Computerworld.com [Hacking News]

China has announced immediate export controls on seven more rare earth elements critical to enterprise IT hardware manufacturing, firing a fresh salvo in the ongoing tech trade war. This move could significantly impact tech giants including Dell Technologies, HP, Apple, and IBM, along with semiconductor leaders such as Intel, Samsung, and TSMC.

The new controls issued by China’s State Council require export licenses for samarium, gadolinium, terbium, dysprosium, lutetium, scandium, and yttrium — along with their alloys, oxides, and compounds. These materials are essential components in data center storage systems, networking equipment, and semiconductors.

Over the weekend, Meta took the opportunity to launch Llama 4, a new series of AI models trained on a large amount of text, images and videos.

According to Meta, Llama 4 is better than its competitors GPT-4o and Gemini 2.0 in a number of areas, including programming, reasoning and language translation.

The two variants Llama 4 Scout and Llama 4 Maverick are available on Llama.com and Hugging Face now, while the top-of-the-line Llama 4 Behemoth will take a little longer.

Techcrunch points out that Meta has chosen to limit the use of the Llama 4 multimodal models within the European Union, likely due to the EU’s AI and data protection rules.

Anthropic plans to add more than 100 roles across Europe, indicating a broader push to expand its global footprint amid rising demand for enterprise-ready AI tools.

The new roles will cover sales, engineering, research and business operations, with most hiring focused in Dublin and London, according to a Reuters report.

The company has also named former Stripe executive Guillaume Princen as its head of Europe, the Middle East, and Africa.

Last month, Anthropic raised $3.5 billion in funding. In February, it launched Claude 3.7 Sonnet, its most advanced model to date and the first to feature hybrid reasoning capabilities.

Targeting expansion and localization

Anthropic’s European hiring spree reflects its broader ambitions for global expansion – a move that mirrors a wider industry trend among well-funded AI firms, according to analysts.

“Every AI startup that has raised $1 billion is competing for global domination,” said Hyoun Park, CEO and chief analyst at Amalgam Insights. “Anthropic’s recent $3.5 billion round is no exception, as it seeks to be the leading model company for the Americas, Europe, Asia, and Africa. And to enable this growth, it makes sense that these European hires mainly focus on sales, marketing, and engineering.”

The company’s regional push also signals the growing importance of localization in AI development – especially as regional regulations and incentives shape how technology is built and deployed.

“Hiring locally in Europe could be to ensure that its offerings down the line are tuned to European requirements that may not fully align with that of the US – particularly related to AI governance,” said Abhishek Sengupta, practice director at Everest Group. “There could also be potential business incentives aligned to local operations as more and more countries announce funds for developing a local AI ecosystem.”

Implications on EU AI adoption

Beyond expansion, Anthropic’s growing presence in Europe may signal the start of direct competition with firms positioning themselves as regionally compliant AI providers.

Anthropic is “confronting” Cohere head-on, Park said, pointing out that the Canadian startup has been actively marketing itself as a European-based, EU-compliant alternative.

“But this market competition may be a gift in disguise, as the EU has been famously aggressive in attacking monopoly,” Park said. “Having anthropic more deeply focused on the EU may give more freedom for Cohere, OpenAI, Microsoft, Google, and other AI firms to be able to build business in what is perceived as a competitive market.”

For AI companies with global ambitions, building a meaningful presence in the EU is no longer optional. Park noted that Anthropic’s more deliberate and safety-focused approach may align better with the demands of highly regulated markets.

“It seems that in the immediate future, the US will be a country of less AI regulation where solutions such as Meta, OpenAI, and xAI may have an advantage,” Park added. “It also seems likely that the EU will be driving a lot of global first world AI policy in the near future, almost by default as other first world global centers are reluctant to define AI standards and governance.”

As generative AI continues to dominate technology news, the hype around AI PCs is growing. It’s easy to see the appeal of personal computers powerful enough to handle genAI processing tasks.

Offered by all the major PC vendors, AI PCs include hardware and software designed to accelerate AI workloads across the central processing unit (CPU), graphics processing unit (GPU), and neural processing unit (NPU). AI is primarily a server-side application, but AI PCs offload at least some of the processing to the client.

[ Related: More AI PC news and insights ]

But these are very early days for the AI PC market, and IT buyers may be cautious, wondering why they should invest in AI PCs when so many genAI tools are accessible through the web.

The short answer is you’re going to get them whether you want them or not, because all business PCs are headed in that direction. In time, AI features will be standard issue just like Wi-Fi and vPro are now. But that’s not the case just yet, so the question is whether to move ahead with AI PC purchases now or hold off until the market matures.

What AI PCs bring to the table

The primary argument for running AI applications on the desktop instead of in the cloud is increased privacy and security. Especially in highly regulated sectors like finance and healthcare, sensitive data might have to stay within the confines of your firewall, so you want to process it locally, not on a cloud provider’s servers.

Another key consideration is cost. Server-based AI processing is extremely expensive, costing anywhere from $50,000 to millions of dollars for in-house large language model (LLM) rollouts. Even using AI-as-a-service (AIaaS) can cost tens of thousands of dollars per month, and you have the aforementioned privacy and security concerns around sending your data to the cloud. Shifting part of the AI workload onto endpoint devices, including PCs, should reduce the need for server-based processing — and speed up response times as well.

AI PCs are more than a marketing gimmick — they are a means to improved productivity, argues Steve Long, senior vice president of Lenovo’s Intelligent Devices Group. “AI PCs give time back to the user,” he said. “Whether it’s generating marketing content, summarizing meetings, or streamlining workflows with apps like Microsoft 365 Copilot or Personal.ai, these devices handle the heavy lifting so teams can stay focused on higher-value work. The result is real productivity.”

AI PCs also have proactive and autonomous functions like real-time threat detection that runs on the device without slowing it down, he says, and they offer predictive maintenance, which helps IT stay ahead of failures.

Finally, Long says AI PCs make for a more personalized experience because these devices adapt to how you work, such as summarizing meetings in your tone, writing emails with M365 Copilot, or surfacing relevant insights using personal AI agents from previous projects.

“It’s like having a digital assistant who gets smarter every day and never sleeps, and because it all happens locally, it’s fast, secure, and tailored to you,” he said.

Should you buy AI PCs this year?

That’s the million-dollar question. New devices need a killer app to justify their purchase. The IBM PC didn’t take off as a business tool until Lotus 1-2-3 came out. AI PCs need their own must-have app.

Gartner predicts that by end of 2026, all PCs will be AI PCs, but not everyone will be using the AI features. “The use of on-device AI features as distinct from cloud-based will evolve more slowly, and enterprises are likely to phase them in over time, to ensure compliance rules are met. Therefore, Gartner estimates only around 24% of employees will use enhanced NPU capabilities by 2028,” said Ranjit Atwal, senior research director at Gartner.

Lenovo’s Long predicts it’s more likely to be 2027 or 2028 before every PC is an AI PC. “It is a little bit of the wild west of experimentations, with people trying to find the killer use case right now, both in enterprise and consumer,” he said.

Lacking a killer app, AI PCs are tough sell. They are not low-end devices. Atwal estimates the price difference between a standard business PC and an AI PC is $200 or more. Multiply that by dozens, hundreds, or even thousands of PCs and the cost really adds up — and for what?

AI vendors will argue that AI PCs provide future-proofing. Atwal does not. “If buyers don’t see the immediate value, they are not prepared to pay the premium,” he said. “Nobody pays for future-proofing, really — especially that amount — when you don’t know exactly what [the benefit] is.”

Another important factor for IT buyers: Windows 10 will reach end of support in October. Organizations with a fleet of Windows 10 PCs will either have to purchase extended support for those devices or upgrade their users to new Windows 11 devices.

If you’re shelling out for new PCs anyway, it might make sense to spend a little more for devices with AI capabilities. Indeed, a recent IDC survey of IT decision-makers from large firms around the world found that 80% of companies plan to deploy AI PCs in 2025.

One company that’s not waiting to jump into the AI PC pool is Transmission, a global B2B marketing and advertising agency. The firm is deploying the new devices companywide to get its staff familiar with AI tools now rather than later.

“We understand that this is the future, and we want to get them to start having some exposure to the tools, getting them familiar with how they can make better and quicker decisions for themselves, but also for clients,” said Jonathan Cocek, global analytics lead at Transmission.

“We’ve initially launched [Microsoft] Copilot onto everybody’s machine, and really it’s just trying to enable them to get them familiar with using AI, how to stay ahead of the curve.”

What to look for when purchasing AI PCs

If you decide to take the plunge with AI PCs now, the configuration choices you make will affect your users’ experience. So what should you look for? There are several checkboxes to fill.

1. An AI-ready processor. AI PCs are equipped with an NPU designed to speed up the processing of machine learning and deep learning tasks and optimized for operations like matrix multiplications, convolutions, and other types of computations commonly used in AI algorithms.

A NPU is part of the CPU and only found on the newest generations of CPUs:

• Intel Core Ultra (Meteor Lake generation)
• AMD Ryzen 7040 series
• Apple M-Series chips in Macs
• Qualcomm Snapdragon X Elite ARM-based chips for Windows AI PCs

2. Top TOPS performance. NPUs are measured by a benchmark called TOPS, or Tera Operations Per Second. TOPS means how many trillions of operations can be done in one second. Forget GHz and other numbers, TOPS is the measure to live by.

An AI PC has three processing units: the CPU, the GPU, and the NPU. The NPU is the most important, but the CPU and GPU also play a part. Intel’s Lunar Lake family of desktop processors have a TOPS rating of 120, which breaks down to 48 TOPS for the NPU, 67 TOPS for the GPU, and 5 TOPS for the CPU.

The NPUs in AMD’s Ryzen AI MAX series of processors have a TOPS rating of 50 — virtually tied with Intel — but AMD does not break out the CPU and GPU performance.

The one thing you can be certain of is that future generations of CPUs will feature considerably faster NPU performance. The generation of Intel processors prior to Lunar Lake featured a TOPS rating of only 10.

3. A good GPU. Business PCs often have integrated graphics on the CPU die, which are not very high performance. AI processors have much better performing GPUs than previous generations, but they are still no match for a discrete GPU. The only challenge is finding apps written to use the discrete GPU. A few exceptions like CAD/CAM software aside, most line-of-business applications don’t utilize the GPU.

But eventually AI apps will be rewritten to utilize the discrete GPU, because discrete GPUs for gaming have incredible TOPS performance. Nvidia’s new top-of-the-line RTX 5090 has a TOPS rating of 3350, for example. It also costs $4,100. Should you decide to go with a discrete GPU from either AMD or Nvidia, the rule of thumb is that the higher the model number, the faster and newer it is.

