Sinking Servers into the Sea? They're Dead Serious About This

marsbitPublished on 2026-05-20Last updated on 2026-05-20

Abstract

Sinking Servers into the Sea: A Serious Undertaking The article details China's launch of the world's first offshore, directly wind-powered, subsea data center in the East China Sea near Shanghai. This 1.95 billion yuan project houses over 2,000 servers in a submerged 10-meter-deep module. It is directly powered by a nearby offshore wind farm (over 95% green energy) and cooled by seawater. This innovative approach tackles the two core challenges of data centers: massive power consumption and heat dissipation. It achieves an exceptional Power Usage Effectiveness (PUE) of 1.15, far better than China's national average of 1.48, saving an estimated 61 million kWh of electricity annually. It also uses no freshwater and requires significantly less land. The concept builds upon earlier experiments, like Microsoft's Project Natick, which proved servers could reliably operate underwater with lower failure rates due to a stable, inert environment. The Shanghai project advances the model by co-locating with wind farms, simultaneously solving both the power source and cooling source problems in an economically viable way. This integration reduces infrastructure costs and eliminates grid transmission losses for the electricity used on-site. Looking ahead, the vision is to integrate data center modules directly into the foundations of future large-scale, deep-sea wind turbines. This synergy could create a distributed network of "compute factories" at sea, powered by cheap, local green ...

"Total investment of 1.6 billion yuan, PUE as low as 1.15, over 95% direct green power supply, saving 61 million kWh of electricity annually. The world's first offshore wind power direct-connected subsea data center begins operations in the East China Sea off Shanghai's Lingang."

The year is 2026. Everyone's happily using AI, but the people building computing power centers are going crazy. The demand for computing power is skyrocketing too fast, and cooling and power supply can't keep up. The industry has reached a stage where imagination is key. Some time ago, there was even talk of the concept of space-based computing power—launching data centers into outer space. And now, someone has actually thrown servers into the sea.

This isn't some futuristic concept we're introducing. It's already happened—an investment of 1.6 billion yuan, throwing over 2000 servers into the sea. Right now, in the East China Sea off Shanghai's Lingang, east of Xiaoyangshan Island, 10 meters underwater beneath an offshore platform, 192 cabinets are packed into a four-level underwater data hall, running computing power non-stop. The whole structure weighs 1950 tons, roughly equivalent to 1300 family cars. About 500 meters away, over 50 wind turbines are spinning, with wind power fed directly in—the green power supply rate exceeds 95%.

Let's look at a few figures first. PUE (Power Usage Effectiveness, a data center energy efficiency metric, closer to 1 is better): this subsea data center is 1.15 (this number is impressive, we'll come back to it later). The national average is 1.48. Freshwater consumption: Zero. Land area: 200 square meters. A land-based facility of the same scale would need 2000 square meters. After full-scale operation, it is expected to save 61 million kWh of electricity annually.

In other words, dunking servers in the sea not only didn't ruin them but also made them more energy-efficient, water-saving, land-saving, and even lowered the failure rate compared to being on land.

Recently, CCTV broadcast this news. After watching it, I dug into the backstory and found it far more interesting than what was reported.

Looking back, this is a path that has been explored and validated for years. Repeated verification was needed to ultimately ensure computing power could be safely thrown into the water. Looking forward, the grand endeavors of computing power centers and green energy happen to converge on this path—a major strategic move taking a crucial step.

It's worth telling from the beginning.

01: Why on Earth Sink Servers into the Sea?

Data centers can be complex or simple. Simplifying it, they need to solve two core problems: power supply and cooling.

Everyone knows servers need electricity. But many don't realize that the electricity used to cool servers might be almost as much as what the servers themselves consume.

The industry has a core metric for measuring data center energy efficiency called PUE, Power Usage Effectiveness. The calculation is straightforward: the total electricity used by the entire data center divided by the electricity used by IT equipment (servers, storage, networking). If the PUE is 2, it means for every 1 kWh the server burns to compute, the air conditioning and other supporting facilities burn another 1 kWh to cool it and keep it running.

Ideally, PUE should be 1, meaning all electricity goes into computing, with none wasted on cooling. But in reality, it can never reach 1, only approach it infinitely.

The average PUE for data centers across China is roughly 1.48. In other words, for every 3 kWh of electricity burned in Chinese data centers, about 1 kWh is used for air conditioning.

