Stay Ahead, Stay ONMINE

Teaching the model: Designing LLM feedback loops that get smarter over time

Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now Large language models (LLMs) have dazzled with their ability to reason, generate and automate, but what separates a compelling demo from a lasting product isn’t just the model’s initial […]

Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now


Large language models (LLMs) have dazzled with their ability to reason, generate and automate, but what separates a compelling demo from a lasting product isn’t just the model’s initial performance. It’s how well the system learns from real users.

Feedback loops are the missing layer in most AI deployments. As LLMs are integrated into everything from chatbots to research assistants to ecommerce advisors, the real differentiator lies not in better prompts or faster APIs, but in how effectively systems collect, structure and act on user feedback. Whether it’s a thumbs down, a correction or an abandoned session, every interaction is data — and every product has the opportunity to improve with it.

This article explores the practical, architectural and strategic considerations behind building LLM feedback loops. Drawing from real-world product deployments and internal tooling, we’ll dig into how to close the loop between user behavior and model performance, and why human-in-the-loop systems are still essential in the age of generative AI.

The prevailing myth in AI product development is that once you fine-tune your model or perfect your prompts, you’re done. But that’s rarely how things play out in production.


AI Scaling Hits Its Limits

Power caps, rising token costs, and inference delays are reshaping enterprise AI. Join our exclusive salon to discover how top teams are:

  • Turning energy into a strategic advantage
  • Architecting efficient inference for real throughput gains
  • Unlocking competitive ROI with sustainable AI systems

Secure your spot to stay ahead: https://bit.ly/4mwGngO


LLMs are probabilistic… they don’t “know” anything in a strict sense, and their performance often degrades or drifts when applied to live data, edge cases or evolving content. Use cases shift, users introduce unexpected phrasing and even small changes to the context (like a brand voice or domain-specific jargon) can derail otherwise strong results.

Without a feedback mechanism in place, teams end up chasing quality through prompt tweaking or endless manual intervention…  a treadmill that burns time and slows down iteration. Instead, systems need to be designed to learn from usage, not just during initial training, but continuously, through structured signals and productized feedback loops.


2. Types of feedback — beyond thumbs up/down

The most common feedback mechanism in LLM-powered apps is the binary thumbs up/down — and while it’s simple to implement, it’s also deeply limited.

Feedback, at its best, is multi-dimensional. A user might dislike a response for many reasons: factual inaccuracy, tone mismatch, incomplete information or even a misinterpretation of their intent. A binary indicator captures none of that nuance. Worse, it often creates a false sense of precision for teams analyzing the data.

To improve system intelligence meaningfully, feedback should be categorized and contextualized. That might include:

  • Structured correction prompts: “What was wrong with this answer?” with selectable options (“factually incorrect,” “too vague,” “wrong tone”). Something like Typeform or Chameleon can be used to create custom in-app feedback flows without breaking the experience, while platforms like Zendesk or Delighted can handle structured categorization on the backend.
  • Freeform text input: Letting users add clarifying corrections, rewordings or better answers.
  • Implicit behavior signals: Abandonment rates, copy/paste actions or follow-up queries that indicate dissatisfaction.
  • Editor‑style feedback: Inline corrections, highlighting or tagging (for internal tools). In internal applications, we’ve used Google Docs-style inline commenting in custom dashboards to annotate model replies, a pattern inspired by tools like Notion AI or Grammarly, which rely heavily on embedded feedback interactions.

Each of these creates a richer training surface that can inform prompt refinement, context injection or data augmentation strategies.


3. Storing and structuring feedback

Collecting feedback is only useful if it can be structured, retrieved and used to drive improvement. And unlike traditional analytics, LLM feedback is messy by nature — it’s a blend of natural language, behavioral patterns and subjective interpretation.

To tame that mess and turn it into something operational, try layering three key components into your architecture:

1. Vector databases for semantic recall

When a user provides feedback on a specific interaction — say, flagging a response as unclear or correcting a piece of financial advice — embed that exchange and store it semantically.