4. Lots of memory. With traditional PCs, most business users can get away with 8GB of system memory, especially if they use a lot of SaaS apps and live in their browser. But if your employees are going to do AI processing, they’re going to need a lot more memory — 16GB, if not 32GB or even 64GB, depending on the size of the model.

5. Integrated AI software. PCs should be compatible with AI frameworks such as TensorFlow and PyTorch and tools like Windows Studio Effects, Adobe Firefly, and other useful AI applications. Software vendors continue to optimize their applications to leverage local AI processing, so having hardware that supports standard tools and frameworks is critical to accelerate AI-enabled apps. 

Intel is reportedly in advanced discussions with TSMC to form a joint venture that could potentially reshape the future of the x86 platform and the global semiconductor landscape.

Under a proposed deal, TSMC would take a 20% stake in a new entity operating Intel’s chipmaking facilities, in exchange for training Intel personnel in its industry-leading manufacturing processes, according to The Information. 

The move — initiated in part by the US government — marks a potential turning point for Intel, which has struggled in recent years to keep up with rivals in advanced chip production.

If finalized, the partnership could allow Intel’s new CEO, Lip-Bu Tan, to double down on chip design innovation while offloading manufacturing operations to the Taiwanese foundry giant.

For enterprise tech buyers, the implications could be far-reaching, touching everything from processor roadmaps and platform stability to supply chain sovereignty and pricing.

But the proposal is reported to be already generating internal resistance at Intel and raising questions about layoffs, equipment overhauls, and the fate of its manufacturing IP issues that could reshape how enterprise customers view their long-term relationship with one of the most iconic names in computing.

Long-term strategic implications

Analysts say the move marks a turning point for Intel and its enterprise customers. Once seen as a vertically integrated leader, the company has struggled in recent years to maintain that position.

“Intel’s foundry business has struggled to meet expectations – most notably failing to deliver the level of technical service, predictability, and yield quality that external customers expect,” said Sanchit Vir Gogia, chief analyst and CEO at Greyhound Research. “Several deals have seen missed milestones and failed test runs, particularly when benchmarked against TSMC’s operational excellence.”

This brings in concerns that Intel is pulling back from manufacturing, signaling a deeper shift in its ability to control its technology stack and maintain supply chain reliability.

“In such a case, they can invest in R&D and catch up on AI processors as well as power-efficient laptop processors,” said Faisal Kawoosa, founder and lead analyst at Techarc. “Intel is facing competition from Nvidia in AI and from Qualcomm, MediaTek, and Apple Silicon on the PC side.”

However, for enterprise buyers, this changes the calculus, Gogia said, pointing out that many chose Intel precisely because of its internal alignment between design and manufacturing – a one-stop shop that offered tighter integration and faster troubleshooting.

Data from Greyhound’s research indicates that 62% of enterprise technology leaders continue to value tight integration between Intel’s chip design and manufacturing. However, that support is weakening amid growing concerns over Intel’s manufacturing performance.

“We’re still loyal to Intel, but we’ve learned to build fallback scenarios. We can’t afford surprises anymore,” one US-based CIO told Greyhound in a recent fieldnote.

A full exit from in-house manufacturing is unlikely, according to Neil Shah, partner and co-founder at Counterpoint Research. Instead, the joint venture may be a way for Intel to protect and stabilize its foundry business.

“If it is able to protect its foundry business, Intel has to manufacture its own chips in the fab to help with the fab’s scale alongside producing chips for other customers,” Shah said. “What could actually happen is Intel could diversify its design beyond x86 by adopting newer architecture which will also give Intel more experience to compete against and alongside Arm-based PC chipset vendors such as Qualcomm, Apple, and AMD in this AI PC race.”

More concerns for enterprise buyers

A JV with TSMC offers clear advantages to Intel, including higher yields, faster time-to-market, and mature processes on advanced nodes. If Intel gains access to these capabilities, customers could see more competitive chips for AI and high-performance computing.

“But this benefit comes with layers of complexity,” Gogia said. “For one, Intel’s foundry operations are still in recovery mode. The business posted a $7 billion operating loss in 2023 – the first such loss since 1986. Attempts to scale it for external customers have struggled, largely due to inconsistent service delivery and a lack of maturity in handling complex customer requirements.”

Outsourcing to TSMC could address Intel’s manufacturing issues but complicates accountability. Customers may have to navigate supply chains spanning multiple firms and countries, with unclear responsibility in the event of defects, delays, or shortages.

“Intel’s restructuring could cause short-to-midterm disruptions, including potential product delays and shifts in support models,” said Manish Rawat, semiconductor analyst at Techinsights. “Additionally, even with US-based manufacturing, Intel’s deeper integration with TSMC still ties part of its supply chain to Taiwan – raising geopolitical and continuity concerns for security-sensitive customers.”

For Intel’s foundry to succeed – and for the US to maintain supply chain security and tech sovereignty – major chipmakers will need to shift some orders from Taiwan to the US, a move likely to drive up chip costs.

“While the US government’s CHIPS Act may help offset higher costs through subsidies, achieving Taiwan-level scale remains a near- to mid-term challenge, potentially driving up IT and electronics product costs for enterprises unless absorbed,” Shah said. “However, this move would safeguard US enterprises from future supply disruptions.”

The AI market could grow to as much as $4.8 trillion by 2033, according to a new report from the United Nations Conference on Trade and Development. At the same time, the money is expected to be concentrated among a few players and could increase inequality between countries, CNBC reports.

“The benefits of AI-driven automation often favor capital over labor, which can increase inequality and reduce the competitive advantage of cheap labor in developing countries,” the report’s authors wrote. To ensure developing countries are not left behind, the UN group recommends they be involved in discussions on AI rules and ethics.

In the same report, the UN also warned that AI could affect 40% of all jobs, even as the technology has the potential to create new industries and empower workers — as long as investments are made in training.

It might be time for Apple-as-a-service as the company looks to plot a cunning course through the post-globalist freeze — and offering its products as a service could help it do just that.

Apple had been expected to introduce just such a service. In fact, Mark Gurman recently said those plans were quite advanced, but the company shelved the idea, presumably because of its potential impact on “normal” iPhone sales and consequent revenue. 

With the recent round of tariffs from US President Donald J. Trump, things have changed. Given a $3,000 iPhone has become a real possibility under the tax regime, Apple knows it’s going to see a decline in iPhone sales anyway. It knows that those iPhones it does sell will get used longer and it knows that the inevitable cost will put a lot of consumers off from buying these devices.

Apple has also bought itself some time to figure out a way forward, thanks to the planeloads of iPhones it imported into the US just before the tariff announcement was made. These should tide the company over into September, reports claim. That makes sense, because the time isn’t right for the iPhone-as-a-service plan quite yet. 

To put the plan in effect, Apple will need the support of its carrier partners who I expect were quite resistant to the idea before — they liked the margins they made on phones sold through their networks.

However, pragmatism changes things, and even they can see that some money from a lower margin is still better than no margin at all. And when analysts predict iPhone prices could hit $3,000, it’s crystal clear sales will decline.

Think about it

Those tariff taxes will impact almost everything that cannot be made, grown, or harvested in the US. They’re going to be felt, particularly by shoppers in more impoverished socio-economic groups (who also use iPhones). Most shoppers will be much too concerned about the cost of eggs to spring thousands for a phone, and while there will be an elite group of consumers for whom it’s business as usual, most people will endure a crisis of confidence. 

Like the 1920s, there may be a lot of dancing, but not much to dance about.  

The economic precipice the world appears to have been pushed over may be enough to make any iPhone-as-a-service plan look a lot more attractive. After all, it enables cash-strapped consumers to use the smartphone they desire (and perhaps also the watch, tablet, and Mac) for a predictable monthly fee, with AppleCare, iCloud+, and Apple services included, and doesn’t require they laden themselves with credit card debt. 

People are ready to accept it

It’s not as if we’re not ready for such a service. Even back in 2022, CIRP Partner and Co-Founder Josh Lowitz said: “Based on current consumer behavior, iPhone users are primed to adopt a subscription service that provides an iPhone bundled with useful apps. Almost half iPhone owners already finance their iPhone purchase, paying monthly for a new phone. And about one-third trade-in their old phone when they buy a new one. So, a significant portion of the user base is accustomed to never owning a phone, instead basically leasing it.”

Apple also gains. In this case, it benefits from potentially lower, but at least recurring, income upon which to balance its stock. And it benefits from the fact that at the end of the subscription period (or during it if the consumer cannot maintain payments), the devices will be returned for refurbishment, resale/let, and/or recycling.

This also opens up the highly lucrative second-user iPhone market, which is an income stream Apple hasn’t yet fully explored. The iPhone is the most widely sold smartphone on the second-user market and holds its value the longest; company management is said to have been eyeing whether they can extract more from those sales. 

Pros and joes

Taking things a speculative step forward, I can easily imagine the company might choose to keep the highest-end devices out of the subscription loop, making these available for sale only. Apple knows its most affluent customers may be more accepting of a higher price in exchange for a truly cutting-edge product. 

That balance of high-end retail sales and subscription-income, bolstered by all the other plans Apple is putting in place to survive the high-tax transformation of the US economy may help it build a good and viable business in times like these. A different business, but a business all the same

Will Apple do it? Will Apple choose to offer up its products on a rental basis? 

Given sales cadence is going to fall anyway and, logically, product costs will increase, the company might see the plan as a way to pass the poison pill of some of these price increases onto consumers in as easy-to-stomach a remedy as possible. It’s not ideal, of course, but right now company management will be focused on finding the least worse options to help support the future of its business. That’s the context in which such a plan makes sense.

You can follow me on social media! Join me on BlueSky,  LinkedIn, and Mastodon.

When Google first launched the NotebookLM beta in late 2023, the company billed it as an experimental note-taking app with a generative AI assistant built in. You can still use it to create notes, but the app has evolved into something much more novel and powerful. Google now calls it “the ultimate tool for understanding the information that matters most to you.”

Here’s our clearer summary of what NotebookLM has become: It’s a web-based workspace where you can upload your documents and other sources and then instruct Google’s Gemini AI chatbot to analyze and synthesize them. Gemini can summarize the source materials, answer questions about them, and create a range of document types based on them, such as study guides or executive briefings. It can even spin up an audio podcast with synthetic hosts discussing the materials.

When used in NotebookLM, Gemini pulls information from the sources you feed it rather than the internet, theoretically making it less prone to hallucinations than many other genAI tools. That said, Google clearly states that NotebookLM can make mistakes. Responses include citations from your sources, which makes it easy to check for accuracy.