In 2024, global data center electricity consumption was approximately 415 terawatt-hours, accounting for 1.5% of total global electricity use. The IEA (International Energy Agency) predicts this number will more than double to 945 TWh by 2030. And that's just the energy consumption of traditional data centers. With the advent of AI, things have gotten even more exaggerated.

A standard CPU server used to consume about 300 watts. Replaced by a GPU server for AI training, the same rack might consume 3000 watts—a tenfold increase. The IEA report states that electricity consumption for AI-specific servers is expected to grow by 30% annually.

Someone with 20 years in the data center industry gave me a vivid picture: For an office building, the air conditioning units on the roof are sufficient for the whole building. But if you convert that building into a data center, cooling requirements increase exponentially. The area occupied by air conditioning and power equipment could even surpass that of the servers themselves. Eventually, filling the roof and the square below with AC outdoor units might still not be enough to dissipate the heat.

That's why the global data center industry has been pondering the same thing for years: how to find cheaper cooling sources. The answers are surprisingly consistent: go to nature.

Facebook once tried building data centers in high-latitude North America, as close to the Arctic Circle as possible, leveraging naturally low temperatures. A few years ago, Tencent built a data center inside a cave in Guizhou, where temperatures are constant year-round. In this matter, the primary siting criterion for big companies isn't transportation or talent; it's wherever it's cool.

China's "East Data West Computing" project follows the same logic: building data centers in Inner Mongolia, Guizhou, Gansu, etc. The west has electricity—cheap coal power and abundant new energy; the weather is cold—places like Ulanqab are below zero most of the year, offering strong natural cooling. The eight computing hubs and ten data center clusters essentially chase cheap electricity and free cooling westward.

But what about eastern cities?

Places like Shanghai, Shenzhen, and Beijing are precisely where computing power demand is strongest. Financial transactions, AI inference, cross-border data processing—many services are extremely latency-sensitive. Data can't always be routed to a cave 2000 kilometers away in Guizhou, computed, and sent back. Yet these cities happen to have the most expensive land, the tightest energy consumption quotas, and scorching hot summers.

Hence, the sea.

Seawater has an average annual temperature of only about 15 degrees Celsius, with extremely strong fluidity, offering cooling capacity dozens of times greater than lake water. Moreover, offshore wind power is undergoing large-scale construction, placing electricity right next door. Cooling source and power source—the two things data centers need most—are both available offshore.

Logically speaking, sinking servers into the sea is actually the most natural answer.

02: How Many Steps to Get Computing Power into the Sea?

Putting data centers on the seabed wasn't a Chinese idea first.

In 2015, Microsoft launched a project called Project Natick. The first experiment was quite simple: throw one down and see if it breaks. They sank a cylindrical sealed capsule about 2.4 meters in diameter to the Pacific seabed, containing servers, and ran it for 105 days to see if servers would actually work underwater.

The conclusion: yes.

In 2018, Microsoft moved to a second round with a formal deployment. Off the coast of Scotland's Orkney Islands, they sank a sealed container holding 864 servers to about 35 meters deep in the North Sea. Powered by local tidal and wind energy, cooled naturally by seawater, and then left alone.

Two years later, in 2020, Microsoft retrieved the unit from the seabed. Upon opening it, the data was astonishing.

Of the 800-plus underwater servers, only 6 had failed—a failure rate of about 0.7%. Meanwhile, Microsoft had a control group on land—135 servers running for the same two years. Eight failed, a failure rate of nearly 6%. The underwater failure rate was about one-eighth of that on land.

This was a counterintuitive result. Microsoft's explanation: The sealed capsule was filled with dry nitrogen—no oxygen, no humidity, no dust, no vibrations or temperature fluctuations from people entering and leaving. The servers operated in an almost sterile environment, significantly slowing hardware aging.

No one touching, no one looking, no dust, no one opening the door and walking in—instead, nothing went wrong. A place completely devoid of humans might just be the servers' ideal working environment.

Microsoft's experiment proved one thing: seabed cooling is feasible. What followed next was done by the Chinese.

In 2020, Hi-Target, a domestic listed company specializing in marine equipment, acquired a Canadian deep-sea equipment team. This team had previously been involved in the engineering work for Microsoft's Natick project. More importantly, they had over 20 years of experience in deep-sea fields. The know-how accumulated through that experience was crucial: what microorganisms grow where, the flow and geological conditions of different sea areas, how to design joints to last 20 years underwater without problems.

With this technical foundation, the first commercial subsea data center landed in Hainan.

