Get ONMINE

Stay Ahead, Stay ONMINE

How to build a better AI benchmark

It’s not easy being one of Silicon Valley’s favorite benchmarks.  SWE-Bench (pronounced “swee bench”) launched in November 2024 to evaluate an AI model’s coding skill, using more than 2,000 real-world programming problems pulled from the public GitHub repositories of 12 different Python-based projects.  In the months since then, it’s quickly become one of the most popular tests in AI. A SWE-Bench score has become a mainstay of major model releases from OpenAI, Anthropic, and Google—and outside of foundation models, the fine-tuners at AI firms are in constant competition to see who can rise above the pack. The top of the leaderboard is a pileup between three different fine tunings of Anthropic’s Claude Sonnet model and Amazon’s Q developer agent. Auto Code Rover—one of the Claude modifications—nabbed the number two spot in November, and was acquired just three months later. Despite all the fervor, this isn’t exactly a truthful assessment of which model is “better.” As the benchmark has gained prominence, “you start to see that people really want that top spot,” says John Yang, a researcher on the team that developed SWE-Bench at Princeton University. As a result, entrants have begun to game the system—which is pushing many others to wonder whether there’s a better way to actually measure AI achievement. Developers of these coding agents aren’t necessarily doing anything as straightforward cheating, but they’re crafting approaches that are too neatly tailored to the specifics of the benchmark. The initial SWE-Bench test set was limited to programs written in Python, which meant developers could gain an advantage by training their models exclusively on Python code. Soon, Yang noticed that high-scoring models would fail completely when tested on different programming languages—revealing an approach to the test that he describes as “gilded.” “It looks nice and shiny at first glance, but then you try to run it on a different language and the whole thing just kind of falls apart,” Yang says. “At that point, you’re not designing a software engineering agent. You’re designing to make a SWE-Bench agent, which is much less interesting.” The SWE-Bench issue is a symptom of a more sweeping—and complicated—problem in AI evaluation, and one that’s increasingly sparking heated debate: The benchmarks the industry uses to guide development are drifting further and further away from evaluating actual capabilities, calling their basic value into question. Making the situation worse, several benchmarks, most notably FrontierMath and Chatbot Arena, have recently come under heat for an alleged lack of transparency. Nevertheless, benchmarks still play a central role in model development, even if few experts are willing to take their results at face value. OpenAI cofounder Andrej Karpathy recently described the situation as “an evaluation crisis”: the industry has fewer trusted methods for measuring capabilities and no clear path to better ones.  “Historically, benchmarks were the way we evaluated AI systems,” says Vanessa Parli, director of research at Stanford University’s Institute for Human-Centered AI. “Is that the way we want to evaluate systems going forward? And if it’s not, what is the way?” A growing group of academics and AI researchers are making the case that the answer is to go smaller, trading sweeping ambition for an approach inspired by the social sciences. Specifically, they want to focus more on testing validity, which for quantitative social scientists refers to how well a given questionnaire measures what it’s claiming to measure—and, more fundamentally, whether what it is measuring has a coherent definition. That could cause trouble for benchmarks assessing hazily defined concepts like “reasoning” or “scientific knowledge”—and for developers aiming to reach the much-hyped goal of artificial general intelligence—but it would put the industry on firmer ground as it looks to prove the worth of individual models. “Taking validity seriously means asking folks in academia, industry, or wherever to show that their system does what they say it does,” says Abigail Jacobs, a University of Michigan professor who is a central figure in the new push for validity. “I think it points to a weakness in the AI world if they want to back off from showing that they can support their claim.” The limits of traditional testing If AI companies have been slow to respond to the growing failure of benchmarks, it’s partially because the test-scoring approach has been so effective for so long.  One of the biggest early successes of contemporary AI was the ImageNet challenge, a kind of antecedent to contemporary benchmarks. Released in 2010 as an open challenge to researchers, the database held more than 3 million images for AI systems to categorize into 1,000 different classes. Crucially, the test was completely agnostic to methods, and any successful algorithm quickly gained credibility regardless of how it worked. When an algorithm called AlexNet broke through in 2012, with a then unconventional form of GPU training, it became one of the foundational results of modern AI. Few would have guessed in advance that AlexNet’s convolutional neural nets would be the secret to unlocking image recognition—but after it scored well, no one dared dispute it. (One of AlexNet’s developers, Ilya Sutskever, would go on to cofound OpenAI.) A large part of what made this challenge so effective was that there was little practical difference between ImageNet’s object classification challenge and the actual process of asking a computer to recognize an image. Even if there were disputes about methods, no one doubted that the highest-scoring model would have an advantage when deployed in an actual image recognition system. But in the 12 years since, AI researchers have applied that same method-agnostic approach to increasingly general tasks. SWE-Bench is commonly used as a proxy for broader coding ability, while other exam-style benchmarks often stand in for reasoning ability. That broad scope makes it difficult to be rigorous about what a specific benchmark measures—which, in turn, makes it hard to use the findings responsibly.  Where things break down Anka Reuel, a PhD student who has been focusing on the benchmark problem as part of her research at Stanford, has become convinced the evaluation problem is the result of this push toward generality. “We’ve moved from task-specific models to general-purpose models,” Reuel says. “It’s not about a single task anymore but a whole bunch of tasks, so evaluation becomes harder.” Like the University of Michigan’s Jacobs, Reuel thinks “the main issue with benchmarks is validity, even more than the practical implementation,” noting: “That’s where a lot of things break down.” For a task as complicated as coding, for instance, it’s nearly impossible to incorporate every possible scenario into your problem set. As a result, it’s hard to gauge whether a model is scoring better because it’s more skilled at coding or because it has more effectively manipulated the problem set. And with so much pressure on developers to achieve record scores, shortcuts are hard to resist. For developers, the hope is that success on lots of specific benchmarks will add up to a generally capable model. But the techniques of agentic AI mean a single AI system can encompass a complex array of different models, making it hard to evaluate whether improvement on a specific task will lead to generalization. “There’s just many more knobs you can turn,” says Sayash Kapoor, a computer scientist at Princeton and a prominent critic of sloppy practices in the AI industry. “When it comes to agents, they have sort of given up on the best practices for evaluation.” In a paper from last July, Kapoor called out specific issues in how AI models were approaching the WebArena benchmark, designed by Carnegie Mellon University researchers in 2024 as a test of an AI agent’s ability to traverse the web. The benchmark consists of more than 800 tasks to be performed on a