Tools like Pinecone, Weaviate or Chroma are popular for this. They allow embeddings to be queried semantically at scale. For cloud-native workflows, we’ve also experimented with using Google Firestore plus Vertex AI embeddings, which simplifies retrieval in Firebase-centric stacks.

This allows future user inputs to be compared against known problem cases. If a similar input comes in later, we can surface improved response templates, avoid repeat mistakes or dynamically inject clarified context.

2. Structured metadata for filtering and analysis

Each feedback entry is tagged with rich metadata: user role, feedback type, session time, model version, environment (dev/test/prod) and confidence level (if available). This structure allows product and engineering teams to query and analyze feedback trends over time.

3. Traceable session history for root cause analysis

Feedback doesn’t live in a vacuum — it’s the result of a specific prompt, context stack and system behavior. l Log complete session trails that map:

user query → system context → model output → user feedback

This chain of evidence enables precise diagnosis of what went wrong and why. It also supports downstream processes like targeted prompt tuning, retraining data curation or human-in-the-loop review pipelines.

Together, these three components turn user feedback from scattered opinion into structured fuel for product intelligence. They make feedback scalable — and continuous improvement part of the system design, not just an afterthought.


4. When (and how) to close the loop

Once feedback is stored and structured, the next challenge is deciding when and how to act on it. Not all feedback deserves the same response — some can be instantly applied, while others require moderation, context or deeper analysis.

  1. Context injection: Rapid, controlled iteration
    This is often the first line of defense — and one of the most flexible. Based on feedback patterns, you can inject additional instructions, examples or clarifications directly into the system prompt or context stack. For example, using LangChain’s prompt templates or Vertex AI’s grounding via context objects, we’re able to adapt tone or scope in response to common feedback triggers.
  2. Fine-tuning: Durable, high-confidence improvements
    When recurring feedback highlights deeper issues — such as poor domain understanding or outdated knowledge — it may be time to fine-tune, which is powerful but comes with cost and complexity.
  3. Product-level adjustments: Solve with UX, not just AI
    Some problems exposed by feedback aren’t LLM failures — they’re UX problems. In many cases, improving the product layer can do more to increase user trust and comprehension than any model adjustment.

Finally, not all feedback needs to trigger automation. Some of the highest-leverage loops involve humans: moderators triaging edge cases, product teams tagging conversation logs or domain experts curating new examples. Closing the loop doesn’t always mean retraining — it means responding with the right level of care.


5. Feedback as product strategy

AI products aren’t static. They exist in the messy middle between automation and conversation — and that means they need to adapt to users in real time.

Teams that embrace feedback as a strategic pillar will ship smarter, safer and more human-centered AI systems.

Treat feedback like telemetry: instrument it, observe it and route it to the parts of your system that can evolve. Whether through context injection, fine-tuning or interface design, every feedback signal is a chance to improve.

Because at the end of the day, teaching the model isn’t just a technical task. It’s the product.

Eric Heaton is head of engineering at Siberia.

Shape
Shape
Stay Ahead

Explore More Insights

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

Alkira advances NaaS for the agentic AI era

The practical difference spans a number of areas including the format of responses. Rather than returning raw JSON that requires parsing and interpretation, the MCP Server can deliver tabular summaries. An operator or AI agent can request a deployment overview and receive structured data showing region counts, segment configurations, connector

Read More »

Cato extends SASE platform to unmanaged devices

Cato Networks this week introduced the Cato Browser Extension that expands the company’s secure access service edge (SASE) platform and its universal zero-trust network access (ZTNA) capabilities to unmanaged devices and distributed bring-your-own-device (BYOD) endpoints. The Cato Browser Extension, according to the company, is a lightweight onramp to its SASE

Read More »

Energy Department Selects Four Companies for Advanced Nuclear Fuel Line Pilot Projects