Located in Lingshui's Qingshui Bay, Hainan, about 3 kilometers offshore at a depth of 40 meters. The design involved sinking a sealed canister to the seabed, connected to an onshore control station via submarine cable, using natural seawater cooling. Trial operations began in 2022.

After running for over three years, several key figures emerged. PUE less than 1.2, far better than the national average of 1.48. Cooling energy consumption saved over 90%, meaning annual electricity savings of about 3 million kWh, freshwater savings of about 15,000 tons, and an onshore footprint of only 400-500 square meters—roughly one-fifth of a land-based facility of the same scale.

It might sound like moving data centers underwater solves everything, but it's far from that.

For the Hainan generation, the cooling source problem was solved, and costs were validated, but the power source remained a短板. Over 70% of Hainan's grid relies on thermal power. The subsea data center uses municipal power from shore, requiring a submarine cable costing tens of millions of yuan. Daily operating costs are indeed low, but factoring in the heavy upfront capital investment during construction, the economics weren't great. Moreover, relying on thermal power wasn't green enough in the long run.

How to secure both cooling source and power source simultaneously? That's where the next step, Shanghai, comes in.

The Shanghai project follows a completely different approach: Located in the East China Sea off Lingang, east of Xiaoyangshan Island, just 500 meters from an existing 200-megawatt offshore wind farm. Wind power is directly fed into the data center via submarine cable, bypassing the land grid entirely, using genuine green power. Both cooling source and power source are finally solved together.

The crucial change lies in the cost structure. The Hainan generation required building its own shore station, laying cables, pulling networks—these infrastructure components accounted for a large proportion of the total investment. For the Shanghai generation, the shore station, cables, network, even some electrical equipment already existed at the wind farm site and were directly repurposed. This alone reduced investment by tens of percent.

The reason this journey took so long involves difficulties beyond technology, because this is something almost no one has done before, requiring the establishment of many standards from scratch.

First, environmental protection. A little-known fact: environmental standards for marine projects are much stricter than on land.

When you put a continuously heat-emitting object in the sea, effective heat dissipation is called cooling; ineffective dissipation turns it into an underwater heater. Someone previously tried using lake water to cool a data center, pumping cold lake water up for cooling and discharging it back. The result was a rise in lake temperature. While it didn't boil the fish, fish growth rates noticeably accelerated, disrupting the ecological balance—environmental standards were not met.

Marine standards are even stricter. Practitioners tell us that on this detail, environmental regulations require that the seawater temperature change around the data center (roughly within 1 meter) cannot exceed 0.1 degrees Celsius. 0.1 degrees—that precision is already demanding. But just meeting the standard isn't enough; you also need continuous monitoring capability and contingency plans for extreme situations. Not everyone possesses these capabilities.

Hailan Cloud, the company behind this project, is a subsidiary of Hi-Target, a listed company in the marine technology field. The parent company has years of experience in marine observation, subsea equipment, and maritime communications. If a company without marine engineering background attempted this, they might be stuck at the environmental hurdle alone.

Then there's the server side. Those willing to throw expensive servers into the water must truly trust the solution. This equipment isn't cheap; ruining it would be genuinely painful. Moreover, such projects inherently require every hardware component to be highly reliable. You can't expect to send someone down frequently to replace cards; it's very troublesome.

For various reasons, this project ultimately brought together major players from different fields across the chain. For instance, Shenergy, operating the wind farm, a major player in East China's energy sector; Shanghai Electric, the established industrial group, responsible for servers here; telecommunications handled by Shanghai Telecom, self-explanatory. In summary, marine engineering, energy supply, computing operations, server manufacturing—all assembled here. If one link fails, the whole endeavor falls apart.

03: Wind + Computing Power, How Big Is the Potential?

Looking further ahead, even the Shanghai project is just a beginning. What truly makes this path seem to have vast potential is its future integration with large-scale, far-offshore wind farms.

Shanghai is planning a deep-sea offshore wind farm with a total capacity of 4300 megawatts. Currently, a medium-to-large data center in Shanghai typically has a capacity around 20 megawatts. This means the generating capacity of this single wind farm could theoretically power over 200 medium-to-large data centers.

Of course, not all would go to data centers. But a relatively conservative proportion has been calculated: taking about 15% of the wind farm's maximum generating capacity is sufficient to supply a large-scale offshore computing cluster. What does 15% mean? Even when the minimum number of turbines are spinning, this portion of electricity is stable—no need for additional energy storage, no worry about fluctuations. It's the best kind of power.

