit of having junior people.

My fear is that companies will say, "We'll just hire senior people." That's the dangerous area. The talent pipeline ends. There are no replacements.

Y Combinator's Garry Tan spends as much as $200K annually on tokens due to his heavy use of AI agents.

More from AI

Ranbir Kapoor discusses AI, cinema, and being a father.

UK cybersecurity leader warns of AI-driven cyberattacks on essential services by 2028.

AI-driven malware expected to likely target UK infrastructure by 2028.

Google Genie 4 advances world models for robot and AI evaluation.