WASHINGTON— The U.S. Department of Energy (DOE) today announced another step forward in the Trump administration’s efforts to strengthen domestic supply chains for nuclear fuel. DOE selected Oklo Inc., Terrestrial Energy Inc., TRISO-X LLC, and Valar Atomics Inc. for its new pilot program to build advanced nuclear fuel lines. Today’s action will help strengthen America’s national security, reduce reliance on foreign sources of enriched uranium and support the Department’s Reactor Pilot Program that aims to have at least three reactors achieve criticality by July 4, 2026. “President Trump has made clear that a strong nuclear sector is a central component of America’s energy security and prosperity,” said Deputy Secretary of Energy James P. Danly. “Restoring a secure domestic fuel supply will ensure that advanced reactors can move quickly from design to deployment and into operation. The ability to produce these fuels is essential to ensuring American leadership in nuclear energy and to meeting the nation’s growing demand for reliable power.” This is the second round of conditional selections under DOE’s Fuel Line Pilot Program. DOE previously selected Standard Nuclear to build and operate TRISO fuel fabrication facilities. The following projects will leverage the Department’s authorization process to ensure a robust supply of fuel is available for research, development, and demonstration purposes— including the 11 reactors initially selected to participate in DOE’s Reactor Pilot Program: Oklo Inc. (Santa Clara, California) – To build and operate three fuel fabrication facilities to support their Aurora and Pluto reactors, and possibly other fast reactors.  Terrestrial Energy Inc (Charlotte, NC) – To develop the Terrestrial Energy Fuel Line Assembly to demonstrate a fuel salt fabrication process in a phased approach.  TRISO-X Inc. (Oak Ridge, Tennessee) – To build and operate an additional fuel fabrication laboratory facility to enable pilot-scale integration, training, and system validation to

Read More »

Treasuries Gain as USA Gov Shutdown May Follow Month-End

Treasuries gained Monday, supported by a slump in oil prices and a rally in UK government bonds, and by anticipation of buying into Tuesday’s month-end index rebalancing. Yields declined as much as five basis points across tenors with the 10-year note’s falling to 4.14%. The 30-year bond’s dipped below 4.70% for the first time since Sept. 18. US benchmark crude oil futures dropped more than 4% on signs OPEC+ will hike production again in November. The prospect of a US government shutdown beginning Wednesday also has implications for the Treasuries market, as shutdowns are associated with gains for bonds based on their potential to restrain the economy. “There’s a global move lower today in yields,” said Angelo Manolatos, an interest-rate strategist at Wells Fargo Securities. “It’s likely a combination of quarter-end flow dynamics and the possibility of a government shutdown. Yields typically drop modestly during government shutdowns that last at least five days.” The market racked up gains even as Cleveland Fed President Beth Hammack — who becomes a voting member of the central bank’s rate-setting committee next year — reiterated her view that inflation remains too high to warrant cutting interest rates. Futures markets continue to anticipate about 100 basis points of additional Fed easing over the next 12 months. Expectations for Fed rate cuts rest mainly on signs of stress in the US labor market, where job creation has slowed precipitously in recent months. September data is set to be released on Friday. Tuesday’s month-end bond index rebalancing — to add eligible bonds created during the month and remove those that no longer fit the index criteria — typically drives buying by passive and other index-tracking investment funds that can support the market if their needs exceed expectations.  The rebalancing will increase the duration of the Bloomberg Treasury index by an estimated 0.06

Read More »