set of cloned websites mimicking Reddit, Wikipedia, and others. Kapoor and his team identified an apparent hack in the winning model, called STeP. STeP included specific instructions about how Reddit structures URLs, allowing STeP models to jump directly to a given user’s profile page (a frequent element of WebArena tasks). This shortcut wasn’t exactly cheating, but Kapoor sees it as “a serious misrepresentation of how well the agent would work had it seen the tasks in WebArena for the first time.” Because the technique was successful, though, a similar policy has since been adopted by OpenAI’s web agent Operator. (“Our evaluation setting is designed to assess how well an agent can solve tasks given some instruction about website structures and task execution,” an OpenAI representative said when reached for comment. “This approach is consistent with how others have used and reported results with WebArena.” STeP did not respond to a request for comment.) Further highlighting the problem with AI benchmarks, late last month Kapoor and a team of researchers wrote a paper that revealed significant problems in Chatbot Arena, the popular crowdsourced evaluation system. According to the paper, the leaderboard was being manipulated; many top foundation models were conducting undisclosed private testing and releasing their scores selectively. Today, even ImageNet itself, the mother of all benchmarks, has started to fall victim to validity problems. A 2023 study from researchers at the University of Washington and Google Research found that when ImageNet-winning algorithms were pitted against six real-world data sets, the architecture improvement “resulted in little to no progress,” suggesting that the external validity of the test had reached its limit. Going smaller For those who believe the main problem is validity, the best fix is reconnecting benchmarks to specific tasks. As Reuel puts it, AI developers “have to resort to these high-level benchmarks that are almost meaningless for downstream consumers, because the benchmark developers can’t anticipate the downstream task anymore.” So what if there was a way to help the downstream consumers identify this gap? In November 2024, Reuel launched a public ranking project called BetterBench, which rates benchmarks on dozens of different criteria, such as whether the code has been publicly documented. But validity is a central theme, with particular criteria challenging designers to spell out what capability their benchmark is testing and how it relates to the tasks that make up the benchmark. “You need to have a structural breakdown of the capabilities,” Reuel says. “What are the actual skills you care about, and how do you operationalize them into something we can measure?” The results are surprising. One of the highest-scoring benchmarks is also the oldest: the Arcade Learning Environment (ALE), established in 2013 as a way to test models’ ability to learn how to play a library of Atari 2600 games. One of the lowest-scoring is the Massive Multitask Language Understanding (MMLU) benchmark, a widely used test for general language skills; by the standards of BetterBench, the connection between the questions and the underlying skill was too poorly defined. BetterBench hasn’t meant much for the reputations of specific benchmarks, at least not yet; MMLU is still widely used, and ALE is still marginal. But the project has succeeded in pushing validity into the broader conversation about how to fix benchmarks. In April, Reuel quietly joined a new research group hosted by Hugging Face, the University of Edinburgh, and EleutherAI, where she’ll develop her ideas on validity and AI model evaluation with other figures in the field. (An official announcement is expected later this month.)  Irene Solaiman, Hugging Face’s head of global policy, says the group will focus on building valid benchmarks that go beyond measuring straightforward capabilities. “There’s just so much hunger for a good benchmark off the shelf that already works,” Solaiman says. “A lot of evaluations are trying to do too much.” Increasingly, the rest of the industry seems to agree. In a paper in March, researchers from Google, Microsoft, Anthropic, and others laid out a new framework for improving evaluations—with validity as the first step.  “AI evaluation science must,” the researchers argue, “move beyond coarse grained claims of ‘general intelligence’ towards more task-specific and real-world relevant measures of progress.”  Measuring the “squishy” things To help make this shift, some researchers are looking to the tools of social science. A February position paper argued that “evaluating GenAI systems is a social science measurement challenge,” specifically unpacking how the validity systems used in social measurements can be applied to AI benchmarking.  The authors, largely employed by Microsoft’s research branch but joined by academics from Stanford and the University of Michigan, point to the standards that social scientists use to measure contested concepts like ideology, democracy, and media bias. Applied to AI benchmarks, those same procedures could offer a way to measure concepts like “reasoning” and “math proficiency” without slipping into hazy generalizations. In the social science literature, it’s particularly important that metrics begin with a rigorous definition of the concept measured by the test. For instance, if the test is to measure how democratic a society is, it first needs to establish a definition for a “democratic society” and then establish questions that are relevant to that definition.  To apply this to a benchmark like SWE-Bench, designers would need to set aside the classic machine learning approach, which is to collect programming problems from GitHub and create a scheme to validate answers as true or false. Instead, they’d first need to define what the benchmark aims to measure (“ability to resolve flagged issues in software,” for instance), break that into subskills (different types of problems or types of program that the AI model can successfully process), and then finally assemble questions that accurately cover the different subskills. It’s a profound change from how AI researchers typically approach benchmarking—but for researchers like Jacobs, a coauthor on the February paper, that’s the whole point. “There’s a mismatch between what’s happening in the tech industry and these tools from social science,” she says. “We have decades and decades of thinking about how we want to measure these squishy things about humans.” Even though the idea has made a real impact in the research world, it’s been slow to influence the way AI companies are actually using benchmarks.  The last two months have seen new model releases from OpenAI, Anthropic, Google, and Meta, and all of them lean heavily on multiple-choice knowledge benchmarks like MMLU—the exact approach that validity researchers are trying to move past. After all, model releases are, for the most part, still about showing increases in general intelligence, and broad benchmarks continue to be used to back up those claims.  For some observers, that’s good enough. Benchmarks, Wharton professor Ethan Mollick says, are “bad measures of things, but also they’re what we’ve got.” He adds: “At the same time, the models are getting better. A lot of sins are forgiven by fast progress.” For now, the industry’s long-standing focus on artificial general intelligence seems to be crowding out a more focused validity-based approach. As long as AI models can keep growing in general intelligence, then specific applications don’t seem as compelling—even if that leaves practitioners relying on tools they no longer fully trust.  “This is the tightrope we’re walking,” says Hugging Face’s Solaiman. “It’s too easy to throw the system out, but evaluations are really helpful in understanding our models, even with these limitations.” Russell Brandom is a freelance writer covering artificial intelligence. He lives in Brooklyn with his wife and two cats. This story was supported by a grant from the Tarbell Center for AI Journalism.