Calculating based on this proportion, 15% of 4300 MW is approximately 600 MW. A 600 MW computing center, located offshore, directly supplied by green power, naturally cooled by seawater.

There's an important economic calculation here: Far-offshore wind farms are over a hundred kilometers from shore. Transmitting electricity from sea to land incurs transmission losses exceeding 10%. But computing power doesn't require transmitting electricity. Converting electricity directly into computing power at sea and sending the computed results back via fiber optics involves almost zero transmission loss. Electricity loss is over 10%, data loss is nearly zero. Both involve bringing something back from the sea; sending bits is far more economical than sending electrons.

Thinking even further. The foundation of an offshore wind turbine is called the tower or monopile—the large cylinder inserted into the seabed. Offshore wind turbines are getting bigger, with a single turbine's power increasing from 2-3 MW on land to 12 to 20 MW offshore. The tower diameter has also grown, now around 18 to 20 meters.

An 18 to 20 meter diameter cylinder, hollow inside, offers considerable space, but no one previously thought about what could be done inside.

What if, when constructing the wind farm, space was reserved inside to place servers? No need to build additional structures, lay extra cables. The wind farm's power supply facilities, submarine cables, network connections already exist. It's like building the server room while constructing the house.

Calculated this way, overall construction costs could drop by tens of percent compared to land-based facilities, not even counting electricity prices. If the cost of power consumed locally at the far-offshore site could be negotiated down to around three or four cents per kWh, overall operating costs could drop another notch.

A single wind turbine: blades spinning above generating electricity, servers running inside the column below. Each turbine scattered across the sea surface becomes a mini computing factory—no grid power, no freshwater, no human maintenance required. Perfect.

This sounds like science fiction, but the underlying logic holds at every step. Reaching this point, you realize the grand undertakings of energy and computing power interlock seamlessly. Wind power needs local consumption to improve economics; computing power needs cheap green electricity and free cooling. Two endeavors, originally on separate paths, converge at sea.

China has a significant advantage in pursuing this: offshore wind power. China possesses the world's largest installed offshore wind capacity, the lowest power generation costs, and the most mature construction supply chain. If anyone can pioneer the integration of computing centers with offshore wind, it's highly likely to be in China.

"East Data West Computing" goes west, chasing coal and cold. Now, someone is heading east, chasing wind and sea. Two routes, solving the same problem: letting computing power use the cheapest electricity and the freest cooling.

This article is from the WeChat public account "Cool Play Lab," author: Cool Play Lab

Related Questions

QWhat are the two core challenges that data centers primarily need to address, and how does the seabed data center in Shanghai tackle them?

AThe two core challenges are power supply and heat dissipation. The Shanghai seabed data center addresses these by being deployed near an offshore wind farm, using direct green wind power supply (over 95% green electricity), and utilizing the naturally cool (around 15°C) and highly mobile seawater for efficient, passive cooling.

QWhat is PUE and why is the PUE value of 1.15 for the Shanghai seabed data center considered significant?

APUE (Power Usage Effectiveness) is a key metric for data center energy efficiency. It is calculated by dividing the total power used by the entire data center by the power used solely by the IT equipment (servers, storage, network). A value closer to 1 indicates higher efficiency. The PUE of 1.15 for the seabed center is significant because it is much lower than China's national average of 1.48, meaning far less energy is wasted on cooling and supporting infrastructure compared to traditional land-based data centers.

QWhat surprising finding did Microsoft's Project Natick reveal about server reliability in an underwater environment?

AMicrosoft's Project Natick found that servers deployed in a sealed, nitrogen-filled underwater capsule had a remarkably lower failure rate. After two years, only about 0.7% of the 864 servers failed, compared to a nearly 6% failure rate in a comparable land-based control group. This was attributed to the stable, vibration-free, oxygen-free, dust-free, and human-interaction-free environment.

QWhat is the key economic and technical advantage proposed for co-locating future data centers within the support structures (monopiles) of large offshore wind turbines?

AThe key advantage is the significant reduction in construction and operational costs. By integrating the data center space into the wind turbine's monopile during construction, there is no need for separate foundations, dedicated power cables, or complex grid connections—the wind farm's existing infrastructure is reused. This, combined with access to cheap, on-site green power and free seawater cooling, could drastically lower the total cost of ownership for the compute power.

