<p><strong><a href="https://www.ft.com/content/11632d03-358c-454b-af95-d06d4072932b?syn-25a6b1a6=1">Anthropic tells investors quarterly revenue up nearly 50% to $16.8bn ahead of IPO</a></strong></p> The phrase "told investors" appears four times in this FT article, which reports on Anthropic's day of hosting investors on Wednesday. Lots of details in here from "<a href="https://simonwillison.net/2023/Nov/22/deciphering-clues/#hints-to-look-out-for">people with knowledge of the matter</a>".</p> <p>Key things I learned from the article:</p> <ul> <li>Q1 revenue (Jan/Feb/Mar) was "about $4.8bn"</li> <li>Q2 (Apr/May/Jun) was $11.5bn</li> <li>Q3 (Jul/Aug/Sep) was $16.8bn</li> <li>In Q3: $1.70 in revenue for every $1 spent on "computing power"</li> <li>Adjusted operating margins (excluding stock-based compensation) were 11%</li> </ul> <p>This point about distillation is worth quoting in full (highlight mine):</p> <blockquote> <p>The start-up has also clamped down on “distillation” — a controversial practice whereby rival AI labs train their models on the outputs of more advanced ones.</p> <p>When rival companies “distil” their models from Anthropic’s models, that involves heavy usage and spending.</p> <p>Anthropic told investors on Wednesday <strong>distillation accounted for 5 per cent to 7 per cent of revenue</strong> and that they had cut it back “materially” in the third quarter of the year, said four people with direct knowledge of the matter.</p> </blockquote> <p>Napkin maths says that if distillation was 5% of their Q2 revenue (before they cut it back in Q3) then unnamed "other labs" spent over half a billion dollars distilling Anthropic's models in that quarter!</p> <p>Also notable: Anthropic appear to no longer be leaning on "annualized run-rate revenue" (<a href="https://simonwillison.net/2026/May/29/anthropic/">previously</a>) in these communications to their investors. <p>Tags: <a href="https://simonwillison.net/tags/ai">ai</a>, <a href="https://simonwillison.net/tags/generative-ai">generative-ai</a>, <a href="https://simonwillison.net/tags/llms">llms</a>, <a href="https://simonwillison.net/tags/anthropic">anthropic</a></p>

Celebrate Teachers’ Day with AI for Educators in Brazilian Portuguese and see how Study Mode can help teachers support active, step-by-step learning.

"Why have we gotten to the point where we have permitted militaries of various states to use civilian commercial data centers for their military purposes?"

Hugh Son / CNBC: BofA says it hired 1,000 contractors in Q2 to support 3,000 internal staff in patching proprietary and vendor software flaws exposed by AI models like Mythos  —  Bank of America has hired an army of external contractors to help it patch software vulnerabilities exposed by advanced artificial …

MacPaw launches Eney for Mac and iPhone, with proactive task suggestions, meeting notes, app integrations, approval controls, and a free 14-day trial. The post MacPaw Launches Eney AI Assistant for Mac and iPhone: Price and Features appeared first on TechRepublic.

Jeffrey Kampman / Tom's Hardware: Nvidia's RTX Spark Superchip in the Surface Laptop Ultra outperforms Apple's M5 Max MacBook Pro by ~25% in agentic AI completion benchmarks using Qwen models  —  Laying the foundation for the reinvention of the PC in the agentic AI era  —  Nvidia CEO Jensen Huang announced the RTX Spark platform …

Dominic Preston / The Verge: Ledger unveils Atom, an $899 data-only Android-based smartphone companion device launching in summer 2027, as it pauses some sales following hardware tampering  —  Atom is a data-only Android device designed to work alongside your smartphone. … Ledger has announced its latest crypto …

I read the exact answer distribution from the logits of open models, base and instruction-tuned, and turned it into bits and guesses-to-crack. The post How Many Guesses Does It Take to Crack an AI's "Random" Number? appeared first on Towards Data Science.