NotebookLM comes in free and paid versions. The paid version, called NotebookLM Plus, comes with either a Google One AI Premium plan or a Google Workspace subscription (Business Standard or higher plan). The free and paid versions share many of the same features. The main differences are that with Plus you have fewer usage limitations, can customize the writing style and length of the content generated by Gemini AI, and have advanced options for collaborating on your workspace with others.

In this guide, we take you through how to set up and use the major features of NotebookLM.

Create your first notebook

In NotebookLM, a “notebook” contains one or more sources. You can create multiple notebooks, each representing a project you’re working on and containing sources that are specific to that project.

Caution: Google says that the data you upload is not used to train Gemini AI and will stay private, but the company also advises, “it’s best to avoid submitting any information you wouldn’t feel comfortable sharing.”

To get started, sign in to NotebookLM with your Google account, then click the Create button.

To start a notebook, click the Create button.

Howard Wen / Foundry

On the “Add sources” panel, you can drag-and-drop or upload files that are on your PC. Accepted file formats include PDF, TXT, and MP3. Toward the bottom of the “Add sources” panel, you can click to add other sources: Google Docs and Google Sheets files in your Google Drive, links to web pages or YouTube videos, and text that you’ve already copied to your PC clipboard.

You can upload a variety of file types to a notebook, or link to an outside source.

Howard Wen / Foundry

Once you’ve added a source, you’ll be taken to the workspace for your new notebook. This page is divided into three columns: Sources, Chat, and Studio.

The new notebook has three columns: one for sources, one for chatting with the AI, and one for notes and audio materials.

Howard Wen / Foundry

On the upper left, the title of the source (or the first several words in it) that you added will appear as the title for this notebook. You can change this by clicking the title and typing in a new one that you prefer.

The Sources column will list the source you added, and the Chat column will show a summary of that source. (We’ll get to the Studio column later.)

Add more sources and manage them

In the Sources column, click Add source to add another source. The sources that you add will appear in the Sources column.

When you add a source, it’s selected by default, which means Gemini will use it for analysis in the Chat and Studio columns. However, you may need to refresh your browser tab before Chat column updates to include additional sources in its summary.

Checked sources are included in Gemini’s analyses; unchecked ones are excluded.

Howard Wen / Foundry

To unselect (or select) a source: If you don’t want Gemini to include a source in a summary, click the checkmark to the right of the source to unselect it. Click the checkbox again to re-select the source.

To remove or rename a source: Move the pointer over the source and click the three-dot icon that appears to the left of it.

Get a single-source summary and kick off a related chat

If you want to delve deeper into one of your sources, click the source in the Sources column. A “Source guide” card will open at the top of the column, where Gemini will generate a summary that describes the contents of the source.

srcset="https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-05-source-guide.png?quality=50&strip=all 753w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-05-source-guide.png?resize=214%2C300&quality=50&strip=all 214w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-05-source-guide.png?resize=729%2C1024&quality=50&strip=all 729w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-05-source-guide.png?resize=497%2C697&quality=50&strip=all 497w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-05-source-guide.png?resize=120%2C168&quality=50&strip=all 120w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-05-source-guide.png?resize=60%2C84&quality=50&strip=all 60w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-05-source-guide.png?resize=342%2C480&quality=50&strip=all 342w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-05-source-guide.png?resize=256%2C360&quality=50&strip=all 256w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-05-source-guide.png?resize=178%2C250&quality=50&strip=all 178w" width="729" height="1023" sizes="(max-width: 729px) 100vw, 729px">

Click a source in the Sources column to get a summary and a list of related topics.

Howard Wen / Foundry

To the right of the summary is a list of key topics. Clicking any of these sends a request to Gemini to elaborate on the topic, and its response will appear in the Chat column.

In this tutorial, though, we’re more concerned with synthesizing data from multiple sources. To dismiss the topic-related chat and return to the summary of selected sources, click the Refresh button at the top right of the Chat column and then click Continue.

Chat with Gemini AI about your sources

The true power of NotebookLM rests in Gemini’s ability to respond to natural-language queries about your sources. You can ask it questions or give it instructions for analyzing the data.  

In the entry box at the bottom of the Chats column, type in your query and click the arrow to the right. The best way to word your requests is in the form of a question.

Or you can use one of the suggested queries that appear below the entry box. You can use the right and left arrows to cycle through them. Click a suggestion and it’ll be posted as a request to Gemini in the chat window.

srcset="https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-06-chat-with-gemini-about-sources.png?quality=50&strip=all 675w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-06-chat-with-gemini-about-sources.png?resize=199%2C300&quality=50&strip=all 199w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-06-chat-with-gemini-about-sources.png?resize=461%2C697&quality=50&strip=all 461w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-06-chat-with-gemini-about-sources.png?resize=111%2C168&quality=50&strip=all 111w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-06-chat-with-gemini-about-sources.png?resize=56%2C84&quality=50&strip=all 56w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-06-chat-with-gemini-about-sources.png?resize=318%2C480&quality=50&strip=all 318w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-06-chat-with-gemini-about-sources.png?resize=238%2C360&quality=50&strip=all 238w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-06-chat-with-gemini-about-sources.png?resize=165%2C250&quality=50&strip=all 165w" width="675" height="1020" sizes="(max-width: 675px) 100vw, 675px">

Type a request in the entry box or click one of the suggested queries below.

Howard Wen / Foundry

When you enter a request, Gemini may take several seconds to process it and post a response. It analyzes only the sources listed in the Sources column that have a checkmark by them.

When the response is posted in the Chat column, there will be numbers denoted throughout it. These are citations, each representing text from one of your sources Gemini used to create the response. When you click one of these numbers, a panel opens to show you the original text from the source so you can check whether the response is accurate.

srcset="https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-07b-gemini-response-with-citations.png?quality=50&strip=all 798w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-07b-gemini-response-with-citations.png?resize=236%2C300&quality=50&strip=all 236w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-07b-gemini-response-with-citations.png?resize=768%2C978&quality=50&strip=all 768w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-07b-gemini-response-with-citations.png?resize=547%2C697&quality=50&strip=all 547w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-07b-gemini-response-with-citations.png?resize=132%2C168&quality=50&strip=all 132w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-07b-gemini-response-with-citations.png?resize=66%2C84&quality=50&strip=all 66w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-07b-gemini-response-with-citations.png?resize=377%2C480&quality=50&strip=all 377w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-07b-gemini-response-with-citations.png?resize=283%2C360&quality=50&strip=all 283w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-07b-gemini-response-with-citations.png?resize=196%2C250&quality=50&strip=all 196w" width="798" height="1016" sizes="(max-width: 798px) 100vw, 798px">

Gemini’s generated response includes citation numbers and links.

Howard Wen / Foundry

Additionally, there are three actions that you can take, which are listed below Gemini’s response:

• Save to note: Click this button to turn the response into a note that will appear under the “Notes” section in the Studio column.
• Copy: Click this icon to copy the response to your PC clipboard.
• Thumbs Up/Down: Click either icon to rate how good you think the response was. This helps to train the AI to give you better results for future requests.

srcset="https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-08-chat-pane-actions-toolbar.png?quality=50&strip=all 799w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-08-chat-pane-actions-toolbar.png?resize=300%2C50&quality=50&strip=all 300w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-08-chat-pane-actions-toolbar.png?resize=768%2C128&quality=50&strip=all 768w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-08-chat-pane-actions-toolbar.png?resize=150%2C25&quality=50&strip=all 150w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-08-chat-pane-actions-toolbar.png?resize=640%2C107&quality=50&strip=all 640w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-08-chat-pane-actions-toolbar.png?resize=444%2C74&quality=50&strip=all 444w" width="799" height="133" sizes="(max-width: 799px) 100vw, 799px">

You can save the response as a note, copy and paste it elsewhere, and/or rate the quality of the response.

Howard Wen / Foundry

To erase the current chat and start over, click the Refresh button at the top right of the Chat column.

Customize Gemini’s responses (in NotebookLM Plus)

The Plus version lets you adjust and customize the conversational style and length of the responses generated by Gemini. Click the icon of slider controls at the top-right corner of the Chat column. On the “Configure chat” panel that opens, select from the following:

• Analyst: Gemini’s results will have a more business-oriented tone.
• Guide: Gemini’s wording will be geared to those you are collaborating or sharing your notebook with who may not be familiar with its contents.
• Custom: Opens an entry box in which you can write a prompt instructing Gemini how you want it to word its responses.
• Longer: Generates more in-depth and lengthy responses.
• Shorter: Generates concise, summarized content.

Create and manage notes

You can create notes in your notebook and optionally use them as additional sources for Gemini to analyze.

To create a new note: In the Studio column under the Notes heading, click Add note. A card for a new note will open at the top of the Studio column.

Click New Note and type in a new title for this note. Then type your text into the body of the note. You can also paste in text and/or images saved on your PC clipboard, format the text (as headings, bold, italic, or bulleted or numbered lists), and insert web links.

Creating a new note.

Howard Wen / Foundry

To save a note: When you’re finished, click the double-arrow icon at the upper right of the note. Your new note will appear in the Notes section of the Studio column. Click it to reopen it so you can read or edit it.

To save a note as a source: With the note open, click the Convert to source button at the bottom of the Studio column. The note will be added to the sources list in the Sources column.

To delete a note: Move the pointer over its title in the Notes list. Click the three-dot icon that appears to the left of it and click Delete note.

To delete all notes at once, or convert all your notes to a source: To the right of the Notes section heading, click the three-dot icon and select Convert all notes to source or Delete all notes.

Selecting Convert all your notes to source combines all your notes together into one source. (Otherwise, you can convert an individual note to a source at any time by opening the note and selecting Convert to source.)

srcset="https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-11-convert-or-delete-all-notes.png?quality=50&strip=all 635w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-11-convert-or-delete-all-notes.png?resize=294%2C300&quality=50&strip=all 294w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-11-convert-or-delete-all-notes.png?resize=165%2C168&quality=50&strip=all 165w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-11-convert-or-delete-all-notes.png?resize=82%2C84&quality=50&strip=all 82w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-11-convert-or-delete-all-notes.png?resize=471%2C480&quality=50&strip=all 471w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-11-convert-or-delete-all-notes.png?resize=353%2C360&quality=50&strip=all 353w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-11-convert-or-delete-all-notes.png?resize=245%2C250&quality=50&strip=all 245w" width="635" height="647" sizes="(max-width: 635px) 100vw, 635px">

You can combine all your notes into a single source, or delete them all.

Howard Wen / Foundry

Create targeted documents from your notes

In addition to responding to queries about your sources, NotebookLM can extract content from the sources you select to create specific types of documents such as study guides, briefing documents, FAQs, or timelines.