QHow does the 'computing power to the sea' strategy complement China's broader 'East Data West Computing' project?

ABoth strategies aim to solve the same fundamental problem: providing compute power with cheap electricity and free cooling. 'East Data West Computing' moves data centers inland to western regions for access to cheaper coal power and cooler climates. The 'computing power to the sea' strategy moves eastward to coastal waters, chasing abundant offshore wind power (green electricity) and the massive heat dissipation capacity of the ocean. Together, they offer dual pathways to build efficient, sustainable, and cost-effective national computing infrastructure.

Related Reads

The Revelation from the Raydium Theft Incident: New DeFi Vulnerabilities Lurking in Forgotten Old Contracts

**Raydium Exploit Reveals DeFi's Hidden Risk: Forgotten "Zombie" Contracts** A recent attack on Raydium's deprecated V3 AMM pools resulted in a loss of approximately $1.34 million. The hacker exploited pools that were no longer supported by Raydium's current UI or SDK but remained fully functional and accessible on-chain. This incident highlights a critical, often overlooked category of risk in DeFi: inactive or legacy smart contracts that projects fail to properly decommission. Since March 2025, there have been at least 8 publicly reported attacks targeting such abandoned contracts, with total losses around $10.8 million. Including older pools and deprecated features, the count rises to 10 incidents with roughly $22.5 million in losses. These "zombie contracts" represent a lifecycle management failure rather than a code vulnerability, yet they are typically misclassified under general "code bug" categories in security reports, masking the true scale of the problem. The root cause is that projects often merely document a contract as "deprecated" without taking essential technical steps to secure it: withdrawing remaining assets, disabling external call functions, and implementing ongoing monitoring. These forgotten, under-monitored components become prime targets for attackers. To address this, the industry needs to recognize "zombie contracts" as a distinct risk category and establish standardized decommissioning protocols. Essential steps should include: 1) a formal retirement announcement, 2) removal of all front-end integrations, 3) withdrawal of locked assets, 4) disabling key contract functions, 5) ongoing security monitoring, 6) clear user communication, and 7) a post-mortem analysis. The value of a DeFi project lies not only in its current TVL but also in the security of its historical codebase, which has now become a new attack surface.

Foresight News14m ago

The Revelation from the Raydium Theft Incident: New DeFi Vulnerabilities Lurking in Forgotten Old Contracts

Foresight News14m ago

Robots Begin to 'Consume Data': The Hidden Production Chain from Indian Data Factories to Billion-Dollar Humanoid Robots

Robots have started to 'consume data,' driving the formation of a new industrial supply chain focused on producing training data for embodied AI. Unlike large language models, which are trained on vast internet text corpora, embodied AI models face a 'data desert' in the physical world. This has created a massive demand for first-person perspective video data (Ego Data), captured by workers wearing cameras in places like Indian garment factories. Companies like Neocambrian AI are establishing 'data factories' where workers perform standardized tasks (e.g., sorting clothes, kitchen organization) to generate thousands of hours of video. Research, such as NVIDIA's EgoScale, demonstrates that scaling this human demonstration data predictably improves robot performance, particularly for dexterous manipulation. This has validated a training path combining large-scale human data for pre-training with smaller amounts of robot-specific data for fine-tuning. The value of different data types varies significantly, forming a 'data pyramid.' The base consists of low-cost, large-scale internet and Ego Data. Higher layers include more expensive motion-capture data (e.g., from data gloves), simulation/synthetic data, and the most costly and scarce layer: real robot teleoperation data. This demand has spawned a layered ecosystem of data suppliers: low-cost data factories, motion capture and alignment specialists, robot-native teleoperation service providers, simulation data companies, and platforms aiming for data standardization. Robot companies themselves are adopting a 'layered procurement' strategy: outsourcing generic Ego Data while building in-house capabilities for robot-specific adaptation data and the critical deployment/failure data generated in real-world applications. The industry is shifting focus from hardware and basic mobility to the data pipelines required for general-purpose capability. While parallels exist to data labeling companies like Scale AI in the LLM boom, the physical complexity of robot data—involving action success ambiguity and sim-to-real gaps—requires more integrated solutions for data collection, annotation, and a continuous feedback loop. The race is on to build the data engines that will teach robots to operate reliably in the unstructured real world.

marsbit2h ago

Robots Begin to 'Consume Data': The Hidden Production Chain from Indian Data Factories to Billion-Dollar Humanoid Robots

marsbit2h ago

Spicy Commentary | Michael Saylor's 'Player Talk'; 60-Year-Old Aunt Liquidated After 'Scamming a Young Man'