If AI systems have been humming along nicely for weeks or months at a stretch, will the human overseeing them be able to take over when things go wrong? Sebastien Bozon/AFP via Getty Images Imagine an artificial intelligence system that has performed reliably for months. Its recommendations are almost always right, and the employee reviewing them has learned that close scrutiny rarely changes the outcome. Then something unusual happens. The AI makes an error or can’t proceed and needs human intervention. The employee is expected to step in and remains responsible for the outcome. But does the employee know how to do the task? I’m a professor of management information systems and the academic director of Drexel University’s Center for Applied AI & Business Analytics. In my research on human-AI relationships, I call this latent risk: the possibility that people may lose capabilities their organizations still depend on them to exercise. AI can continue to perform well while the people overseeing it become less practiced at exercising independent judgment and less prepared to intervene. The risk becomes latent when the organization still depends on that human capability to step in but has little cause to notice that that capability may be weakening. The concern is not entirely new. The idea of latent risk builds on decades of research into how people respond to automation. For example, studies of automation bias have shown that people can miss problems or follow automated recommendations even when other evidence suggests they should not. What is different today is the role AI is beginning to play in organizations. It is beginning to assume larger parts of the analysis and decision-making that people previously performed. Human in the loop explained. The success paradox The better AI becomes, the more the nature of human work changes. Work that once required an employee to analyze a problem and reach a decision may instead be performed by AI, leaving the employee to review its recommendation. The employee is no longer doing the same work with AI assistance – rather, their job has shifted from doing the work to overseeing what the AI does. This shift is already beginning to change how some companies train employees. Accounting giant Deloitte, for example, is redesigning its graduate audit training as AI takes over some of the manual work traditionally done by junior auditors. The firm is introducing simulations so trainees can still develop the judgment they will need as they move into more complex work. More than four decades ago, cognitive psychologist Lisanne Bainbridge described one of the “ironies of automation”: As systems take over more of the work, instead of simply having their work loads lightened, people can be left facing new challenges in monitoring the systems and handling the unusual situations the systems cannot. Later studies found another problem: As people become less involved in a task, they can become less prepared to step in when needed. Researchers have called this the out-of-the-loop performance problem. This reduced involvement may affect more than a person’s readiness to intervene. It can change the capabilities available to the organization. What work remains for people does not necessarily become easier. It may become harder. These are the situations where experience and independent judgment matter most. The result is a paradox: People may be called upon to exercise independent judgment less often, even as the judgments left to them become more difficult. The accountability problem Accountability may not change as quickly as the work does when organizations deploy AI systems. An employee may now oversee far more decisions than they make themselves, while remaining responsible for the outcomes. The organization’s expectations may still reflect how the work was divided before AI took on a larger role. The organization can therefore end up holding people accountable for decisions over which they have less control. This creates a mismatch. Research on meaningful human control suggests that responsibility should match a person’s authority and ability to act. An employee may retain the authority to override an AI recommendation and remain responsible for the outcome, yet lose the capability to do so. Authority, responsibility and capability can begin to separate. Latent risk becomes an organizational design problem, not simply a human performance problem. The measurement problem Organizations typically judge AI-enabled processes by measures such as accuracy and productivity. Those numbers may look good even as the human role is changing. Suppose employees override AI recommendations less frequently over time. Is that a sign of growing dependence? Perhaps. But it could also mean that the AI has become more accurate and there is less reason to intervene. The number of overrides alone tells little about whether employees remain ready to exercise their own judgment. The key is to understand how the division of work between people and AI is changing. Which decisions have migrated from people to AI? What responsibilities remain with people? Looking at the performance of the employee or the AI separately may miss what is changing between them. The governance problem Organizational governance – defining roles and responsibilities – is also under pressure from AI. Someone may remain formally in the loop with an AI system even as their actual role changes. Defining human and AI roles when a system is introduced is only a starting point. As AI improves and people become accustomed to using it, those roles are likely to change over time. There is nothing inherently wrong with that change. The problem arises when the relationship changes but the assumptions surrounding it do not. It’s important for organizations to periodically ask: What relationship between people and AI did we design, and what relationship do we actually have today? In short, organizational governance needs to be dynamic. Maintaining the human backstop Not every human capability needs to be preserved. If AI can perform a task well, there is little reason to require people to keep doing the same work. Organizations need to decide which capabilities they still expect people to have when human judgment is required. If people may still be called upon to use a capability, it needs to be maintained. Overseeing AI is not the same as exercising independent judgment. If people rarely make the decision themselves, the organization cannot assume they will be ready when they need to. One way to test this is to occasionally ask employees to make a decision without first seeing the AI recommendation. The organization needs to know that the human backstop it relies on will actually work when needed. As AI takes on more work, some human capabilities will inevitably be used less often. Some may no longer matter. Others may still be essential when AI reaches its limits. The challenge is knowing the difference and making that choice deliberately. For me, the central question is this: Which human capabilities does an organization still depend on, and will they be there when needed? Murugan Anandarajan does not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