Russia Expects Oil and Gas Revenue at Lowest in Years

The Russian government expects the oil and gas industries’ contributions to the national budget this year to drop to the lowest since the pandemic of 2020 after prices for its fuels dipped and the ruble strengthened. Moscow will gather about 8.65 trillion rubles ($100 billion) in taxes from the oil and gas industry, according to amendments to Russia’s 2025 budget. That’s about 22 percent less than last year’s revenues, according to the amendments, published Monday on the website of the State Duma, the lower chamber of the parliament.  Levies from oil and gas producers are critical for the Kremlin, as they are projected to account for almost a quarter of revenues into state coffers this year. Moscow plans to keep boosting military spending to finance the ongoing war in Ukraine, and will raise taxes like VAT and increase borrowing to bridge the budget gap. In a drive to further bear down on Russia’s energy revenues, western nations and allies have adopted wide-ranging sanctions. Most recently, US President Donald Trump has been putting pressure on partners in the North Atlantic Treaty Organization, including Turkey, to stop buying Russia’s barrels altogether.  Urals, Russia’s main export blend, is projected to average $58 a barrel this year, compared with $66.60 last year. Despite sanctions, the decline is mainly driven by lower crude prices amid concerns over global economic growth.  Russia’s government sees the average Urals discount to global benchmark Brent at $12 a barrel. The gap has shrunk compared with the earlier years of the war, but is still markedly wider than the historic discount of $2-4 a barrel because of sanctions.  The strengthening of the national currency is another reason behind declining revenues. For this year, the government projects the exchange rate at 86.1 rubles to a US dollar, compared with 92.4 rubles a dollar

Read More »

Exxon to Cut 2,000 Jobs

Exxon Mobil Corp. plans to cut about 2,000 jobs globally as the Texas oil company consolidates smaller offices into regional hubs as part of its long-term restructuring plan.  The reductions represent about 3% to 4% of Exxon’s global workforce and are part of the company’s ongoing efficiency drive, Chief Executive Officer Darren Woods said in an memo to employees Tuesday. About half will be in the Europe and most of rest in Canada at Calgary-based Imperial Oil Ltd., which is nearly 70% owned by Exxon. Chevron Corp., ConocoPhillips and BP Plc are among major oil companies to have also announced thousands of job cuts in recent months as crude prices tumbled this year in response to increased supplies from OPEC and its allies. Exxon, however, has been on a major internal restructuring push since 2019 as Woods sought to simplify the company’s sprawling global footprint that came as a result of the merger with Mobil two decades ago.  Exxon is making “tough decisions” that build upon a years-long effort to improve competitiveness, Woods wrote in the memo. “The changes we’ve announced today will further strengthen our advantages and grow the gap with our competition, helping to keep us in the lead for decades to come,” he said.   The oil giant will cut 1,200 positions in the European Union and Norway by the end of 2027, with layoffs making up half of the reductions, it said in a statement. Imperial will cut about 900 positions, or 20% of its workforce, in the same time period, helping reduce operational expenses by C$150 million ($108 million) annually.  The regional hubs will focus on Exxon’s major growth initiatives such as oil in Guyana, liquefied natural gas along the Gulf Coast and trading globally. For example, the company recently announced plans to move employees from Brussels and Leatherhead, UK, to central London, where it’s recently

Read More »

Democratic House bill would reverse Trump energy policies, bolster RTO oversight

Democratic Reps. Sean Casten of Illinois and Mike Levin from California are preparing to introduce sweeping energy legislation that would reverse Trump administration policies, including by restoring clean energy tax credits and limiting the U.S. Department of Energy’s ability to declare “energy emergencies” to keep fossil-fueled power plants from retiring. The Trump administration is driving up the cost of electricity by creating barriers to clean energy investment to support higher-cost fossil fuel sectors, according to Casten. “You’ve now got this surging demand [from data centers and other loads] — and particularly if we’re not going to have an administration that’s going to encourage competitive markets — then that means that the only people who are going to be able to build stuff are regulated utilities with mandatory capital amortization and so it’s very hard to see anything but upward pressure on electric prices,” Casten said Monday in an interview. The draft Cheap Energy Act aims to put downward pressure on electricity bills by bolstering energy efficiency efforts, supporting grid-enhancing technologies and helping provide lower-cost renewable energy projects access to the grid, according to Casten. One section of the legislation would increase oversight of regional transmission organizations. “Everybody in the electric regulatory world, including lots of [Federal Energy Regulatory Commission] commissioners … and frankly a lot of utility executives, knows that there’s a massive governance problem in the RTO space,” Casten said. Governors in the PJM Interconnection footprint, for example, are pressing for a role in the RTO’s decisionmaking. Provisions in the bill would require FERC to reform the governance and stakeholder participation practices of RTOs and independent system operators, according to a summary of the draft bill. They would also require independent transmission monitors to facilitate the cost-effective construction of transmission. Partly, those provisions would make it easier for outsiders