To create a document: In the Studio column, under the Notes section, click one of the following:

• Study guide: Your selected sources are turned into a guide that’s structured for the reader to learn about them, featuring quiz questions, essay format questions, and a glossary of key terms.
• Briefing doc: This is a summary of key information that Gemini extracts from the selected sources in your notebook. Sections typically include an executive summary, main themes and ideas, and a conclusion.
• FAQ: Information from the selected sources is formatted into a list of Frequently Asked Questions and answers.
• Timeline: Data from the selected sources will be formatted into a chronological representation of events. This format obviously works best with sources that include specific events and dates.

Once you’ve selected one of the above, Gemini will analyze the selected sources in your notebook and generate a draft version of the document type. It’ll appear as a note listed under the Notes section of the Studio column. Depending on how extensive the content is in your selected sources, the draft may look incomplete, so you’ll need to make edits to it.

A briefing document and a timeline generated from selected sources.

Howard Wen / Foundry

Share and collaborate on your notebook

It’s not only you who can benefit from a notebook you create. You can also share it with others, and they’ll be able to query Gemini about the source material too.

When you share a notebook, you can either restrict others’ permissions so that they can only view and query the notebook’s contents, or allow them to edit it, such as by adding, changing, or removing sources or notes in it. Everyone you share a notebook with can interact with Gemini and copy its responses to their own PC clipboard. But only those with Editor status can add a Gemini response as a note or as a new source to your notebook.

To share a notebook: Click the Share button that’s toward the upper right of your notebook’s workspace. A Share panel opens over the workspace.

Inside the entry box, type the name or email address of someone who’s in your Google contacts. (Note: they must already be in your Google contacts. You cannot type in any email address here.) They will be added to the “People who have access” list.

Sharing a notebook.

Howard Wen / Foundry

By default, a person you add will be granted Viewer access to your notebook. If you want them to collaborate with you on the notebook, click Viewer and change it to Editor.

When you are done adding people and setting their access levels, click the Send button. They’ll be notified by email that you’re sharing this notebook with them (unless you unchecked “Notify people”).

In NotebookLM Plus: Under “Viewer access” on the Share panel, you can allow those you’ve granted only Viewer access to see the sources and notes in your notebook. Or you can restrict them from seeing these — they will only be allowed to chat with Gemini about your notebook.

Get AI-spoken discussions of your notes

The splashiest feature that Google has added to NotebookLM since its launch is Audio Overview, which generates a podcast-like spoken discussion between two AI “hosts” (currently supported in English only). They banter back and forth in a friendly tone, sounding eerily like real people, as they summarize key points in your sources.

This feature may come off as an impressive tech gimmick. But it can be a convenient, and entertaining, way for you (and others you share your notebook with) to listen to your notebook content as if it’s a podcast during your daily commute.

Create an Audio Overview

In the Studio column, under the Audio Overview section, click the Generate button. Or, if you click the Customize button: You can type an instruction to guide what you want the AI hosts to focus their discussion on about your sources.

Gemini will analyze your selected sources and generate an audio file. This usually takes several minutes. The exact amount of time depends on the word length and complexity of your sources.

Note: Ensure that the text in your selected sources reads clearly. This will help generate a discussion between the AI hosts that sounds coherent. Source documents that have headings and subheadings can also help to create a logical flow in their spoken discussion.

Once the audio file is generated, it will appear in the Audio Overview section. Click it to start listening.

srcset="https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-14-audio-overview-playback.png?quality=50&strip=all 628w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-14-audio-overview-playback.png?resize=300%2C214&quality=50&strip=all 300w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-14-audio-overview-playback.png?resize=236%2C168&quality=50&strip=all 236w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-14-audio-overview-playback.png?resize=118%2C84&quality=50&strip=all 118w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-14-audio-overview-playback.png?resize=506%2C360&quality=50&strip=all 506w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-14-audio-overview-playback.png?resize=351%2C250&quality=50&strip=all 351w" width="628" height="447" sizes="(max-width: 628px) 100vw, 628px">

Listening to an audio overview in NotebookLM.

Howard Wen / Foundry

Along the top-right of the audio player are icons that let you rate the overall production quality of the audio summary (thumbs up/down) and share it publicly for others to listen to online. Click the three-dot icon to change the playback speed or to download the audio summary as a WAV file.

Interact with an Audio Overview (experimental feature)

Not only can you listen to the two AI hosts talk about your notebook sources, you can join in on their discussion — literally. Click Interactive mode, which will open a panel in the Studio column showing a sine wave responding to their voices. At the bottom of the column, click the Play audio icon (a right-pointing triangle), and when you feel like taking part in their talk, click the Join button.

Speaking into your PC’s microphone, you can interrupt the hosts and ask them a question or make a request to clarify a point or explain something in more detail. It’s as if you’ve called in to a live radio show.

srcset="https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-15b-interact-with-audio-summary.png?quality=50&strip=all 753w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-15b-interact-with-audio-summary.png?resize=227%2C300&quality=50&strip=all 227w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-15b-interact-with-audio-summary.png?resize=527%2C697&quality=50&strip=all 527w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-15b-interact-with-audio-summary.png?resize=127%2C168&quality=50&strip=all 127w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-15b-interact-with-audio-summary.png?resize=64%2C84&quality=50&strip=all 64w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-15b-interact-with-audio-summary.png?resize=363%2C480&quality=50&strip=all 363w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-15b-interact-with-audio-summary.png?resize=272%2C360&quality=50&strip=all 272w, https://b2b-contenthub.com/wp-content/uploads/2025/03/google-notebooklm-15b-interact-with-audio-summary.png?resize=189%2C250&quality=50&strip=all 189w" width="753" height="996" sizes="(max-width: 753px) 100vw, 753px">

To join in the AI hosts’ discussion, click Join.

Howard Wen / Foundry

Create and manage more notebooks

Once you discover how useful a NotebookLM notebook is for one project, you’ll want to create more.

On the notebook workspace, click the NotebookLM icon at the upper-left corner to go to the NotebookLM home page. This shows all the notebooks under your account.

Find and manage all your notebooks on the NotebookLM home page.

Howard Wen / Foundry

To create a new notebook: Simply click the Create new button, which will open a new workspace for a blank notebook.

To manage your notebooks: On the NotebookLM home page, you can change the title of a notebook or delete it by clicking the three-dot icon on the upper-right of its card. (If your notebooks are shown in list view, this icon will be to the right of the notebook’s title.)

This article was originally published in February 2024 and updated in April 2025.

Related:

• How to use Gemini AI to write (and rewrite) in Google Docs and Gmail
• How to use Gemini AI to make templates in Google Sheets
• Google Workspace: 7 great ways to use the Gemini AI sidebar
• Google Workspace tips and tutorials

Mesmerized by the scalability, efficiency and flexibility claims from generative AI (genAI) vendors, enterprise execs have been all but tripping over themselves trying to push the technology to its limits. 

The fear of flawed deliverables — based on a combination of hallucinations, imperfect training data and a model that can disregard query specifics and can ignore guardrails — is usually minimized. 

But the Mayo Clinic is trying to push back on all those problematic answers. 

In an interview with VentureBeat, Matthew Callstrom, Mayo’s medical director, explained: “Mayo paired what’s known as the clustering using representatives (CURE) algorithm with LLMs and vector databases to double-check data retrieval.

“The algorithm has the ability to detect outliers or data points that don’t match the others. Combining CURE with a reverse RAG approach, Mayo’s [large language model] split the summaries it generated into individual facts, then matched those back to source documents. A second LLM then scored how well the facts aligned with those sources, specifically if there was a causal relationship between the two.”

(Computerworld reached directly to Callstrom for an interview, but he was not available.)

There are, broadly speaking, two categories for reducing genAI’s lack of reliability: humans in the loop (usually, an awful lot of humans in the loop) or some version of AI watching AI. 

The idea of having more humans monitoring what these tools deliver is typically seen as the safer approach, but it undercuts the key value of genAI — massive efficiencies. Those efficiencies, the argument goes, should allow workers to be redeployed to more strategic work or, as the argument becomes a whisper, to sharply reduce that workforce.

But  at the scale of a typical enterprise, genAI efficiencies could replace the work of thousands of people. Adding human oversight might only require dozens of humans. It still makes mathematical sense.

The AI-watching-AI approach is scarier, although a lot of enterprises are giving it a go. Some are looking to push any liability down the road by partnering with others to do their genAI calculations for them. Still others are looking to pay third-parties to come in and try and improve their genAI accuracy. The phrase “throwing good money after bad” immediately comes to mind.  

The lack of effective ways to improve genAI reliability internally is a key factor in why so many proof-of-concept trials got approved quickly, but never moved into production. 

Some version of throwing more humans into the mix to keep an eye on genAI outputs seems to be winning the argument, for now. “You have to have a human babysitter on it. AI watching AI is guaranteed to fail,” said Missy Cummings, a George Mason University professor and director of Mason’s Autonomy and Robotics Center (MARC). 

“People are going to do it because they want to believe in the (technology’s) promises. People can be taken in by the self-confidence of a genAI system,” she said, comparing it to the experience of driving autonomous vehicles (AVs).

When driving an AV, “the AI is pretty good and it can work. But if you quit paying attention for a quick second,” disaster can strike, Cummings said. “The bigger problem is that people develop an unhealthy complacency.”

Rowan Curran, a Forrester senior analyst, said Mayo’s approach might have some merit. “Look at the input and look at the output and see how close it adheres,” Curran said. 

Curran argued that identifying the objective truth of a response is important, but it’s also important to simply see whether the model is even attempting to directly answer the query posed, including all of the query’s components. If the system concludes that the “answer” is non-responsive, it can be ignored on that basis.

Another genAI expert is Rex Booth, CISO for identity vendor Sailpoint. Booth said that simply forcing LLMs to explain more about their own limitations would be a major help in making outputs more reliable.

For example, many — if not most — hallucinations happen when the model can’t find an answer in its massive database. If the system were set up to simply say, “I don’t know,” or even the more face-saving, “The data I was trained on doesn’t cover that,” confidence in outputs would likely rise.

Booth focused on how current data is. If a question asks about something that happened in April 2025 — and the model knows its training data was last updated in December 2024 — it should simply say that rather than making something up.  “It won’t even flag that its data is so limited,” he said. 

He also said that the concept of “agents checking agents” can work well — provided each agent is assigned a discrete task. 

But IT decision-makers should never assume those tasks and that separation will be respected. “You can’t rely on the effective establishment of rules,” Booth said. “Whether human or AI agents, everything steps outside the rules. You have to be able to detect that once it happens.”