It’s not easy being one of Silicon Valley’s favorite benchmarks. 

SWE-Bench (pronounced “swee bench”) launched in November 2024 to evaluate an AI model’s coding skill, using more than 2,000 real-world programming problems pulled from the public GitHub repositories of 12 different Python-based projects. 

In the months since then, it’s quickly become one of the most popular tests in AI. A SWE-Bench score has become a mainstay of major model releases from OpenAI, Anthropic, and Google—and outside of foundation models, the fine-tuners at AI firms are in constant competition to see who can rise above the pack. The top of the leaderboard is a pileup between three different fine tunings of Anthropic’s Claude Sonnet model and Amazon’s Q developer agent. Auto Code Rover—one of the Claude modifications—nabbed the number two spot in November, and was acquired just three months later.

Despite all the fervor, this isn’t exactly a truthful assessment of which model is “better.” As the benchmark has gained prominence, “you start to see that people really want that top spot,” says John Yang, a researcher on the team that developed SWE-Bench at Princeton University. As a result, entrants have begun to game the system—which is pushing many others to wonder whether there’s a better way to actually measure AI achievement.

Developers of these coding agents aren’t necessarily doing anything as straightforward cheating, but they’re crafting approaches that are too neatly tailored to the specifics of the benchmark. The initial SWE-Bench test set was limited to programs written in Python, which meant developers could gain an advantage by training their models exclusively on Python code. Soon, Yang noticed that high-scoring models would fail completely when tested on different programming languages—revealing an approach to the test that he describes as “gilded.”

“It looks nice and shiny at first glance, but then you try to run it on a different language and the whole thing just kind of falls apart,” Yang says. “At that point, you’re not designing a software engineering agent. You’re designing to make a SWE-Bench agent, which is much less interesting.”

The SWE-Bench issue is a symptom of a more sweeping—and complicated—problem in AI evaluation, and one that’s increasingly sparking heated debate: The benchmarks the industry uses to guide development are drifting further and further away from evaluating actual capabilities, calling their basic value into question. Making the situation worse, several benchmarks, most notably FrontierMath and Chatbot Arena, have recently come under heat for an alleged lack of transparency. Nevertheless, benchmarks still play a central role in model development, even if few experts are willing to take their results at face value. OpenAI cofounder Andrej Karpathy recently described the situation as “an evaluation crisis”: the industry has fewer trusted methods for measuring capabilities and no clear path to better ones. 

“Historically, benchmarks were the way we evaluated AI systems,” says Vanessa Parli, director of research at Stanford University’s Institute for Human-Centered AI. “Is that the way we want to evaluate systems going forward? And if it’s not, what is the way?”

A growing group of academics and AI researchers are making the case that the answer is to go smaller, trading sweeping ambition for an approach inspired by the social sciences. Specifically, they want to focus more on testing validity, which for quantitative social scientists refers to how well a given questionnaire measures what it’s claiming to measure—and, more fundamentally, whether what it is measuring has a coherent definition. That could cause trouble for benchmarks assessing hazily defined concepts like “reasoning” or “scientific knowledge”—and for developers aiming to reach the much–hyped goal of artificial general intelligence—but it would put the industry on firmer ground as it looks to prove the worth of individual models.

“Taking validity seriously means asking folks in academia, industry, or wherever to show that their system does what they say it does,” says Abigail Jacobs, a University of Michigan professor who is a central figure in the new push for validity. “I think it points to a weakness in the AI world if they want to back off from showing that they can support their claim.”

The limits of traditional testing

If AI companies have been slow to respond to the growing failure of benchmarks, it’s partially because the test-scoring approach has been so effective for so long. 

One of the biggest early successes of contemporary AI was the ImageNet challenge, a kind of antecedent to contemporary benchmarks. Released in 2010 as an open challenge to researchers, the database held more than 3 million images for AI systems to categorize into 1,000 different classes.

Crucially, the test was completely agnostic to methods, and any successful algorithm quickly gained credibility regardless of how it worked. When an algorithm called AlexNet broke through in 2012, with a then unconventional form of GPU training, it became one of the foundational results of modern AI. Few would have guessed in advance that AlexNet’s convolutional neural nets would be the secret to unlocking image recognition—but after it scored well, no one dared dispute it. (One of AlexNet’s developers, Ilya Sutskever, would go on to cofound OpenAI.)

A large part of what made this challenge so effective was that there was little practical difference between ImageNet’s object classification challenge and the actual process of asking a computer to recognize an image. Even if there were disputes about methods, no one doubted that the highest-scoring model would have an advantage when deployed in an actual image recognition system.

But in the 12 years since, AI researchers have applied that same method-agnostic approach to increasingly general tasks. SWE-Bench is commonly used as a proxy for broader coding ability, while other exam-style benchmarks often stand in for reasoning ability. That broad scope makes it difficult to be rigorous about what a specific benchmark measures—which, in turn, makes it hard to use the findings responsibly. 

Where things break down

Anka Reuel, a PhD student who has been focusing on the benchmark problem as part of her research at Stanford, has become convinced the evaluation problem is the result of this push toward generality. “We’ve moved from task-specific models to general-purpose models,” Reuel says. “It’s not about a single task anymore but a whole bunch of tasks, so evaluation becomes harder.”

Like the University of Michigan’s Jacobs, Reuel thinks “the main issue with benchmarks is validity, even more than the practical implementation,” noting: “That’s where a lot of things break down.” For a task as complicated as coding, for instance, it’s nearly impossible to incorporate every possible scenario into your problem set. As a result, it’s hard to gauge whether a model is scoring better because it’s more skilled at coding or because it has more effectively manipulated the problem set. And with so much pressure on developers to achieve record scores, shortcuts are hard to resist.

For developers, the hope is that success on lots of specific benchmarks will add up to a generally capable model. But the techniques of agentic AI mean a single AI system can encompass a complex array of different models, making it hard to evaluate whether improvement on a specific task will lead to generalization. “There’s just many more knobs you can turn,” says Sayash Kapoor, a computer scientist at Princeton and a prominent critic of sloppy practices in the AI industry. “When it comes to agents, they have sort of given up on the best practices for evaluation.”