Read More »

EPA mulls postponing coal plant wastewater compliance, changes to Regional Haze Rule

Dive Brief: The U.S. Environmental Protection Agency on Monday proposed giving coal power generators additional time to comply with new wastewater disposal guidelines and said it also would consider regulatory changes to the Clean Air Act’s Regional Haze Rule. EPA’s proposals are part of a package of coal-supporting actions the Trump administration unveiled this week, including $625 million to retrofit and recommission coal plants from the Department of Energy. Extending compliance for EPA’s effluent limitations guidelines would reduce electricity costs by approximately $30 million to $200 million annually, the agency estimated. Sierra Club Beyond Coal Campaign Director Laurie Williams countered that every day the requirements are delayed means more people “will be exposed to higher levels of toxic pollution.” Dive Insight: Residual waste from burning coal contains contaminants including mercury, cadmium and arsenic, according to the EPA. The federal government strengthened regulation on coal combustion residuals following a disastrous coal ash spill from the Tennessee Valley Authoriy’s Kingston plant in 2008, which contaminated some 300 acres of land, making it one of the largest industrial spills in U.S. history.  EPA revised its ELG requirements last year to include stronger protections around coal ash wastewater pollution. The changes included requiring plants to halt some types of discharges by 2029 or to commit by the end of this year to cease burning coal by 2034. EPA’s proposal would extend both the compliance and the notice deadlines. The agency said the move would “reduce costs for facilities and help with electricity reliability and affordability.” Comments are due 30 days after the proposal is published in the Federal Register. EPA’s current ELGs for wastewater discharges from steam electric power plants “are potentially costly to an electric power sector that is struggling with increasing demand as AI is booming, data centers are being constructed and operated around

Read More »

New Partnerships Form to Deploy Digital Energy Twins in Data Centers

Of course, partnerships have been ongoing in the data center and AI realms. For example, Microsoft and AMD have a long-standing collaboration to provide powerful cloud solutions for enterprise workloads, high-performance computing, and AI. Google Cloud and NVIDIA have a deep partnership across the AI stack, allowing Google Cloud to integrate NVIDIA’s latest hardware and software into its services. Colocation providers are collaborating with tech companies to offer turnkey AI solutions, which hardware, AI research labs, and cloud providers are joining forces in cross-industry alliances to support AI development. What’s Next for Digital Energy Twins Analysts anticipate that the next wave of digital energy twins will feature deeper integration with AI and the Internet of Things (IoT) for self-updating, autonomous operations, and enhanced predictive capabilities, moving beyond monitoring to proactive simulation and optimization. Key developments include increased integration with quantum computing, the development of interoperable and standardized ecosystems, and a greater focus on cybersecurity to protect these complex systems. These advancements aim to drive greater operational efficiency, improve asset performance, reduce downtime, and support sustainability goals across the energy sector.

Read More »

South Korea’s data center fire triggers global scrutiny of lithium-ion batteries and DR architecture