Another popular concept for making genAI more reliable is to force senior management — and especially the board of directors — to agree on a risk tolerance level, put it in writing and publish it. This would ideally push senior managers and execs to ask the tough questions about what can go wrong with these tools and how much damage they could cause. 

Reece Hayden,  principal analyst with ABI Research, is skeptical about how much senior management truly understands genAI risks.

“They see the benefits and they understand the 10% inaccuracy, but they see it as though they are human-like errors: small mistakes, recoverable mistakes,” Hayden said. But when algorithms go off track, they can make errors light years more serious than humans.

For example, humans often spot-check their work. But “spot-checking genAI doesn’t work,” Hayden said. “In no way does the accuracy of one answer indicate the accuracy of other answers.”

It’s possible the reliability issues won’t be fixed until enterprise environments adapt to become more technologically hospitable to genAI systems.

“The deeper problem lies in how most enterprises treat the model like a magic box, expecting it to behave perfectly in a messy, incomplete and outdated system,” said Soumendra Mohanty, chief strategy officer at AI vendor Tredence. “GenAI models hallucinate not just because they’re flawed, but because they’re being used in environments that were never built for machine decision-making. To move past this, CIOs need to stop managing the model and start managing the system around the model. This means rethinking how data flows, how AI is embedded in business processes, and how decisions are made, checked and improved.”

Mohanty offered an example: “A contract summarizer should not just generate a summary, but it should validate which clauses to flag, highlight missing sections and pull definitions from approved sources. This is decision engineering defining the path, limits, and rules for AI output, not just the prompt.”

There is a psychological reason execs tend to resist facing this issue. Licensing genAI models is stunningly expensive. And after making a massive investment in the technology, there’s natural resistance to pouring even more money into it to make outputs reliable.

And yet, the whole genAI game has to be focused on delivering the goods. That means not only looking at what works, but dealing with what doesn’t. There’s going to be a substantial cost to fixing things when these erroneus answers or flawed actions are discovered.

It’s galling, yes; it is also necessary. The same people who will be praised effusively about the benefits of genAI will be the ones blamed for errors that materialize later. It’s your career — choose wisely.

At a special 50th anniversary event on Friday, Microsoft executives reflected on the company’s storied past and on how it’s now reinventing itself for an AI-focused future.

With previous CEOs Bill Gates and Steve Ballmer in attendance, current CEO Satya Nadella boasted that the company is “leading this new wave of AI innovation, and more importantly, democratizing just like we did it with the PC.”

Details about the company’s plans were laid out by Microsoft Executive Vice President Mustafa Suleyman, who noted that the ability to customize Windows to every person’s specific needs is coming. “Years ago, Bill laid out a bold ambition, which at the time probably felt like a pretty crazy dream — to put a PC on every desk and in every home.

“Today, we’re creating a Copilot for everyone,” Suleyman said at the event, which was webcast. 

Suleyman talked about how the company is transforming its generative AI (genAI)-based Copilot into a personal assistant. Microsoft is replicating key sensory features from humans into the software.

“Today, we’re taking the very first steps towards rich memory and personalization, the very foundations of an AI companion,” Suleyman said.

Copilot is gradually adding a “Memory” feature that can personalize the tool to remember human preferences, dates, events and more. Suleyman pointed out how the AI agent over time will be able to remember birthdays, and provide reminders on tasks. It will also provide advice on how users go through each step in training sessions on specific topics and even memorize individual traits, such as whether a person greets others formally or informally.

The memory feature works with others such as “Actions,” which can complete tasks in the background.

Microsoft is also developing avatars for Copilot that make interacting with it more fun. Suleyman showed off avatars as animated characters, and in jest showcased the dreaded Clippy — of old Microsoft Office fame — as an avatar.

The company’s main announcements included Copilot Vision, a mobile app that can help users interact with the real-world. The app uses the phone’s camera to capture images and in real-time provide context of the surroundings.

“With our new mobile app, Copilot can actually see what you see and talk to you about it in real time,” Suleyman said.

The second piece to Copilot Vision is a Windows app, which can take a snapshot of a user’s PC screen and help explain what is being displayed. The app works across applications, browser tabs or files.

“It will read the screen and interact with the content. You’ll be able to use Copilot to search, change settings, organize files and collaborate on projects without switching between files or apps,” the company said in a blog post.

“With my permission it can see my screen like a second set of eyes,” Suleyman said at the event said. “It’s my sounding board. And most importantly, it can respond in the context of what I’m seeing on my screen.”

Suleyman made no reference to Windows Recall, the controversial Copilot feature that uses snapshots to log the history of activity on a PC. Recall was unveiled last year and quickly ran into a storm of controversy related to privacy concerns.

Microsoft has also started rolling out Copilot Search, with AI integrated into a conventional Bing search to provide better search results. The search results will be personalized and dynamically generated on the screen.

“With Copilot’s new search capabilities, you can get many magazine style cards made just for you, on any topic that you care about, with text, images, videos, and maps built right in,” Suleyman said.

Microsoft also unveiled “Podcasts,” an AI feature that can instantly generate podcasts with video and audio, and new AI technologies for Azure AI Foundry. 

For enterprise users, Microsoft recently rolled out Research and Analyst agents to boost enterprise search and employee productivity. 

AI will be the biggest change to the PC since the graphical user interface (GUI), and it may be as important as the first databases for enterprise users, said Jack Gold, principal analyst at J. Gold Associates.

But harnessing its potential is a challenge, with numerous usability, privacy and security challenges. “The ability to make AI most useful and efficient for enterprise needs still needs a lot of work. We’re in the first innings,” Gold said.

Microsoft’s challenge with AI is not just in the OS, but also in apps that support enterprise users, where Microsoft has a large installed base.

“While Copilot may make the way we interact with our devices through agents that implement and execute tasks for us more personal, it’s what AI may do to enhance our insights from our increasingly complex enterprise informational environment that could be a game changer,” Gold said.

It’s likely to be a decade-long maturing process before enterprises see the same level of maturity and creativity users have grown to expect in day-to-day go-to apps.

“Enterprises need to start down the path now, but don’t expect to achieve the end state in the short term,” Gold said.

The tech market lost 29,000 jobs in March, even as the unemployment rate for tech-related jobs slipped 0.1% to 3.1%, according to an analysis of US Labor Department data released today.

At the same time, tech sector employment fell by 8,428 jobs last month, driven largely by job cuts in tech services and telecommunications, according to CompTIA, a non-profit IT trade association that analyzed figures from the US Bureau of Labor Statistics (BLS)

The BLS data, CompTIA said, sent conflicting signals about the IT job market as uncertainty around it “accelerated.”

“With many employers in wait-and-see mode, the jobs data is about in line with expectations for the month,” said Tim Herbert, chief research officer at CompTIA.

Overall, US job growth beat expectations, with the broader economy adding 228,000 positions. That’s nearly 100,000 more than the 135,000 economists had predicted would be added. Conversely, the federal government lost 4,000 jobs, due to reduction in force efforts by the unofficial Department of Government Efficiency (DOGE).

Victor Janulaitis, CEO of IT industry research firm Janco Associates, said DOGE’s impact is causing IT contracts to be delayed, leaving government IT professionals uncertain about their future.

The overall US unemployment rate rose slightly to 4.2% in March, just 0.1% higher than in February. Previous US job gains for January and February were revised down by a total of 48,000, with January reduced by 14,000 and February by 34,000, according to BLS data.

In the first quarter of 2025, the IT job market shrank by 34,200 jobs, according to Janulaitis. The number of unemployed IT professionals dropped slightly in March to 133,000 from 148,000 at the start of the year. “We believe that many low-skilled, legacy-skilled IT pros, or displaced IT professionals have stopped looking for jobs in the IT sector,” he said.

CompTIA said new employee job postings for tech occupations rose slightly to more than 213,000. In all, there were nearly 478,000 active tech job postings last month. Industry sectors adding the most new tech job postings included professional, scientific and technical services (52,526), administrative and support services (26,099) and manufacturing (21,975).

Software developers and engineers, tech support specialists, systems engineers and architects, and cybersecurity engineers and analysts were in highest demand, according to CompTIA’. Positions in artificial intelligence (AI) or that require AI skills accounted for 21% of all active tech job postings. And half of all tech job postings did not require a four-year academic degree.

Several industry experts saw the uptick in overall hiring and steady unemployment rates as a good omen.

Martha Heller, CEO of executive search firm Heller, called the BLS jobs report healthy, and said it reinforces “the plain truth that technology innovation (primarily AI) will drive up employment. In every sector, companies need people to realize the return on their AI investments, and while those people are often technologists, the AI employment boom will expand to include a broader range of skills.

“With the nearly certain instability that our current economic policy brings, technology and the people who can use it to navigate uncertainty will have job security for a long time,” Heller said.

Kye Mitchell, head of US tech recruiting firm Experis North America, sees AI as a boon for all businesses. “In the tech sector, our real-time data shows that businesses are addressing their AI needs by focusing on their data to be prepared to best leverage AI in this new era,” she said.

Quarter over quarter, demand for data analysis and cleansing has risen sharply, he said. For example, roles for data scientists are up 238%. Database architect positions rose 142%. And, job openings for mathematicians were up 1,272%, Mitchell said.

“Additionally, there’s a growing need for executive management, indicating businesses are relying on leadership to drive AI integration,” she said. “The impact of recent economic policy shifts on AI worker demand is still unfolding, and market uncertainty may lead to longer decision timelines and adjustments.”

Others familiar with the tech industry saw the latest jobs report as positive, including Ger Doyle, US country manager at global staffing firm ManpowerGroup. “Today’s jobs report is a welcome sign, given the negative indicators we saw leading into it. While the US labor market is proving to be resilient, there are signs of cooling that are consistent with employers navigating uncertainty.”

ManpowerGroup data shows a 2% overall decline in all job postings and a 3% drop in new job postings. While the labor market appears stable, Doyle said employers remain cautious, with the exception of healthcare and education sectors.

“Uncertainty is keeping both employers and employees from making moves, leading businesses to focus on maintaining current workforces but considering adjustments if the uncertainty persists,” he said.

Janulaitis said interviews he performed with more than 100 CFOs and CIOs revealed optimism about Trump administration, though many anticipate a mild economic downturn in mid-2025. “Unemployed IT pros found jobs faster than expected as CIOs quickly filled positions while hiring was still authorized,” he said.

`

We are still within the sequence of energies described in Newton’s Third Law of Motion, “every action has an equal and opposite reaction.” In this case, Apple, already badly bruised by Trump’s tariffs and their impact on its business, will now take another kicking as nations react to those oddly calculated trade taxes. Europe is gunning for Apple’s services, while China is about to stick some of the world’s rarest components behind its own trade wall. 