People across the U.S., England and Australia agree that teachers are most like coaches when considering comparable professions. Drazen Zigic/iStock via Getty Images Plus Three answers, three countries, one open-ended question: “Teaching is similar to what other occupation or activity?” We asked this question to more than 3,000 people across the United States, England and Australia between May 2025 and March 2026. Coaching was the front-runner across all three countries. But the similarities partially diverged there. Respondents in England and Australia more often compared teaching with professional roles such as nursing and social work. U.S. respondents, meanwhile, more often likened teaching to being a parent and religious roles, such as a cleric or pastor. In the three countries, however, one other idea stood out: Teaching was primarily associated with a caring relationship. We have previously examined how Republicans and Democrats in the U.S. find common ground on what makes a good teacher, despite the polarization in the country on education and many other issues. This time, we wanted to know whether people across countries had similar ideas of teachers. Common ground In all three countries, respondents most often described teaching as centered in the relationship between teacher and student: a form of coaching, caregiving, parenting and other types of personal support. Coaching was the most common answer in each country, but its prominence varied. Coaching was named by 30% of U.S. respondents, compared with 23% in England and 23% in Australia. The differences became more pronounced when participants named other occupations and activities. Respondents in England and Australia were more likely to associate teaching with institutional, credentialed and paid professional roles, such as nursing, lecturing and social work. “Nursing. They care for and look after people whilst they are at their most vulnerable. Teachers do a similar thing by caring and teaching young people some of the skills they will need later in life,” a 39-year-old Australian man told us. “A lecturer, you are passing on knowledge to people, usually the younger generation,” another 32-year-old man in England said. Compared with respondents in those countries, U.S. respondents were more likely to describe teaching through caregiving roles, such as parenting, and through religious figures, such as priests, pastors or rabbis. “Being a parent, you are caring for young minds and shaping them through learning,” a 33-year-old woman in the U.S. told us. “As a preacher, he teaches the word of God. How they deliver the word determines how well the people will understand it,” a 23-year-old woman in the U.S. said. While international comparisons are informative, they do not represent a single view of teaching. Instead, they suggest that respondents in three countries likely drew on somewhat different cultural repertoires when describing the profession. We were surprised by just how much agreement we found. Despite differences in schooling models and cultural norms, people in the United States, England and Australia had remarkably similar ideas about how relationships are at the heart of teaching, closely echoing our other work showing that Republicans and Democrats view teachers similarly. While AI is fueling educational debates, people don’t associate the technology with teachers. Connect Images/iStock via Getty Images AI is not at the center of people’s thinking In September 2026, The New York Times conducted an informal survey of 37 education researchers who gave 30 ideas about how to fix schools. Their list consists primarily of interventions such as more per-pupil funding, banning cellphones and restricting artificial intelligence in the classroom. These ideas appeared far less often in our participants’ explanations of what teaching resembles. In fact, out of more than 3,000 responses to our survey, AI came up only a handful of times and never as the reason teaching and another occupation were said to be alike. Given how central AI has become to political and academic debates about the future of education, its near-total absence in these public descriptions of teaching is notable. At the same time, experts and citizens agree that what separates very good teaching from poor execution is likely the quality of the relationship between educators and children. We think that serious conversation about student achievement must include clear descriptions of very good teachers: the practices they use to encourage and enable students to learn. Our research shows that teaching should make room for both the transmission of knowledge and the meaningful and caring relationships that make learning possible. The authors do not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.

Bloomberg: Amazon launches two AI wearables, the $75 Bee Frontier wristband and $100 Bee Eclipse magnetic phone-attachment recorder, with a $12/month Bee Pro subscription  —  Amazon.com Inc. launched two new wearables that use artificial intelligence to record and analyze a person's day …

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. Pig kidneys that were “frozen” for a year have been brought back to life Pig kidneys that were cooled to -150 °C and stored in a freezer for months have been…

Cryptocurrency firms are increasingly turning to the world of traditional finance for their next leg of growth by positioning themselves as competitors to Wall Street’s entrenched institutional businesses.

Amazon-owned Bee has launched its first new AI devices since its acquisition, including an updated wearable and a credit-card-sized recorder that attaches to smartphones. The devices can summarize conversations, create to-do lists, and integrate with Alexa and Amazon shopping, with prices starting at $74.99.