Kasthuri Jagadeesan, research director at Everest Group, said enterprises should audit centralization risks by mapping interdependencies across UPS, cooling, and shared power zones. “The NIRS case illustrates that redundancy alone cannot protect against weak compartmentalization,” she said, noting that outages can cost millions per hour. “Geographic redundancy is only effective if failover processes are tested and staff are trained to execute under pressure,” Jaura said. “CIOs must validate that DR plans are operational, not theoretical. This means regular, realistic testing, cross-functional engagement, and continuous improvement.” IDC research shows that centralized facilities offer economies of scale but concentrate risk, while distributed and modular approaches enhance resilience and reduce single points of failure, according to Jaura. “Diversify datacenter locations to mitigate regional risks,” he advised. “Invest in modular and mobile data center solutions for flexibility and rapid recovery.” Market implications Rai said the incident may instigate “heightened due diligence and a more cautious pace of adoption,” but lithium-ion technology’s advantages remain compelling. “What is likely to change is that enterprises will demand stronger safety certifications and vendor accountability, and accelerate investment in disaster recovery planning, geographical redundancy, and resilience frameworks.” Kalyani Devrukhkar, senior analyst at Everest Group, said both enterprises and regulators will be more demanding about safety standards. “Some organizations may look at alternatives like sodium-ion or advanced valve-regulated lead-acid, and insurers will almost certainly raise premiums where risk is seen as high,” she said, noting that NFPA 855 and newer International Fire Code editions now include stricter requirements for lithium-ion battery systems. “Enterprises are increasing budgets for business continuity management, IT disaster recovery planning, and alternative site management,” Jaura said. For CIOs, Jaura recommended a business impact analysis-driven framework that balances efficiency with safety and compliance. “The decision is not binary — mitigation such as advanced monitoring, fire suppression, and

Read More »

Zayo launches DynamicLink NaaS platform with core-based service delivery

At customer locations, only a lightweight demarcation device is required. This equipment creates a clean handoff point but contains no processing logic or intelligence. All routing, security, traffic management and network services run on the pre-deployed hardware in Zayo’s facilities. “We’ve basically taken the logic out of those boxes that used to sit on premise, sucked that into the core of the network, and then put it on hardware that can do extremely fast multi-terabyte level, single-packet processing. You know, line rate, speed performance,” Long explained. Service model: One port, multiple functions DynamicLink’s commercial model centers on port-based pricing. Organizations purchase ports at locations where they need connectivity. This includes headquarters, branch offices, colocation facilities, cloud on-ramps or any location Zayo’s network reaches. Each port provides flexible capacity that can be allocated across different services. A single 10 Gbps port might simultaneously support point-to-point Ethernet between facilities, dedicated internet access and cloud connectivity. Customers reconfigure these allocations through a self-service portal. “If you have a dynamic link port, that, if you have, like, a dynamic link in three different locations, let’s say it’s at a headquarters location and two of your data centers, you can use that to have a point to point Ethernet,” Long said. “If you want to use those 10 gigs at one location to go out to the internet, you can use it to go out to the internet.” Previously, each of these functions required ordering a separate service. An organization wanting dedicated internet access ordered a DIA port. Point-to-point Ethernet between data centers required two separate ports. Each service involved manual provisioning and fixed configuration.

Read More »

Scaling Stargate: OpenAI’s Five New U.S. Data Centers Push Toward 10 GW AI Infrastructure

Stargate is OpenAI’s massive AI infrastructure initiative, developed as a joint venture in partnership with Oracle and SoftBank. Formally announced in January 2025, the program is accelerating rapidly with the disclosure of five new U.S. data center sites. These additions—along with the flagship development in Abilene, Texas, and other ongoing projects—bring Stargate’s total planned capacity to nearly 7 gigawatts (GW). The cumulative investment estimate has now topped $400 billion as the program heads toward its ultimate goal: a 10 GW, $500 billion buildout. While the initiative focuses on building capacity with non-Microsoft partners, Microsoft remains a key technology partner and OpenAI’s primary cloud provider (Azure). Where Are the Five New Sites? The next wave of Stargate capacity is landing in regions already familiar with large-scale data center development. Based on public reporting and company statements, the five identified sites are: Shackelford County, Texas (greater Abilene expansion): An extension of the area already hosting Vantage Data Centers’ Frontier project, a $25 billion development on 1,200 acres. Milam County, Texas (Central Texas growth corridor): Previously announced as the home of a SoftBank-led Stargate data center campus. Doña Ana County, New Mexico (Las Cruces area): Linked to Project Jupiter, a proposed $165 billion build spearheaded by BorderPlex Digital Assets, with Stack Infrastructure reported as a potential participant. Lordstown, Ohio (Eastern PJM/FirstEnergy territory): Redevelopment of a former GM/Foxconn complex, being repositioned as a large-scale AI campus through a collaboration between OpenAI, Oracle, and SoftBank. An additional Midwest site (TBD): Location yet to be disclosed. These builds are being advanced under partnership models, with Oracle expected to lead three of the sites and SoftBank/SB Energy two. Together, they reinforce Stargate’s path toward a 10 GW national roadmap. Scale and Performance Goals With the addition of the five new campuses, plus Abilene and other previously announced