The third law

China, predictably, isn’t happy. Not only has it slapped a reciprocal 34% tariff against US goods, but it also introduced export restrictions on rare earth materials — this will have a direct and significant impact on tech, as these are used in almost every electronic device.

That includes Macs and PCs, smartphones and tablets, as well as weapons systems, energy-generation tech, and electric cars. China produces around 90% of the world’s refined rare earths and the new export restrictions are being applied against exports to all countries. The exports haven’t been banned, but the latest move does give Beijing the strategic power to restrict them or even turn them off. (Not all is lost, of course. Other nations also have some of these materials, including Canada, for example).

France, meanwhile, is warning that the EU’s second line response to these tariffs will be aimed at digital services, such as those provided by all the Big Tech firms — Amazon, Apple, Google, Meta, and Microsoft.

This is also a strategic approach, given that services are popular across the region and have already emerged as the bulwark against the recession those firms all knew was coming but never called out, preferring socialism for the rich. (And given Apple’s really under pressure in the EU, at what point will the US decide to relent in some fields to give its big tax revenue generator a break?)

Also, in the run up to this chapter, don’t be too surprised if Apple — and other companies selling products into the US — imported bigly piles of hardware to meet short-term demand. 

Tactical silence

Don’t expect to hear much from Apple’s leadership quite yet. Execs recognize that the sensible approach is to stay under water until others throw their own responding stones into the pool as the rocks of Trump’s tariff troubles ripple across an angry world trade pond.

They may have gamed out a whole range of potential scenarios, but must now wait to see what’s left after the storm. It’s only after both action and reaction have had time to play out that defensive plans can be put into effect.

I don’t imagine they feel particularly optimistic in the medium term. Investment bank JP Morgan now sees a 60% chance of global economic recession this year, up from a still uninspiring 40% before. A recession would further deepen the damage to consumer confidence, and no doubt cut tithes from product sales across every category, not just iPhones. But America’s biggest company will be especially impacted, and given it already supports 2.9 million US jobs, what’s good for Apple is good for America. 

As is what’s bad.

Battle of the bean counters

While the extent of the tariffs and the unpredictable nature of reactions to them makes the future hard to see, Apple’s leadership is shrewd. They’ll have read The Art of War enough times to understand the need to preserve what resources they do possess and use what defensive opportunities they can exploit.

Look at it this way: every company involved in hardware production rationally knows that the cost of repatriating production to the US would be trillions of dollars, and would take years, even if there were enough skilled workers to handle all that.

With trillions at stake, what do they do? Backstage, anticipate political maneuvering, lobbying, and a chance for a cohort of people not usually celebrated too much at Apple to shine. I’m talking about lawyers, accountants, tax experts, PR types, and corporate staff.

This really is their time.

Way I see it, just as product design saved Apple in the late ’90s with a message that echoed the zeitgeist then, tactical use of obscure tax and territorial law has now become equally important to the future of Apple and any other multinational company exposed to risk. In terms of profit and revenue, this is the time when the back room professionals must save the day. Defend. Delay. Navigate. Obfuscate. Be ready to innovate. Your shareholders are counting on you.

You can follow me on social media! Join me on BlueSky,  LinkedIn, and Mastodon.

On April 4, 1975, two childhood friends, Paul Allen and Bill Gates, started what is today one of the world’s leading technology companies. Inspired by the January cover of Popular Electronics magazine, they created Microsoft (then Micro-Soft, referring to microprocessors and software) to develop software for the Altair 8800, one of the first personal computers on the market.

Half a century later, and after establishing its operations center in Redmond, Washington, in 1986, the company has diversified its activity, from Windows operating systems to the cloud and the video game industry, reaching a present dominated by artificial intelligence (AI).

In the stock market — where it listed in March 1986 — it has always occupied the top positions on the Nasdaq, where it is known as one of “the magnificent seven” along with other major players in the digital landscape.

Throughout these five decades it has only had three global CEOs: Bill Gates himself, Steve Ballmer and Satya Nadella, in an industry accustomed to the constant dance of names. But in that time it has acquired and merged with hundreds of companies, including Hotmail, Nokia’s mobile devices division, and more recently Skype, GitHub, Linkedin and Activision Blizzard, the latter, related to gaming, being the most expensive at $68.7 billion.

But let’s look back to some of the earliest milestones in the development of the organization. In 1980, Gates and Allen signed an agreement to provide an operating system for IBM’s first personal computer. After presenting its MS-DOS PC in 1981, Windows came along in 1985 and, by the end of the decade, it was already the most recognized software brand on the planet. The first version of Microsoft Office arrived around that time, the productivity suite that today, under the Microsoft 365 brand, is used by more than 1.5 billion people.

One of the first images of the Microsoft team, TAKEN shortly after its founding. In the bottom row are Bill Gates (left) and Paul Allen (right).

Microsoft

During the 1990s, Bill Gates’ vision of the potential of the World Wide Web led to the integration of the company’s products and services into the network and the creation of one of the most famous browsers in the digital world, Internet Explorer. Gates has become one of the leading gurus of the computer industry and his thinking has permeated the world through numerous lectures, public appearances and published books.

In 1995 he set out a roadmap for the IT revolution in his first major text, The Road Ahead. Of the changes to come, he told Time magazine in a 1997 interview, “People always fear change. People feared electricity when it was invented, didn’t they?” That year he created the Bill and Melinda Gates Foundation to address health and education issues of our time.

The arrival of the new century crowned a new CEO, Steve Ballmer. Gates stepped back to let the company’s first business manager continue with his strategy. Through 2013 Ballmer laid the first stones of the company’s video game business with the launch of the Xbox game console, and of the cloud with the 2008 launch of Azure. It was in 2011 that Office made the leap to the cloud with Office 365, and the company too a further step in collaboration with the purchase of the Skype video calling tool. In Ballmer’s final years in charge the Surface line of tablets emerged.

Bill Gates and Steve Ballmer during a meeting in the 1990s.

Microsoft

In 2014 comes the Satya Nadella era, characterized by a hitherto unknown acceleration of the technology industry around the planet. It stands out for the consolidation of cloud business and video games while introducing new trends, such as AI. Along the way, Windows 10, the purchase of LinkedIn, GitHub, the launch of Teams, which became the fastest growing application for the company, and the strong investment in OpenAI, the creator of ChatGPT. His great achievement, perhaps, is to take Microsoft to a market value of $3 trillion dollars by 2024 thanks to this strategy. But it was in 2015 that he turns the technology around with a new corporate mission: “To empower every person and every organization on the planet to achieve more” through IT.

Today, Nadella leads a company of 228,000 employees spread across more than 190 countries.

Supplied Art (with Permission)

Paco Salcedo, head of the company’s Spanish subsidiary, says, “The most relevant learning in these 50 years is to have gone from believing we were a company ‘that knew everything’ to wanting to become a brand that ‘wants to learn everything’. And, if we look to the future, we believe that AI can enable the next billion jobs in all parts of the economy.”

A marathon journey

Reviewing this trajectory, Foundry consultant Fernando Maldonado says the company has adapted to change. “You have to know how to rise to disruptions in time,” he said, noting the company’s entry into the cybersecurity sector under the premise that “without protection there is nothing” when hardly anyone understood this need. Also, the company’s switch to embrace open source, even before its acquisition of GitHub. “At the beginning, open source was the enemy and, today, we can say that it also knew how to adapt. […] A lot of its software is already open source.”

“All this gives you an idea that it has always been able to react in time to macro-trends,” Maldonado said. Although, he added, the company has taken several hits, inevitable for a company that has been around for half a century, such as the flop of the Bing search engine. “If it hadn’t had any friction, we wouldn’t be talking about its achievements,” he says. “But it has worked on a very solid foundation with Office, which is the anchoring of all types of companies.”

Now, it remains to be seen how it fares in the still very young world of AI, with the billion-dollar investments it is making, although there is some slowdown in terms of data centers. In any case, says the analyst, “it is also making a very strong bet on the future”.

Nadella’s reflection regarding Microsoft’s anniversary is in line with the expert’s analysis: “I’ve found myself reflecting on how Microsoft has remained a consequential company decade after decade in an industry with no franchise value. And I realize that it’s because -time and time again, when tech paradigms have shifted- we have seized the opportunity to reinvent ourselves to stay relevant to our customers, our partners and our employees. And that’s what we are doing again today,” he said.

Two of the biggest names in artificial intelligence are independently developing new AI tools that encourage learning, at a time when the technology has been criticized for dumbing down smart users in the enterprise and discouraging critical thinking. While the new initiatives from OpenAI and Anthropic are aimed at transforming how AI is used in higher education, the opportunities they open up extend beyond universities.

The approach taken by the two AI firms, emphasizing learning, retention, and critical thinking, has implications for enterprises that are increasingly integrating AI into their workflows.

Such education-focused initiatives arrive at a crucial moment, as research increasingly shows that improper use of AI in workplaces leads to what experts call “cognitive atrophy” — a decline in employees’ ability to think critically when AI handles too many cognitive tasks.

Like it or not, those who use Windows Update to obtain and install updates, patches, and fixes will occasionally encounter issues. Sometimes, a specific update may not download. At other times, an update may fail to install. Every now and then, Windows Update (a.k.a. WU) may refuse to interact with the Windows servers from which updates come.

That’s where these steps come into play. Try these troubleshooting techniques in the order outlined below to get WU operational again.

Although I’ve illustrated the WU troubleshooting steps in Windows 11 here, the same techniques also work with Windows 10, with only minor variations. This story covers both Windows OSes quite nicely.

1. If at first you don’t succeed…try again

Though the most basic of all troubleshooting techniques for Windows Update doesn’t always work, it can help sometimes. That is, if WU fails to work as expected, the first thing to try is running the Windows Update request again.

For best results, restart the target PC before trying again. It’s not absolutely necessary, but it may clear some pending or transitory conditions (such as an already-downloaded update that needs to be applied, or some other pending system action) that will put Windows Update back to rights.

Here’s what to do (see Figures 1 and 2):

1. Click Start, then the power button icon (far right), then Restart from its pop-up menu.

Figure 1: After saving all work, click the power button in the Start menu (lower right), and then click Restart in the menu.

Ed Tittel / Foundry

2. When the desktop reappears, click Start > Settings > Windows Update, then click the Check for updates button.

Figure 2: Click Start > Settings > Windows Update. Then click Check for updates. It may work!

Ed Tittel / Foundry

In my own experience, this works one-quarter to one-half of the time when a WU problem manifests. The upload and install simply succeeds on a second try. Could it have been a download hiccup, or perhaps cosmic rays? Whatever: getting it on that next try is a good thing!

But if restarting the PC provides no relief, move onto the next section.