In a paper from last July, Kapoor called out specific issues in how AI models were approaching the WebArena benchmark, designed by Carnegie Mellon University researchers in 2024 as a test of an AI agent’s ability to traverse the web. The benchmark consists of more than 800 tasks to be performed on a set of cloned websites mimicking Reddit, Wikipedia, and others. Kapoor and his team identified an apparent hack in the winning model, called STeP. STeP included specific instructions about how Reddit structures URLs, allowing STeP models to jump directly to a given user’s profile page (a frequent element of WebArena tasks).

This shortcut wasn’t exactly cheating, but Kapoor sees it as “a serious misrepresentation of how well the agent would work had it seen the tasks in WebArena for the first time.” Because the technique was successful, though, a similar policy has since been adopted by OpenAI’s web agent Operator. (“Our evaluation setting is designed to assess how well an agent can solve tasks given some instruction about website structures and task execution,” an OpenAI representative said when reached for comment. “This approach is consistent with how others have used and reported results with WebArena.” STeP did not respond to a request for comment.)

Further highlighting the problem with AI benchmarks, late last month Kapoor and a team of researchers wrote a paper that revealed significant problems in Chatbot Arena, the popular crowdsourced evaluation system. According to the paper, the leaderboard was being manipulated; many top foundation models were conducting undisclosed private testing and releasing their scores selectively.

Today, even ImageNet itself, the mother of all benchmarks, has started to fall victim to validity problems. A 2023 study from researchers at the University of Washington and Google Research found that when ImageNet-winning algorithms were pitted against six real-world data sets, the architecture improvement “resulted in little to no progress,” suggesting that the external validity of the test had reached its limit.

Going smaller

For those who believe the main problem is validity, the best fix is reconnecting benchmarks to specific tasks. As Reuel puts it, AI developers “have to resort to these high-level benchmarks that are almost meaningless for downstream consumers, because the benchmark developers can’t anticipate the downstream task anymore.” So what if there was a way to help the downstream consumers identify this gap?

In November 2024, Reuel launched a public ranking project called BetterBench, which rates benchmarks on dozens of different criteria, such as whether the code has been publicly documented. But validity is a central theme, with particular criteria challenging designers to spell out what capability their benchmark is testing and how it relates to the tasks that make up the benchmark.

“You need to have a structural breakdown of the capabilities,” Reuel says. “What are the actual skills you care about, and how do you operationalize them into something we can measure?”

The results are surprising. One of the highest-scoring benchmarks is also the oldest: the Arcade Learning Environment (ALE), established in 2013 as a way to test models’ ability to learn how to play a library of Atari 2600 games. One of the lowest-scoring is the Massive Multitask Language Understanding (MMLU) benchmark, a widely used test for general language skills; by the standards of BetterBench, the connection between the questions and the underlying skill was too poorly defined.

BetterBench hasn’t meant much for the reputations of specific benchmarks, at least not yet; MMLU is still widely used, and ALE is still marginal. But the project has succeeded in pushing validity into the broader conversation about how to fix benchmarks. In April, Reuel quietly joined a new research group hosted by Hugging Face, the University of Edinburgh, and EleutherAI, where she’ll develop her ideas on validity and AI model evaluation with other figures in the field. (An official announcement is expected later this month.) 

Irene Solaiman, Hugging Face’s head of global policy, says the group will focus on building valid benchmarks that go beyond measuring straightforward capabilities. “There’s just so much hunger for a good benchmark off the shelf that already works,” Solaiman says. “A lot of evaluations are trying to do too much.”

Increasingly, the rest of the industry seems to agree. In a paper in March, researchers from Google, Microsoft, Anthropic, and others laid out a new framework for improving evaluations—with validity as the first step. 

“AI evaluation science must,” the researchers argue, “move beyond coarse grained claims of ‘general intelligence’ towards more task-specific and real-world relevant measures of progress.” 

Measuring the “squishy” things

To help make this shift, some researchers are looking to the tools of social science. A February position paper argued that “evaluating GenAI systems is a social science measurement challenge,” specifically unpacking how the validity systems used in social measurements can be applied to AI benchmarking. 

The authors, largely employed by Microsoft’s research branch but joined by academics from Stanford and the University of Michigan, point to the standards that social scientists use to measure contested concepts like ideology, democracy, and media bias. Applied to AI benchmarks, those same procedures could offer a way to measure concepts like “reasoning” and “math proficiency” without slipping into hazy generalizations.

In the social science literature, it’s particularly important that metrics begin with a rigorous definition of the concept measured by the test. For instance, if the test is to measure how democratic a society is, it first needs to establish a definition for a “democratic society” and then establish questions that are relevant to that definition. 

To apply this to a benchmark like SWE-Bench, designers would need to set aside the classic machine learning approach, which is to collect programming problems from GitHub and create a scheme to validate answers as true or false. Instead, they’d first need to define what the benchmark aims to measure (“ability to resolve flagged issues in software,” for instance), break that into subskills (different types of problems or types of program that the AI model can successfully process), and then finally assemble questions that accurately cover the different subskills.

It’s a profound change from how AI researchers typically approach benchmarking—but for researchers like Jacobs, a coauthor on the February paper, that’s the whole point. “There’s a mismatch between what’s happening in the tech industry and these tools from social science,” she says. “We have decades and decades of thinking about how we want to measure these squishy things about humans.”

Even though the idea has made a real impact in the research world, it’s been slow to influence the way AI companies are actually using benchmarks. 

The last two months have seen new model releases from OpenAI, Anthropic, Google, and Meta, and all of them lean heavily on multiple-choice knowledge benchmarks like MMLU—the exact approach that validity researchers are trying to move past. After all, model releases are, for the most part, still about showing increases in general intelligence, and broad benchmarks continue to be used to back up those claims. 

For some observers, that’s good enough. Benchmarks, Wharton professor Ethan Mollick says, are “bad measures of things, but also they’re what we’ve got.” He adds: “At the same time, the models are getting better. A lot of sins are forgiven by fast progress.”