Read More »

Q3 Executive Roundtable Recap

AI-scale workloads are reshaping the fundamentals of data center design. For Data Center Frontier’s Q3 2025 Executive Roundtable, three industry leaders tackled the most urgent challenges: managing thermal and water risk at scale, balancing CapEx vs. OpEx in the race to build, and breaking down silos as cooling, water, and power systems converge. <!–> Sept. 26, 2025 3 min read –>

Read More »

‘Nomads at the Summit’ Podcasts – Recorded Live at DCF Trends Summit 2025

@import url(‘https://fonts.googleapis.com/css2?family=Inter:[email protected]&display=swap’); a { color: #1796c1; } .ebm-page__main h1, .ebm-page__main h2, .ebm-page__main h3, .ebm-page__main h4, .ebm-page__main h5, .ebm-page__main h6 { font-family: Inter; } body { line-height: 150%; letter-spacing: 0.025em; font-family: Inter; } button, .ebm-button-wrapper { font-family: Inter; } .label-style { text-transform: uppercase; color: var(–color-grey); font-weight: 600; font-size: 0.75rem; } .caption-style { font-size: 0.75rem; opacity: .6; } #onetrust-pc-sdk [id*=btn-handler], #onetrust-pc-sdk [class*=btn-handler] { background-color: #1796c1 !important; border-color: #1796c1 !important; } #onetrust-policy a, #onetrust-pc-sdk a, #ot-pc-content a { color: #1796c1 !important; } #onetrust-consent-sdk #onetrust-pc-sdk .ot-active-menu { border-color: #1796c1 !important; } #onetrust-consent-sdk #onetrust-accept-btn-handler, #onetrust-banner-sdk #onetrust-reject-all-handler, #onetrust-consent-sdk #onetrust-pc-btn-handler.cookie-setting-link { background-color: #1796c1 !important; border-color: #1796c1 !important; } #onetrust-consent-sdk .onetrust-pc-btn-handler { color: #1796c1 !important; border-color: #1796c1 !important; background-color: undefined !important; } Welcome to Nomads at the Summit, a new podcast series from Data Center Frontier in partnership with the Nomad Futurist Foundation. Recorded live at the 2025 Data Center Frontier Trends Summit (Aug. 26-28), here we sit down with industry leaders, innovators, and change-makers shaping the future of digital infrastructure. Join hosts Nabeel Mahmood and Phillip Koblence of Nomad Futurist, alongside DCF editorial leadership including Editor at Large Melissa Farney and Senior Editor David Chernicoff, for these candid conversations that highlight the ideas, talent, and technologies driving the next chapter of the data center industry. Whether you attended the DCF Trends Summit in person or are just now tuning in from afar, Nomads at the Summit gives you a behind-the-scenes look at the people and innovations defining what’s next in digital infrastructure. <!–> EPISODE LIST ]–> Waste Heat to Water – The Path Towards Water Positive Data Centers In this DCF Trends-Nomads at the Summit Podcast episode, Matt Grandbois, Vice President at AirJoule, introduces a game-changing approach to one of the data center industry’s most pressing challenges: water sustainability. As power-hungry, high-density environments collide with growing water scarcity

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 »