2. Try the Windows Update troubleshooter (or Get Help app)

As the trite software saying goes, “There’s an app for that.” For common Windows issues — which definitely include WU — it’s equally valid to observe, “There’s a troubleshooter for that.” Troubleshooters are built-in Windows mini-apps meant to diagnose and fix specific Windows problems.

In Windows 11, troubleshooters are being deprecated in favor of a new universal Get Help app, but for now, at least, the Windows Update troubleshooter remains the go-to for WU problems in both Windows 10 and Windows 11. Indeed, if you enter troubleshoot Windows Update errors into the Get Help app, it takes you to this very same troubleshooter.

At some as-yet-unknown point in the future, that troubleshooter may be replaced with another tool available via Get Help. In the meantime, here’s how to fire up the WU troubleshooter directly:

1. Click Start > Settings.

2. Insider the Settings app, type trouble into search. You’ll see something like the list shown in Figure 3 appear. Click Troubleshoot Other problems.

Figure 3: The WU troubleshooter falls under Troubleshoot Other problems.

Ed Tittel / Foundry

3. In the “Other troubleshooters” pane, Windows Update appears in position 4 (from top) under the “Most frequent” heading, as shown in Figure 4. Click the Run button at the right-hand side of that entry.

Figure 4: Click the Run button to the right of Windows Update to launch that troubleshooter.

Ed Tittel / Foundry

4. The Windows Update troubleshooter runs in its own window. You’ll pick one of two options: I can’t download, install, or uninstall an update or I installed an update, but that update has caused a problem. Pick the one that matches your situation. (For this story, I picked I can’t download…) The troubleshooter asks permission to run an automated diagnostic. Choose Yes, and it runs — be patient, this can take a minute or two.

When it finishes, you’ll see a final report window like the one in Figure 5.

Figure 5: When the error check completes, the troubleshooter asks if its fix worked. How can you tell?

Ed Tittel / Foundry

The troubleshooter asks if its fix solved the problem, but it’s difficult to tell that without running another update check in WU. I usually answer Yes to the question shown in Figure 5, restart the PC, and run Windows Update again. Only then is the real Yes/No answer clear.

In my experience, the WU troubleshooter actually fixes what’s wrong one-third to one-half of the time. If it doesn’t, advance to the next section, please.

3. Completely reset WU

If the WU troubleshooter doesn’t address your issue, there is heavier artillery you can bring to bear. It’s fully described in the excellent Windows 11 Forum tutorial entitled “Reset Windows Update in Windows 11.” This tutorial includes a download link to a special batch file that turns off all WU related services, resets all Windows Update components and policies, then restarts all the services it turns off. (There’s a similar tutorial for resetting WU in Windows 10 on Windows 10 Forums, complete with its own batch file.)

The download ends in a ZIP extension, so you’ll need to unzip it after you download. This will reveal the batch file named Reset_Reregister_Windows_Update_Components_for_Windows11.bat.

Next, you’ll need to open an administrative command prompt or PowerShell session. (Type either cmd or powershell in the Windows search box, right-click Command Prompt or Windows PowerShell in the list of results, select Run as administrator, and enter the password if prompted.)

Next, navigate to the folder where the aforenamed batch file resides. (It’s so handy, I keep it on the desktop on my Windows 10 and 11 PCs, as shown in Figure 6.)

Figure 6: The filename is so long you need to open Properties to see it on the desktop.

Ed Tittel / Foundry

Cut and paste the filename into the command line, and then hit the Enter key to run it. That’s all there is to it.

This batch file takes a while to run and involves checking and stopping WU services (bits, wuauserv, and cryptsvc). It then flushes the DNS cache, clears pending update requests, deletes temporary files and folders related to WU downloads, and resets Windows Update policies in the Windows Registry. Then it restarts the services it stopped, and finally ends by offering a Restart button. Use it, because the PC must restart for all these changes to take effect.

I’ve used this batch file on most of my Windows 10 and 11 PCs (and 7, 8, and 8.1 PCs before that), and it’s worked in every case except two to put WU back in action. On the machines where it failed, it turned out I had a hardware problem (a failing SSD) on one of them. On the other, Microsoft’s update was later documented to pose problems for certain PCs.

IMO, the reset-reregister batch file is something of a silver bullet for WU problems. Thus, if you don’t mind waiting the time it can take to work through its processes (under 2 minutes on my 8th-gen i7 Lenovo X12 hybrid tablet), it’s always worth a try.

After I ran it on my first test PC, it cleared the error condition the troubleshooter found but couldn’t fix, and successfully downloaded and installed a pending Cumulative Update (CU).

If the silver bullet doesn’t work, then what?

On one of the two occasions when the batch file didn’t fix WU, I had to take my affected PC to the shop, where they were able to detect and diagnose a failing drive (SSD) that I did not catch on my own. Once that component was replaced, the system returned to normal working order.

On the other machine, I learned from the chatter around KB5053390 on Windows 11 Forum that others were having similar problems. After I used DISM to install that package manually, WU resumed working correctly. Thus, online research can also help prevent unnecessary flailing about when WU problems present.

Before you take a PC that gets to this stage to the shop, I’d recommend trying an in-place repair install. And if that fails, try a clean OS install next. Only then should a trip to the shop be needed. Hopefully, you’ll never get that far!

Additional reading on WU issues and errors

The Microsoft Learn collection called “Windows deployment documentation” covers the ins and outs of Windows deployment, including WU and updates. It includes a Troubleshoot section broken into two parts — namely, “Resolve Windows upgrade errors” and “Windows Update issues troubleshooting.” Both dispense lots of useful advice and techniques, plus discussions of (and references to) Windows Update error codes. For those who want to learn and do more about WU and its workings, these are must-reads.

This article was originally published in November 2022 and updated in April 2025.

The Social Security Administration (SSA) payment system has more than 60 million lines of COBOL code. That makes it one of the world’s largest, not to mention oldest, codebases. Everyone agrees it needs to be updated. But, sorry, there’s no way it can be updated in mere months.

Nevertheless, Elon Musk and his “Department of Government Efficiency” (DOGE) claim they can convert the ancient SSA code to Java or some other modern language in only a few months. Reminder DOGE isn’t a real department — and that modernization effort isn’t likely to happen anytime soon.

Seriously, I’ve seen this rodeo before. For example, the Internal Revenue Service (IRS) has been trying to replace the 200,000-line assembly language Individual Master File (IMF) with Java for decades. Hasn’t happened. The latest plan was to replace the IMF by 2028. I say “was” because President Donald J. Trump, Musk, and DOGE have cut the IRS modernization funds to zilch. 

As for SSA, it was going to replace COBOL with Java in 2017, a code translation expected to take five years; it failed. (That’s why we’re having this conversation now.) Part of the problem was due to the COVID-19 pandemic. But, as Ernst & Young noted, the agency also didn’t have a real plan for modernizing its legacy IT systems, nor one to identify and track the systems.

Guess what? There’s no real plan today, either. All we know is that Musk, yes-man Steve Davis and  nine kid software engineers, and some undefined AI program will miraculously transform the code into Java.

I doubt any of them can even read COBOL code. This is, after all, the same crew who thought there were hordes of 150-year-old fraudsters ripping off Social Security. What they didn’t realize was that COBOL doesn’t have a standardized way of dealing with dates, which led to records for numerous century-old-plus people who were not — I repeat- — not getting an unfair penny of government funds.

Why should they know? Over a decade ago, fewer than one in four colleges offered COBOL classes.

Thinking of dates, that’s not the only data in that codebase kept in formats incompatible with modern languages. Converting this data accurately is crucial and fraught with risks. Any errors during the conversion process could result in data loss or corruption, which would have catastrophic consequences for the millions of Americans who rely on Social Security benefits.

Commerce Secretary Howard Lutnick believes that if Social Security checks didn’t go out one month, his mother-in-law “wouldn’t call and complain.” Maybe she wouldn’t — after all, she has a billionaire son-in-law. I don’t. I bet the vast majority of the 73 million other Americans who depend on it don’t have billionaires in the family either. 

Even if these youngsters did grok the language, knowing COBOL is only part of the problem. You see, COBOL is more than just a coding language. It also incorporates organizational processes and policies. That’s not an approach modern developers deal with.

In addition, COBOL systems were not designed to integrate seamlessly with modern technologies. A migration requires extensive customization to ensure compatibility with newer software and hardware, which is both costly and complex. This integration challenge was exacerbated by the need to maintain the system’s functionality during the  transition, a delicate balancing act that demands meticulous planning and execution.

Meticulous is not the word I’d use for Musk and his move-fast and break-things bunch.

As Waldo Jaquith, former director of technology acquisition for the Inflation Reduction Act, said on Bluesky about the scheme, “I’m an expert in modernizing legacy government software systems. This is profoundly stupid and will definitely fail, and it’s just a question of whether our social security system fails along with it.

“I cannot think of a single legitimate reason for DOGE to perform such a modernization, especially on such a time scale.”

Jason Fichtner, a former SSA deputy commissioner, told CNBC: “If you start messing with the system’s code, that could impact those who are currently getting benefits now, and that’s a new front-and-center concern.

“You can’t just flip a switch one night and expect to be able to upgrade. It takes due diligence, and you have to understand the complexity of the programs,” he added.

In light of DOGE’s employee cuts at the agency, Fichtner and Kathleen Romig, director of Social Security and Disability Policy at the Center on Budget and Policy Priorities, warned in a column: “If Social Security’s computer systems experience an outage, which has happened twice in recent years, the agency may lack the expertise to resolve it.” 

Unless Congress acts soon to block DOGE’s misguided technology plans, Social Security, which has never missed a payment in its nearly 90 years, might finally not deliver American workers’ hard-earned retirement checks. Yes, legacy systems should be replaced and upgraded. No one’s arguing they shouldn’t be.

But waving a magic AI wand powered up by a political agenda isn’t the way to do it.

Enterprise IT leaders are facing a double-whammy of uncertainties complicating their data center building decisions: The ever-changing realities of genAI strategies, and the back-and-forth nature of the current tariff wars pushed by the United States.

“This is obviously a fluid situation. The stated goal of the [US] administration is to bring more development into the US,” said Forrester Senior Analyst Alvin Nguyen. “But with some of these activities, there is the potential that it draws some manufacturing and other capabilities of the data center away from the US.”

Nguyen, who advises enterprises on data center strategies, said the tariffs are adding complexity and uncertainty into the already volatile genAI data center strategies.

“Right now, there’s too much variability. With all of the tariffs, this may be the thing that slows down AI,” Nguyen said. “And if you slow down AI, that will slow down the data centers.”