For now, the industry’s long-standing focus on artificial general intelligence seems to be crowding out a more focused validity-based approach. As long as AI models can keep growing in general intelligence, then specific applications don’t seem as compelling—even if that leaves practitioners relying on tools they no longer fully trust. 

“This is the tightrope we’re walking,” says Hugging Face’s Solaiman. “It’s too easy to throw the system out, but evaluations are really helpful in understanding our models, even with these limitations.”

Russell Brandom is a freelance writer covering artificial intelligence. He lives in Brooklyn with his wife and two cats.

This story was supported by a grant from the Tarbell Center for AI Journalism.

Shape
Shape
Stay Ahead

Explore More Insights

More Articles

Stay ahead with more perspectives on cutting-edge power, infrastructure, energy,  bitcoin and AI solutions. Explore these articles to uncover strategies and insights shaping the future of industries.

Shape

How Jeetu Patel made Cisco unrecognizable

From dashboard sprawl to Cloud Control The most visible proof point of the new Cisco is Cloud Control, the unified management plane that now spans networking, security, compute, observability, collaboration, and an expanding ecosystem of third-party tools. Cisco is careful to note that this is not just another single pane

Read More »

IBM, ServiceNow team to bring AI to legacy enterprise systems

Decades of deeply interconnected legacy systems are the biggest barrier to moving fast on AI, the companies stated. Their pairings will take advantage of Big Blue’s expertise in working with large systems, such as its mainframe environment, and extensive legacy applications, along with ServiceNow’s workflow and agent management platforms. “Most

Read More »

Energy Secretary Keeps Coal-Fired Power Generation Alive in the Northwest

WASHINGTON—U.S. Secretary of Energy Chris Wright today issued an emergency order to keep affordable, reliable, and secure coal generation online and address critical grid reliability issues facing the Northwestern region of the United States. The emergency order directs TransAlta Centralia Generation LLC (TransAlta) to ensure that Unit 2 of the Centralia Generating Station in Centralia, Washington, a coal-fired power plant, remains available to operate. Centralia Unit 2 was scheduled to shut down at the end of 2025. The order minimizes the risk and cost of unnecessary blackouts. “Taking reliable generation off the grid compromises energy reliability and needlessly raises energy costs for Americans,” said Energy Secretary Wright. “During peak summer demand, Northwesterners deserve continued access to affordable, reliable, and secure energy to power and cool their homes.” Thanks to President Trump’s leadership, coal plants across the country are being saved from premature retirement and reversing plans to shut down. In 2025, more than 17 gigawatts of coal-power electricity generation were saved from going offline. As outlined in DOE’s Resource Adequacy Report, power outages could increase by 100 times by 2030 if the U.S. continues to take reliable power offline. The availability of Centralia to operate will continue to be an asset to maintain reliability in the Western Electricity Coordinating Council (WECC) Northwest region. The North American Electric Reliability Corporation’s (NERC) 2025 Long-Term Reliability Assessment assessed that the WECC Northwest region is at high risk of energy shortfalls over the next five years, noting that “rapid forecasted demand growth is driving the need for more resources” and that “periods of unserved energy are projected for both summer and winter.” This order is in effect beginning on June 15, 2026, through September 12, 2026. Background: According to the U.S. Environmental Protection Agency’s data, in 2025, Centralia generated an average of approximately 340,000 MWh per month, providing vital generation capacity to the region.  ###

Read More »

United States, Cyprus, Greece, Israel and Rice University To Establish Eastern Mediterranean Energy Center in Houston

HOUSTON, TEXAS—U.S. Secretary of Energy Chris Wright today signed a Declaration of Intent (DOI) with the Minister of Energy, Commerce, and Industry of the Republic of Cyprus Michael Damianos, Minister of Environment and Energy for Greece Stavros Papastavrou, Israeli Ambassador to the United States Dr. Yechiel Leiter, and President of Rice University Reginald DesRoches to establish the Eastern Mediterranean Energy Center (EMEC). The agreement establishes a framework to strengthen cooperation between the respective nations through the Eastern Mediterranean Energy Center (EMEC). It also advances a key initiative envisioned under Secretary Rubio’s Eastern Mediterranean Security and Energy Partnership Act of 2019. The agreement advances President Trump’s commitment to strengthening America’s partnerships with key allies while expanding opportunities for U.S. energy development, innovation, and investment. As global energy demand continues to grow, the United States, Cyprus, Greece, and Israel will work together to promote energy security, strengthen critical infrastructure, support emerging technologies, and advance long-term economic growth throughout the Eastern Mediterranean. “The Eastern Mediterranean Energy Center will help fulfill President Trump’s vision of prosperity and energy security at home and abroad,” said Secretary Wright. “The Eastern Mediterranean is an increasingly important region for global energy development, and this agreement strengthens cooperation among key allies while advancing our shared goals of energy abundance, economic prosperity, and regional stability. By establishing the Eastern Mediterranean Energy Center at Rice University in Houston, we are ensuring all member nations of this agreement will benefit from a lasting partnership bound together by the brightest minds and industry leaders in hydrocarbon development.” The partnership will support collaboration on shared priorities including natural gas development, U.S. LNG infrastructure, energy transportation networks, grid reliability, critical infrastructure resilience, and emerging technologies. It will also facilitate scientific and technical exchanges, research partnerships, workforce development initiatives, and engagement with industry stakeholders. The Trump

Read More »

Energy Secretary Secures Carolinas’ Grid Ahead of Period of Hot Weather

WASHINGTON—The U.S. Department of Energy (DOE) today issued an emergency order to mitigate blackouts in the Carolinas’ ahead of a period of hot weather. Issued pursuant to Section 202(c) of the Federal Power Act, the order authorizes Duke Energy Carolinas, LLC (“DEC”) and Duke Energy Progress, LLC (“DEP”) (collectively, “Duke Energy”) to operate specified units located within Duke Energy’s service territory to operate up to their maximum generation output levels, notwithstanding air quality or other permit limitations arising under federal, state, or local law or regulation, or other applicable source of law. The order was issued subsequent to Duke Energy’s application. The order will mitigate the risk of unnecessary blackouts brought on by unusually high load forecasts and high temperatures across the region. “Maintaining affordable, reliable, and secure power in the Duke Energy service territory is non-negotiable,” said U.S. Secretary of Energy Chris Wright. “The previous administration’s energy subtraction policies weakened the grid, leaving Americans more vulnerable during events like this. Thanks to President Trump’s leadership, we are reversing those failures and using every available tool ensuring Americans in the Carolinas’ have continued access to affordable, reliable, and secure energy to power and cool their homes.” On day one, President Trump declared a national energy emergency after the Biden administration’s energy subtraction agenda left behind a grid increasingly vulnerable to blackouts. The order is in effect beginning at 4:00 PM ET on June 11, 2026, and shall expire at 10:00 PM ET on June 12, 2026. Background: Duke Energy stated that some generating units are limited in providing needed generation because of conditions and limitations in their environmental permits. As a result, the system “may not have sufficient generation available to meet this unusually high demand and [Duke Energy] may be forced to curtail load in order to maintain security

