More than 100 people lost their lives in July when flash floods ravaged Central Texas, illustrating the devastation that’s possible when mounting natural disasters are met with inefficient government responses. Agencies operated in complete silos as the floods quickly destroyed communities — county emergency systems couldn’t share real-time data with state coordinators while rescue teams worked from incompatible dispatch networks — resulting in desolation that was completely amplified by the failure of civic communications technology.

Today, our cities and towns operate like 18th-century mansions rewired with modern gadgets: functional in calm weather, yet lethal in storms. And as climate disasters intensify, cyber warfare evolves and AI-powered threats emerge, the “smart city” model that promised efficiency through sensors and dashboards has proven dangerously inadequate in recent years.

The path forward is not more tech — it’s new architecture. We need cities that don’t just collect data, but process that data in real time to adapt, evolve and take action as unified living systems. They can’t just be “smart,” but instead must be “omni” cities: urban ecosystems with integrated AI nervous systems that coordinate every element of civic life with both precision and purpose.

THE FRAGMENTATION CRISIS

Today’s cities are digital archipelagos. A drone from one vendor can’t share data with a robot from another, while emergency systems speak different languages. This is not inefficiency, but a structural failure seen across virtually every U.S. metropolis.

This fragmentation has been our downfall during the greatest of tragedies. When Hurricane Helene struck in 2024, outdoor sirens stayed silent while cell networks collapsed — not because the technology failed, but because nothing was designed to work together. Several months later when California wildfires forced 200,000 evacuations earlier this year, communication breakdowns and uncoordinated shelters led to 31 preventable deaths.

Yet these communication issues are not isolated to climate disasters. In fact, they’re happening in small ways every day, from 911 dispatch failures to fractured public transport services. And as time passes, new threats are emerging faster than cities can adapt, from AI-powered cyber attacks targeting infrastructure vulnerabilities, to weaponized drone swarms exploiting communication gaps. Our fragmented civic networks don’t just lag behind these challenges — they amplify them.

BEYOND ‘SMART’: THE OMNI CITY VISION

Just as smart cities promised efficiency in the 2010s, omni cities will deliver the resilience needed to face the looming threats of the next decade.

As suggested by its name, an omni city operates as a unified organism in which every component shares a common protocol for crisis response. When a wildfire approaches an omni city, the city doesn’t just sound alarms — it automatically reroutes traffic, opens shelters and coordinates evacuation routes while emergency teams receive real-time data from every connected system.

This isn’t science fiction; cities like Houston are already testing integrated frameworks that link climate response, public safety and mobility systems. The difference of an omni city, however, lies in treating cities as ecosystems rather than collections of isolated smart devices.

The key is interoperability — not just between machines, but between machines and humans. This requires a city-scale operating system that allows autonomous tools, public responders and ethical protocols to work in lockstep. Instead of retrofitting isolated apps, this OS treats cities like systems, linking drones, sensors, robotics, transport and emergency teams through a unified protocol layer.

When cities become more intelligent, they must also become more accountable, particularly in regards to equity. The smart city movement failed in part because it prioritized convenience for the wealthy (or those who can access technology in the first place) over resilience for everyone. Omni cities must flip this model by prioritizing resilience testing in the places most vulnerable to system failures, with the communities that legacy infrastructure has consistently ignored. This isn’t charity, it’s engineering. Systems that can’t serve everyone can’t truly serve anyone.

THE MOMENT FOR ACTION

While exploring the next iteration of a “smart city” may feel daunting, local governments don’t need federal permission to begin. They can start by requiring interoperability standards for new public technology, creating transparent audit systems for automated decisions and prioritizing deployments in underserved communities. The goal isn’t to build perfect cities overnight, but to create the foundations for urban systems that can evolve with emerging challenges.

The era of omni cities begins with recognizing that in a world of cascading crises, our urban infrastructure must become our first responder. Cities that understand this won’t just survive — they’ll define what governance means in the age of artificial intelligence.

When infrastructure thinks as fast as threats emerge, resilience becomes possible. The question is not whether cities will evolve, but which ones will evolve first.

Cesar R. Hernandez is an Equity Fellow in the Center for Public Leadership at Harvard Kennedy School.

Scale AI, which helps tech companies prepare data to train their AI models, filed a lawsuit against one of its former sales employees and its rival Mercor on Wednesday. The suit claims the employee, who was hired by Mercor, “stole more than 100 confidential documents concerning Scale’s customer strategies and other proprietary information,” according to a copy seen by TechCrunch.

Scale is suing Mercor for misappropriation of trade secrets and is suing the former employee, Eugene Ling, for breach of contract. The suit also claims the employee was trying to pitch Mercor to one of Scale’s largest customers before he officially left his former job. The suit calls this company “Customer A.”

Mercor co-founder Surya Midha denies that his company used any data from Scale, although he admits that Ling may have been in possession of some.

“While Mercor has hired many people who departed Scale, we have no interest in any of Scale’s trade secrets and in fact are intentionally running our business in a different way. Eugene informed us that he had old documents in a personal Google Drive, which we have never accessed and are now investigating,” Midha told TechCrunch in an emailed statement. 

“We reached out to Scale six days ago offering to have Eugene destroy the files or reach a different resolution, and we are now awaiting their response,” Midha said.