Microsoft is urging Office 2016 and 2019 customers to upgrade to Microsoft 365 before support ends Oct. 14, but analysts said viable alternatives are available outside Microsoft’s walled garden.

“Continuing to use unsupported software can expose your organization to security vulnerabilities, compliance risks, and operational disruptions,” the company warned in a blog post.

Microsoft 365 is the cloud-based version of Microsoft Office, which also includes Teams, Copilot and web-based collaborative features. It is available only by subscription, with prices starting at $9.99 a month, or $99.99 a year for the Personal edition.

The alternative, a standalone desktop version of Microsoft Office, doesn’t have the AI and collaboration features, nor does it have Copilot or Teams. Users looking to replace Office 2016 or 2019 could also opt for the convenience of a one-time purchase of Office 2024, which costs $149. 

Microsoft in the blog post talked only about upgrading to M365, which never expires. Office 2024 support ends in October 2029.

Analysts said enterprise customers might find upgrading to Microsoft 365 worthwhile for its generative AI (genAI) tools, collaborative features and security. Or, depending on enterprise AI and productivity needs, they could jump ship for rivals such as Google Workspace. 

For enterprise customers worried about data privacy in the cloud, the desktop edition of Office or free open-source alternatives such as LibreOffice might be more attractive. The $149 price of Office 2024 might also be cheaper than Microsoft 365 in the long run, analysts said.

Microsoft wants to move customers to Microsoft 365 subscription services, said Jason Wong, vice president at Gartner for on app design and development.

“This makes support easier and lowers the cost of products for Microsoft, while at the same time it opens up many up-sell and cross-sell opportunities such as security products, the Power Platform tools, and of course M365 Copilot,” Wong said.

Basic Microsoft 365 editions for home users or business users don’t include Teams. But users can get features that include Intune, Defender, Clipchamp and Loop at higher subscription prices.

For those who only need Microsoft Office apps that include PowerPoint and Word, the standalone option could be attractive, especially if they already use something like Google or Zoho for mail, calendar, and document storage, said Irwin Lazar, principal analyst at Metrigy.

“For those wishing to take advantage of AI, an upgrade to M365 is a requirement. I expect that for [small and mid-sized businesses], Microsoft now offering M365 without Teams at a lower price could prove attractive,” Lazar said. 

Metrigy in a recent study noted that about 25% of Microsoft customers were evaluating the unbundled option. The study, Employee Experience Optimization: 2025, was published in November.

“Google’s recent price increases for Workspace are likely to help Microsoft, especially for SMBs, though Google includes Gemini AI now with Workspace,” Lazar said.

Enterprises would see value in moving from legacy on-prem, disconnected apps to Microsoft 365 though at this point those were probably Lotus or legacy on-prem Exchange/Sharepoint customers.  “The savings is likely to be minimal if customers are already using cloud-based services for document, email, and calendar,” Lazar said.

Some organizations are resisting the push to the cloud — primarily European-based companies with stricter data requirements and regulations. “Cost also plays a factor in staying on-premise, but typically organizations realize they won’t be getting the latest features and capabilities, like generative AI and Copilot, if they choose this path,” Wong said. 

Gartner sees clients evaluating rival suites to see what life looks like outside of the Microsoft ecosystem. “It typically comes down to familiarity of products and features for the workers, and the overall security and cost of ownership for IT to consider whether to switch or not,” Wong said.

Microsoft, for its part, painted a doomsday scenario to get users to upgrade to Microsoft 365 if they don’t quit Office 2016 or 2019 by the time support expires. 

“You may have started noticing limitations,” the company wrote. “Your apps are stuck on your desktop, limiting productivity anytime you’re away from your office. You can’t easily access your files or collaborate when working remotely or traveling, creating unnecessary friction for your team. Perhaps you’ve seen your company’s IT expenses creep upwards as you’ve added separate solutions for email, file storage, and virtual meetings.”

Readers of CIO, Computerworld, CSO, InfoWorld, and Network World this week sought to learn how corporate culture can affect digital transformation, wanted to know more about the rise of cloud-based ERP, and had questions about desktop-based office software.

Transform Faster

Disruptive forces dictate the need for speed when it comes to key digital initiatives. This week we reported 12 ways in which IT leaders are overhauling strategies and processes to streamline IT for quicker success. We explained how to transform: faster.  

Transformation is not only a question of technology, of course. And the readers of CIO were keen to understand how big a part organizational culture plays in transformation projects. According to Smart Answers: a lot.  

It says that organizational culture impacts time to business outcomes and the development of digital core competencies. And that a culture committed to digital transformation, such as a cloud-committed culture, can accelerate the adoption of digital technologies. Transformation requires a change in mind set, and culture is a big part of that. 

Find out: How does organizational culture impact digital transformation speed?  

SAP is Rising

Last week we reported on problems in migrations to SAP’s S4/HANA. This week we brought the better news that SAP adoption is surging (in Europe at least) as enterprises embrace cloud. But why? What is driving the trend of cloud ERP adoption? 

The accepted answer would be that organizations are looking to accelerate digital transformation efforts and seek greater operational efficiency through advanced technologies. Smart Answers takes that ball and runs with it, adding other factors: a broader shift toward scalable and efficient IT infrastructure; SAP’s RISE with SAP program, designed to facilitate cloud migration; and the cost savings from not having to refresh systems is another factor driving the adoption of cloud-based ERP systems. 

Find out: What is driving the trend of cloud ERP adoption?  

The Price of Productivity

Recently we reported that LibreOffice downloads are on the rise as users look to avoid subscription costs. We said that the free open-source Microsoft Office alternative is being downloaded by nearly 1 million users a week.  

It makes sense, right? Something free is better than something expensive. But is it a fair comparison. Readers of Computerworld wanted to know if a desktop productivity suite could compete with a cloud-based solution such as O365. The answer is yes… but Microsoft’s desktop software is the best solution. At least that is the opinion of Smart Answers.  

Find out What are the advantages of using a desktop office productivity suite over a cloud-based one?  

About Smart Answers 

Smart Answers is an AI-based chatbot tool designed to help you discover content, answer questions, and go deep on the topics that matter to you. Each week we send you the three most popular questions asked by our readers, and the answers Smart Answers provides.  

Developed in partnership with Miso.ai, Smart Answers draws only on editorial content from our network of trusted media brands—CIO, Computerworld, CSO, InfoWorld, and Network World—and was trained on questions that a savvy enterprise IT audience would ask. The result is a fast, efficient way for you to get more value from our content. 

Apple’s leadership is no doubt scrambling to identify a silver lining (if there is one) as the storm of US President Donald J. Trump’s punishing range of global tariffs rains heavily across the company’s supply chain. Even Apple’s attempt to mitigate the impact of anticipated tariffs on Chinese goods with big investments in manufacturing in India, Thailand, and elsewhere wasn’t enough. 

Most of the world will be affected by Trump’s tariffs, which the company is unlikely to be able to swallow whole without raising product prices.

The scale is huge — 54% (the 34% hike announced last night in addition to an existing 20% tariff) on Chinese imports; India gets a 26%, tariff, Vietnam, 46%, Taiwan, 25%, and Thailand gets a 36% slap.

Insert synonym for ‘ouch’ here

These tariffs are consequential. 

Rosenblatt analyst Barton Crocket estimates Apple must realistically prepare itself for an additional $39.5 billion in costs as a result, which will directly impact iPhones sold in its biggest market (the US) and made in China. This will also affect China’s economy in terms of lost sales.

Naturally, Apple’s stock price fell, and was off a dramatic 8.9% by early afternoon today.

To claw that cash back, the company will have little choice but to boost hardware prices across the board, including those goods outside the US. When it does, it will no doubt lean in on its services business in an attempt to help mitigate the consequences of these tariffs. What else can a business with a responsibility to its shareholders do?

One thing it will do is try to win exemptions on the application of these tariffs against its products. Perhaps, for example, Apple can point to the investment it and its partners are making in processor manufacturing in the US. The argument: any tariff should be waved as that component will eventually be made in the US, but the factories aren’t ready yet. Perhaps that is part of the reason semiconductors are to be exempted from reciprocal tariffs.

Apple might also consider shifting more production to India, particularly if the government there secures positive trade deals with the US. Other responses, according to analyst Ming-Chi Kuo could include raising iPhone Pro prices as consumers of those products might be more accepting. The company could also increase carrier subsidies, he said, as well as cutting trade-in values and, of course, squeezing suppliers for lower costs.

Another thing Apple can do is use this moment as a golden opportunity to re-purchase its shares. It announced a $90-billion-share repurchase budget in January, and the state-mandated damage its stock is currently taking means it will be able to buy more shares for the same money — assuming it doesn’t use some of this capital to defer the impact of the tariffs (unlikely).

Why manufacturing won’t come back

What Apple won’t be able to do is move all of its manufacturing to the US. There are several reasons, but perhaps the most significant one is that there is nothing like enough trained staff for some of the most advanced manufacturing jobs available in the US. There’s been little investment in training up people for those jobs, and there is no way that training can be delivered before the tariffs strike later this week. 

That means any manufacturer, including Apple, rushing to migrate more manufacturing to the US will choose to deploy automation and AI in their factories. Make no mistake, Apple and its manufacturers already make copious use of smart manufacturing systems, which means any manufacturing facilities they open in the US will be automated. Not only this, but there’s a matter of scale. There are physical, financial, and human limits to how many manufacturing lines can physically be built in any given time frame. As a result, building enough factories within a stone’s throw of the “Gulf of America” at the scale needed to meet US market demands is probably not going to happen. 

The art of the deal

Despite this, Apple may have one positive thing it can achieve on the back of these Trump taxes (which will soon be felt by US consumers in the form of higher prices): the new tariff against India can be directly seen as a challenge for the tariff-led protectionism that exists in that market. Ironically, India and other nations hit by these tariffs may now think dropping their own tariffs on US goods could help them at least reduce the tariff applied to their exports. That certainly seems to be the point of these taxes.

Will they work? 

Realistically, it is possible the US can convince some other nations to submit to its approach to taxation, though it could just as easily fan the flames of nativist nationalism, fostering retaliatory tariffs from others and further raising prices for US consumers. No man is an island, after all, and we are all involved in the economy mankind makes.

Follow the money

In my view, a lot now depends on how the administration chooses to deploy the tens of billions it raises through the exercise. Will this money be fed directly into the US economy in a positive way, or does the government simply intend to build up a huge pile of currency for bragging rights or tax cuts?

I don’t know the answer. But in the short term I do think it reasonable to predict Apple, like every manufacturer, must become accustomed to a higher cost of doing business in America.

While we wait, Apple’s stock is down sharply — economic gravity in action.

You can follow me on social media! Join me on BlueSky,  LinkedIn, and Mastodon.