**"Spicy Commentary": Three Tales of Crypto's Wild Week** This week's "Spicy Commentary" column highlights three dramatic stories from the cryptocurrency world. First, **MicroStrategy's Michael Saylor** addressed the controversy over his company potentially selling Bitcoin. At the BTC Prague event, he clarified, "I never said the company can't sell Bitcoin. I told *you* never to sell *your* Bitcoin." This "do as I say, not as I do" stance was criticized by netizens as peak linguistic gymnastics, noting a history of him previously stating the company would "never" sell. Second, a **bizarre fraud case** emerged from Beijing. A 60-year-old woman, obsessed with getting rich from crypto but unwilling to risk her own savings, posed online as the 20-something "god-daughter" of a high-ranking official. She catfished a young man, convincing him to give her over 200,000 yuan for fabricated emergencies. She then invested all the stolen money into cryptocurrency with 10x leverage, only to lose everything in a market crash. The woman was sentenced to four years in prison for fraud. Finally, a **sobering trader's tale** surfaced on Reddit. A user posted "Tale of a crypto trader," confessing their net worth had plummeted from a peak of $45 million to roughly $17,200, primarily due to holding meme coins too long. The post, described as a crypto "book of confessions," sparked reactions ranging from sympathy to critique about greed, poor risk management, and the perils of treating meme coins as long-term investments instead of taking profits. The column concludes that this week featured masterful rhetoric, elaborate scams, and extreme financial volatility, stitching together another chapter in crypto's unpredictable theater.

Foresight News3h ago

Spicy Commentary | Michael Saylor's 'Player Talk'; 60-Year-Old Aunt Liquidated After 'Scamming a Young Man'

Foresight News3h ago

Tremble Humans, AI Continues Its Accelerated Sprint

Trembling, Humans: AI Continues Its Accelerated Sprint Yes, AI is still rapidly accelerating. While deep learning seemed to stall quickly in its early years, large models after years of development show no sign of hitting their ceiling. At the Zhiyuan Conference 2026, the focus is on enabling AI to move from the digital world into the physical world. Scaling Law remains effective, continuing to drive advancements in both large language models and multimodal models. The industry is now entering a phase of pursuing World Models, though unresolved technical paths and data issues mean this exploration may take 3-5 more years. Concurrently, breakthroughs in Agents are accelerating AI's real-world application in fields like healthcare and meetings. Making Agents truly useful requires key hardware-software co-design, evident from the strong presence of chip vendors at the conference. We stand at a new historical threshold where AI is becoming a foundational force reshaping the world. The first day of the conference highlighted AI's evolution from "knowing how to chat" to "knowing how to work." Scaling Law persists, World Models are the next key battleground, and Agents are transitioning from usable to好用 (user-friendly). Scaling Law is not ending but diversifying. New models like Anthropic's Fable 5 demonstrate scaling through parameter size, synthetic data, and reinforcement learning. Advancements in AI Coding and Agent deployment are enabling a trend of AI self-evolution, potentially allowing AI to take over digital world iterations. World Models represent the next frontier for large models extending into the physical realm, but no current model is truly impressive at solving real-world problems. Technical consensus is lacking, with debates on data sources (video, simulation, real-world). Different approaches are emerging: language-centric, pixel-centric, 3D-structure-centric, and visual-representation-centric models. Zhiyuan Institute is exploring a fifth path: unified latent space modeling fusing language and visual representations, and introduced its own under-development World Model, Physis-v0.1. On the product side, Agents are key to bringing AI into daily life. Since 2025, the "Year of the Agent," products have become more proactive and capable of complex tasks. Zhiyuan showcased four vertical Agents for cardiac diagnosis, autonomous research, meeting summarization, and protein risk discovery. However, technical challenges remain, particularly in context engineering like memory and orchestration. "Harness" – the engineering framework around an Agent – is crucial for maximizing its capabilities by clarifying intent, designing workflows, and incorporating validation and feedback. In summary, AI's breakneck pace continues on multiple fronts: foundational model scaling, the ambitious pursuit of World Models for physical understanding, and the ongoing refinement of practical Agents. The journey from capable to truly reliable and useful AI systems is well underway.

marsbit3h ago

Tremble Humans, AI Continues Its Accelerated Sprint

marsbit3h ago

Trading

Spot
Futures
活动图片