Scale alleges that these documents contained the specific data that would allow Mercor to serve Customer A, as well as several other of Scale’s most important clients.

Scale wanted Mercor to give it a full list of the files in the drive, and to prevent Ling from working with Customer A. It alleges in the suit that Mercor refused. Ling did not immediately respond to TechCrunch’s request for comment.

There are scant clues in the suit about the identity of Customer A. The suit does say that if Scale’s rival did win this customer away, it would be a contract “worth millions of dollars to Mercor.”

Whatever the details of this suit, it does show one thing: Scale is clearly concerned enough about the threat of Mercor to pursue legal action. As TechCrunch previously reported, even with Meta’s multibillion-dollar investment into Scale, TBD Labs — the core unit within Meta tasked with building AI superintelligence — is still using Mercor and other LLM data training service providers.

Mercor is rising in the LLM training arena because it is known for hiring content specialists, often PhDs, to train LLM data in their areas of expertise.

In June, Scale announced that Meta was investing $14.3 billion for a 49% stake in Scale and was hiring away its founder. Shortly after that, several of Scale AI’s largest data customers, who are competitors to Meta’s efforts, reportedly cut ties with it.

CoreWeave, Inc. (NASDAQ: CRWV) has announced the acquisition of OpenPipe Inc., a leader in reinforcement learning (RL) platforms for training AI agents, marking a strategic move to strengthen its AI cloud capabilities. OpenPipe’s technology is designed to enable developers to train agents using advanced machine learning techniques, allowing the agents to learn from experience and improve over time in accuracy, performance, and reliability. The platform includes Agent Reinforcement Trainer (ART), one of the most widely used open-source RL toolkits for training agents [5].

Brian Venturo, Co-founder and Chief Strategy Officer at CoreWeave, highlighted the importance of reinforcement learning in enhancing model performance for agentic and reasoning tasks. The acquisition integrates OpenPipe’s self-learning tools with CoreWeave’s high-performance AI cloud, creating a more comprehensive platform for developers to build scalable intelligent systems. Kyle Corbitt, Co-founder and CEO of OpenPipe, added that the partnership with CoreWeave enables the expansion of their vision to accelerate the development of reliable, high-performing, and cost-effective AI systems [5].

The acquisition builds upon CoreWeave’s recent acquisition of Weights & Biases, a move that aligns with the company’s strategy to deepen its vertical integration across its technology stack. By incorporating new reinforcement learning and fine-tuning capabilities, CoreWeave is offering customers greater flexibility to train, adapt, and optimize their AI models. This expansion also supports AI labs and enterprises in solving complex problems autonomously, as reported by industry experts in the field [5].

CoreWeave’s AI cloud platform is purpose-built for the scale, performance, and expertise required to power AI innovation. The company operates a growing network of data centers across the U.S. and Europe, and it has been recognized as one of the TIME100 most influential companies and featured on Forbes Cloud 100 in 2024. The acquisition of OpenPipe further positions CoreWeave to meet the growing demand for AI infrastructure, particularly in the realm of autonomous decision-making and learning systems [5].

In addition to this strategic acquisition, CoreWeave recently participated in the Goldman Sachs Communacopia + Technology Conference, where CEO Michael Intrator and Chief Development Officer Brannin McBee provided insights into the company’s vision and roadmap for the future [6]. This engagement with major financial institutions underscores CoreWeave’s ongoing efforts to enhance transparency and communication with its investors and the broader market.

The market has seen mixed reactions to CoreWeave’s recent developments. While the stock initially saw gains following NVIDIA’s strong AI GPU earnings, recent insider selling by top executives and major shareholders like Magnetar Financial triggered a decline in its share price [2]. Despite these short-term fluctuations, CoreWeave remains a key player in the AI infrastructure sector, with strategic partnerships such as the recent expansion with Applied Digital (APLD) further solidifying its position in the market [4].

Source:

[1] CoreWeave, Inc. (CRWV) Is One Of The Biggest Beneficiaries Of NVIDIA’s Booming AI GPU Demand, Says Jim Cramer (https://finance.yahoo.com/news/coreweave-inc-crwv-one-biggest-192935144.html)

[2] CoreWeaves’ Stock Slides as Insider Selling Sparks Investor Concerns (https://www.marketwatch.com/story/coreweaves-stock-slides-as-insider-selling-sparks-investor-concerns-fef032fe)

[3] How Does Reinforcement Learning Power Agentic AI Systems (https://www.getmonetizely.com/articles/how-does-reinforcement-learning-power-agentic-ai-systems)

[4] CoreWeave Just Gave This Data Center Stock a Big Boost (https://finance.yahoo.com/news/coreweave-just-gave-data-center-162455061.html)

[5] CoreWeave to Acquire OpenPipe, Leader in Reinforcement Learning (https://www.businesswire.com/news/home/20250903667712/en/CoreWeave-to-Acquire-OpenPipe-Leader-in-Reinforcement-Learning)

[6] CoreWeave to Participate in the Goldman Sachs Communacopia + Technology Conference (https://investors.coreweave.com/news/news-details/2025/CoreWeave-to-Participate-in-the-Goldman-Sachs-Communacopia–Technology-Conference/default.aspx)