Read More »

Energy Department Issues RFP to Advance President Trump’s 172-Million-Barrel Strategic Petroleum Reserve Exchange

WASHINGTON—The U.S. Department of Energy (DOE) today issued a Request for Proposal (RFP) for an exchange of up to 40 million barrels of crude oil from the Strategic Petroleum Reserve (SPR). Today’s solicitation opens competitive bidding, continuing DOE’s execution of President Trump’s 172-million-barrel release as part of a coordinated 400-million-barrel action by International Energy Agency (IEA) member nations’ strategic reserves. Under President Trump’s leadership, DOE has advanced an unprecedented series of large-scale SPR exchange solicitations at record speed. These actions have moved critical crude oil supplies into the market to address short term supply disruptions and bolster energy security for the United States and its allies. The crude oil will originate from the SPR’s Big Hill and Bryan Mound sites. This action builds on the Department’s four previous solicitations that collectively awarded more than 133 million barrels across three completed exchanges. DOE’s earlier exchanges demonstrated the SPR’s ability to rapidly deliver crude under emergency authorities while achieving a 26 percent premium in returned barrels—expanding the reserve at no additional cost to American taxpayers. “With today’s announcement, we are accelerating the President’s commitment to a coordinated and strategic release that stabilizes global oil markets,” said DOE Acting Assistant Secretary for the Hydrocarbons and Geothermal Energy Office Curt Coccodrilli. “This exchange will help move oil swiftly to refiners, ease short-term supply pressures, and ensure the Strategic Petroleum Reserve continues to grow stronger through the return of premium barrels.” Under DOE’s exchange authority, participating companies will return the 40 million borrowed barrels with additional premium barrels, ensuring immediate market supply while increasing the SPR’s long-term inventory. Bids for this solicitation are due no later than 11:00 A.M. Central Time on Monday, June 15, 2026. For more information on the SPR, please visit DOE’s website. 

Read More »

DOE’s Hydrocarbons and Geothermal Energy Office Invests $3.6 Million to Modernize America’s Coal-Fired Power Plants

WASHINGTON—The U.S. Department of Energy’s (DOE) Hydrocarbons and Geothermal Energy Office (HGEO) today announced $3.6 million for nine design and engineering projects that will support the refurbishment or retrofit of existing coal power plants with transformational technologies that address wastewater systems and improve the efficiency, reliability, flexibility, and performance of coal and natural gas use. By upgrading our nation’s existing coal facilities, these initiatives will help strengthen the backbone of America’s power grid and ensure all American’s have access to affordable, reliable, and secure energy when they need it most. These efforts help to advance President Trump’s Executive Orders Reinvigorating America’s Beautiful Clean Coal Industry and Strengthening the Reliability and Security of the United States Electric Grid to restore common-sense energy policies that prioritize dependable power, affordability, and American workers. “America’s coal fleet is an undeniable pillar of our energy dominance and economic strength, but for too long, policies have undermined this vital industry and the dedicated workforce behind it, threatening our grid’s stability and driving up costs for everyday Americans,” said DOE Acting Assistant Secretary of the Hydrocarbons and Geothermal Energy Office Curt Coccodrilli. “With the project investments announced today, we are decisively moving to champion our existing coal plants, ensuring they continue to deliver affordable, reliable power, keep the lights on, and fuel America’s progress for generations to come.” Projects have been selected under three topic areas to provide a path forward to rapidly and cost-effectively restore the stability of the nation’s bulk power system while also finding beneficial uses for wastes generated by coal-based energy production. The projects will be executed in three phases, with design and engineering completed in Phase I, final engineering and detailed design completed in Phase II, and technology implementation and validation completed in Phase III. Selectees to receive Phase I funding include: Baker Hughes Energy Transition LLC (Houston, Texas),

Read More »

Energy Department Releases Finalized Fusion Science and Technology Roadmap to Accelerate Commercial Fusion Power

WASHINGTON—The U.S. Department of Energy (DOE) today released the finalized Fusion Science and Technology (FS&T) Roadmap, a national strategy to accelerate the development and commercialization of fusion energy on the most rapid, responsible timeline in history. Building on earlier roadmap efforts, the finalized roadmap brings together fusion science, technology, infrastructure, workforce development, and commercialization priorities into a single national strategy to support fusion pilot plants and commercial fusion power in the mid-2030s. Fusion is the process that powers the sun and stars. For decades, scientists and engineers have worked to bring that same process to Earth as a source of abundant, reliable energy. The finalized roadmap outlines how DOE, industry, universities, and national laboratories will work together to accelerate the path toward commercial fusion energy in the United States. This effort advances President Trump’s energy dominance agenda and reinforces the Administration’s commitment to expanding reliable American energy production, strengthening domestic supply chains, and maintaining U.S. leadership in critical technologies. By accelerating progress toward commercial fusion power, DOE is helping secure a future of abundant and reliable energy. “Fusion energy has entered a new era defined by extraordinary scientific progress and public-private momentum,” said DOE Under Secretary for Science Dr. Darío Gil. “With this roadmap, we now have the clarity, coordination, and sustained commitment needed to turn the promise of fusion into a reality for the American people.” Developed with input from more than 800 scientists and engineers across the public and private sectors, the finalized FS&T Roadmap reflects contributions from more than 15 private companies, over 10 National Laboratories, and more than 70 universities. The roadmap identifies the critical science and technology gaps that must be closed to realize fusion pilot plants and strengthen U.S. leadership in the global fusion industry. The FS&T Roadmap establishes a unified strategy for the U.S.