Discussion | Link

A Delaware jury found that Arm breached its licensing agreement with Qualcomm but awarded the chipmaker just $2 in damages, leaving a potentially multibillion-dollar dispute over royalty payments unresolved. The jury ruled on October 13 that Arm had failed to provide software patches and development tools required under its agreement with Qualcomm. It also found that Arm had interfered with Qualcomm’s business relationship with Meta Platforms, while rejecting several other allegations against Arm. Qualcomm is seeking to invoke a contractual provision that would allow it to stop paying royalties to Arm for five years. Arm has argued that the provision is unenforceable, leaving US District Judge Maryellen Noreika to decide whether Qualcomm is entitled to the remedy, Reuters reported. Qualcomm welcomed the finding that Arm breached its licensing obligations, while Arm emphasized that it prevailed on several other claims and was ordered to pay only nominal damages. The verdict follows an earlier dispute over Qualcomm’s 2021 acquisition of chip designer Nuvia. Arm argued that Qualcomm needed its consent to use technology developed under Nuvia’s license. A jury sided with Qualcomm on key issues in December 2024, and a judge ruled in its favor on the remaining claim in September 2025. Arm has appealed that judgment. The dispute comes as Arm expands beyond its traditional business of licensing processor technology. In March 2026, the company unveiled its first data center CPU, developed with Meta Platforms. Qualcomm, meanwhile, has been expanding beyond smartphones with its Snapdragon processors for Windows computers. PC market challenges unresolved The Arm–Qualcomm dispute is unlikely to significantly impact enterprise adoption of Snapdragon-powered Windows laptops, said Pareekh Jain, CEO of Pareekh Consulting. Charlie Dai, principal analyst at Forrester, said the remaining dispute is more focused on licensing remedies and commercial terms than on the viability of existing products. While the ruling reduces one source of uncertainty for Qualcomm, its prospects in the PC market remain dependent on factors beyond the legal dispute. The chipmaker accounted for less than 1% of the overall PC market, excluding tablets, in the second quarter of 2026, said Ben Yeh, principal analyst at Omdia, citing the firm’s data. “At present, the main challenges facing Qualcomm’s expansion in the PC market remain consumers’ limited confidence in the compatibility of Windows apps for the Arm-based architecture, their continued reliance on existing suppliers such as Intel and AMD, and the relatively high price of Qualcomm’s chips,” Yeh said. Yeh does expect Qualcomm’s share of PC sales to increase modestly in the second half of 2026, supported by the launch of its Snapdragon C products and supply shortages at Intel. For enterprise buyers, the more immediate considerations concern how well Snapdragon-powered Windows PCs fit into their existing IT environments. Dai said CIOs should evaluate device management and AI capabilities alongside application compatibility and vendor support when considering Snapdragon-powered laptops. Deepika Giri, head of research for AI, analytics and data in Asia Pacific at IDC, said the remaining legal uncertainty was manageable, but advised buyers to obtain multiyear commitments on product roadmaps and support from PC manufacturers and Qualcomm. Jain recommended maintaining a mix of Arm-based and x86 devices to reduce procurement risks. Arm’s relationships with chipmakers face scrutiny The jury’s finding that Arm failed to provide required development tools could bring greater scrutiny to its obligations under licensing agreements. Dai said clearer licensing obligations and stronger protections for licensees could encourage more custom processor development, potentially increasing competition. “What changes is leverage: licensees will push harder on terms and invest more in custom silicon,” Giri said. “Over the next few years, enterprises could benefit from broader hardware choices and improved performance differentiation, although any increase in licensing costs or legal complexity could also put upward pressure on processor pricing and ultimately on enterprise hardware spending,” Dai said. Arm’s expansion into chip development could encourage licensees to diversify without abandoning its technology altogether. Giri expects companies to continue using Arm for its established software ecosystem while also investing in RISC-V and their own processor designs. A wider range of processor architectures could give enterprises access to more specialized hardware, but could also complicate software certification and infrastructure planning, Dai said. To manage that complexity, Giri recommended that enterprises adopt architecture-agnostic software and procurement strategies, reducing their dependence on any single processor platform.

As AI enables smaller, faster teams, companies should hire curious, adaptable people who challenge established practices and can be trusted to make independent decisions, according to guest author Eric Francia, who writes that giving employees real responsibility, encouraging questions and investing in continuous learning helps teams innovate, stay engaged and retain talent.

For the third quarter of 2026, global cybersecurity- and privacy-focused startups pulled in a total of $6.8 billion, per Crunchbase data. That’s up about 50% from the prior quarter and tied for the highest quarterly tally in years.

New York-based Chariot Claims has raised $10 million in Series A funding led by Silverton Partners to help consumers pursue claims over alleged privacy violations, hidden fees and subscriptions that are difficult to cancel.

Pasting in the whole chat history cost 20 to 31 times more than looking up notes, and it did no better on the tasks. The post I Tested Four Ways to Give an AI Assistant a Memory. The Obvious One Was the Worst Value. appeared first on Towards Data Science.