Read More »

Amazon claims its data centers are 7x more water-efficient than the industry average

“Amazon is on the leading edge, but it’s not a secret recipe,” he said. What sets the company apart is scale, execution, facility design, geographic mix, and its aggressive pursuit of energy goals. Others are doing the similar things, if through different avenues: Microsoft is investing in closed-loop cooling systems that dramatically reduce evaporative water loss. Google is heavily focused on reclaimed water and using AI to optimize data centers. Meta has long relied on outside-air cooling. And overall, the industry is moving toward liquid cooling for dense AI deployments, “which changes the water equation again,” said Kimball. One of the big variables is location: Climate influences water efficiency, so where a company builds its infrastructure is as important as its cooling methods. Further, power-consumptive AI changes the discussion, he emphasized; traditional enterprise workloads and dense AI training clusters create very different thermal profiles.

Read More »

Marvell announces 102.4 Tbps switch silicon built for AI

Data movement has become an important concern in modern AI data centers. In the past, a cluster of a few servers could adequately handle back-office applications and databases. But with AI’s gigantic models, all sections of the data center need to move and receive data at high speeds. That requires a lot more power use than in the past. GPU- and XPU-based systems are approaching 120KW per rack, and switching and networking components consume approximately 15-25% of total rack power, making low-power switch silicon a strategic requirement. The Teralynx T100 delivers up to 25% lower power consumption than competitive solutions at a higher data rate. This enables AI infrastructures to deploy more accelerators within existing power envelopes without requiring additional power infrastructure. “As AI workloads evolve and scale exponentially, hyperscalers require network architectures that optimize latency, power and scalability simultaneously,” said Rishi Chugh, vice president and general manager of the data center switch business unit at Marvell, in a statement.

Read More »

From the data center to the edge: How to build secure, effective enterprise AI infrastructure

While hyperscalers and neo-cloud providers may get the lion’s share of attention for providing AI infrastructure, many enterprises are taking a build-it-themselves approach to meet their specific AI requirements. The success of such projects is crucial to achieving business objectives, yet companies face significant challenges as they try to scale pilots to production. Organizations must keep up with the dynamic, ever-changing demands that AI applications place on compute and network infrastructure, from the data center to the edge. That means architecting systems to grow as demand warrants and to avoid performance bottlenecks. The architecture must also account for AI-driven security vulnerabilities and ensure appropriate defenses are in place. Yes, it’s a tall order. But here, in simplified form, is a three-step plan for meeting those objectives. Step one: Go modular Integrating all the required components in piecemeal fashion for an AI factory is complex, costly, and fraught with integration risk. Start with a modular design, based on proven NVIDIA reference architectures. A modular approach combines pre-validated accelerated computing hardware, AI software, and orchestration platforms, as well as networking and storage capabilities. A modular strategy speeds implementation and creates a faster time to value for your AI infrastructure. Using modules that combine compute, networking, and storage makes it easier to scale capacity as needed, whether in the data center or at edge facilities. In addition, the modular approach simplifies the job of addressing varying requirements, from inferencing engines at the edge to massive-scale model training in the data center, while staying within the same solution family. The same applies to easing integration processes, as modular platforms offer pre-validated software. The Cisco Secure AI Factory with NVIDIA approach, for example, includes hardware (Cisco AI PODS) that is pre-validated to work with NVIDIA AI Enterprise software; Cisco Security and Splunk Observability software; orchestration

Read More »

OpenAI weighs Nvidia-backed lease for 10 GW Ohio data center campus

OpenAI would control the computing equipment under a 20-year lease and begin payments once the site starts operating, with the first phase expected in 2028. Nvidia is expected to supply the hardware and guarantee both OpenAI’s lease obligations and the developer’s financing, the report added. The reported structure highlights a broader shift in AI infrastructure strategy, where model developers, chip suppliers, and energy providers are forging increasingly long-term partnerships to secure compute capacity amid surging demand. “These types of symbiotic deals are becoming the norm as AI infrastructure rolls out,” said Neil Shah, vice president for research and partner at Counterpoint Research. “If a CIO picks OpenAI to be the base layer, they shouldn’t just accept whatever infrastructure comes with it. CIOs need to negotiate and demand that OpenAI uses a mix of capacity so all your eggs are not in one premium basket like Nvidia.” OpenAI and Nvidia did not immediately respond to requests for comment.

Read More »

Arista unveils 1.6T rack-scale switch family for AI infrastructure

The new Arista family joins a growing ecosystem of vendors looking to tap into the 1.6T Ethernet world, which includes Cisco, Nvidia, Celestica and others. “Arista Network’s new 7060XE7 Series is a strong signal of where large-scale AI fabrics are heading: higher bandwidth, better power efficiency, and tighter integration between compute, optics, silicon, cooling, and network operating software,” wrote Sameh Boujelbene, vice president, data center switch and AI networks market research for Dell Oro, in a LinkedIn post. Among the features that stand out to her are “strong customer and ecosystem validation from Microsoft Azure, Oracle Cloud Infrastructure, Meta, AMD, and Broadcom.”

Read More »

Water Emerges as a Critical Constraint for AI Data Centers

“There really has been a major shift within the last couple of years,” Bajpayee said. “I would even say within the last 12 months is where we have seen suddenly a rapid increase in the data center operators’ desire to control their water destiny.” For Gradiant, the MIT-born water technology company that built its reputation serving semiconductor manufacturers, pharmaceutical companies, and industrial customers worldwide, that shift has translated into a rapidly expanding pipeline of data center opportunities. More importantly, Bajpayee believes it signals a fundamental change in how the industry thinks about water itself. The conversation is no longer centered primarily on sustainability metrics or corporate environmental goals. Instead, operators increasingly view water as a business continuity issue. “We’re seeing operators themselves come to us and tell us that these are issues they are facing,” Bajpayee said. “They want to make sure they don’t get stalled, their permits don’t get pulled, their business doesn’t get stopped, and communities don’t push them out because they didn’t figure out a way to control their water.” From Water Treatment to Water Strategy That shift is occurring as Gradiant expands deployments of its recently announced HyperSolved platform, an end-to-end cooling water management system purpose-built for AI data centers. The company says HyperSolved is now being deployed with several of the world’s largest hyperscale operators across North America, Europe, and Asia, reflecting growing industry demand for integrated approaches to water infrastructure. While compute, networking, and power systems have evolved rapidly during the AI era, water management often remains fragmented, requiring operators to coordinate multiple vendors responsible for sourcing, treatment, cooling, wastewater management, reuse, discharge, and regulatory compliance. Gradiant’s approach seeks to consolidate those functions into a single integrated platform and operating model. The timing reflects the growing scale of the challenge. New AI data center

Read More »

Microsoft will invest $80B in AI data centers in fiscal 2025

And Microsoft isn’t the only one that is ramping up its investments into AI-enabled data centers. Rival cloud service providers are all investing in either upgrading or opening new data centers to capture a larger chunk of business from developers and users of large language models (LLMs).  In a report published in October 2024, Bloomberg Intelligence estimated that demand for generative AI would push Microsoft, AWS, Google, Oracle, Meta, and Apple would between them devote $200 billion to capex in 2025, up from $110 billion in 2023. Microsoft is one of the biggest spenders, followed closely by Google and AWS, Bloomberg Intelligence said. Its estimate of Microsoft’s capital spending on AI, at $62.4 billion for calendar 2025, is lower than Smith’s claim that the company will invest $80 billion in the fiscal year to June 30, 2025. Both figures, though, are way higher than Microsoft’s 2020 capital expenditure of “just” $17.6 billion. The majority of the increased spending is tied to cloud services and the expansion of AI infrastructure needed to provide compute capacity for OpenAI workloads. Separately, last October Amazon CEO Andy Jassy said his company planned total capex spend of $75 billion in 2024 and even more in 2025, with much of it going to AWS, its cloud computing division.

Read More »

John Deere unveils more autonomous farm machines to address skill labor shortage

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Self-driving tractors might be the path to self-driving cars. John Deere has revealed a new line of autonomous machines and tech across agriculture, construction and commercial landscaping. The Moline, Illinois-based John Deere has been in business for 187 years, yet it’s been a regular as a non-tech company showing off technology at the big tech trade show in Las Vegas and is back at CES 2025 with more autonomous tractors and other vehicles. This is not something we usually cover, but John Deere has a lot of data that is interesting in the big picture of tech. The message from the company is that there aren’t enough skilled farm laborers to do the work that its customers need. It’s been a challenge for most of the last two decades, said Jahmy Hindman, CTO at John Deere, in a briefing. Much of the tech will come this fall and after that. He noted that the average farmer in the U.S. is over 58 and works 12 to 18 hours a day to grow food for us. And he said the American Farm Bureau Federation estimates there are roughly 2.4 million farm jobs that need to be filled annually; and the agricultural work force continues to shrink. (This is my hint to the anti-immigration crowd). John Deere’s autonomous 9RX Tractor. Farmers can oversee it using an app. While each of these industries experiences their own set of challenges, a commonality across all is skilled labor availability. In construction, about 80% percent of contractors struggle to find skilled labor. And in commercial landscaping, 86% of landscaping business owners can’t find labor to fill open positions, he said. “They have to figure out how to do

Read More »

2025 playbook for enterprise AI success, from agents to evals

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More 2025 is poised to be a pivotal year for enterprise AI. The past year has seen rapid innovation, and this year will see the same. This has made it more critical than ever to revisit your AI strategy to stay competitive and create value for your customers. From scaling AI agents to optimizing costs, here are the five critical areas enterprises should prioritize for their AI strategy this year. 1. Agents: the next generation of automation AI agents are no longer theoretical. In 2025, they’re indispensable tools for enterprises looking to streamline operations and enhance customer interactions. Unlike traditional software, agents powered by large language models (LLMs) can make nuanced decisions, navigate complex multi-step tasks, and integrate seamlessly with tools and APIs. At the start of 2024, agents were not ready for prime time, making frustrating mistakes like hallucinating URLs. They started getting better as frontier large language models themselves improved. “Let me put it this way,” said Sam Witteveen, cofounder of Red Dragon, a company that develops agents for companies, and that recently reviewed the 48 agents it built last year. “Interestingly, the ones that we built at the start of the year, a lot of those worked way better at the end of the year just because the models got better.” Witteveen shared this in the video podcast we filmed to discuss these five big trends in detail. Models are getting better and hallucinating less, and they’re also being trained to do agentic tasks. Another feature that the model providers are researching is a way to use the LLM as a judge, and as models get cheaper (something we’ll cover below), companies can use three or more models to

Read More »

OpenAI’s red teaming innovations define new essentials for security leaders in the AI era

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More OpenAI has taken a more aggressive approach to red teaming than its AI competitors, demonstrating its security teams’ advanced capabilities in two areas: multi-step reinforcement and external red teaming. OpenAI recently released two papers that set a new competitive standard for improving the quality, reliability and safety of AI models in these two techniques and more. The first paper, “OpenAI’s Approach to External Red Teaming for AI Models and Systems,” reports that specialized teams outside the company have proven effective in uncovering vulnerabilities that might otherwise have made it into a released model because in-house testing techniques may have missed them. In the second paper, “Diverse and Effective Red Teaming with Auto-Generated Rewards and Multi-Step Reinforcement Learning,” OpenAI introduces an automated framework that relies on iterative reinforcement learning to generate a broad spectrum of novel, wide-ranging attacks. Going all-in on red teaming pays practical, competitive dividends It’s encouraging to see competitive intensity in red teaming growing among AI companies. When Anthropic released its AI red team guidelines in June of last year, it joined AI providers including Google, Microsoft, Nvidia, OpenAI, and even the U.S.’s National Institute of Standards and Technology (NIST), which all had released red teaming frameworks. Investing heavily in red teaming yields tangible benefits for security leaders in any organization. OpenAI’s paper on external red teaming provides a detailed analysis of how the company strives to create specialized external teams that include cybersecurity and subject matter experts. The goal is to see if knowledgeable external teams can defeat models’ security perimeters and find gaps in their security, biases and controls that prompt-based testing couldn’t find. What makes OpenAI’s recent papers noteworthy is how well they define using human-in-the-middle

Read More »
More Articles