[Werner Vogels, Amazon CTO] Technology is now deeply intertwined with our daily lives around the world, influencing almost everything. It shapes our relationships, the care we pursue, how we work, the actions we take to protect ourselves and even what we learn and when. On that reality alone, it is not unreasonable for it to feel like the dystopian nightmares depicted by E.M. Forster or Ernest Cline.

But we are now standing at the threshold of a fundamentally different phase. We are already catching glimpses of a future that values autonomy, empathy and individual expertise. Interdisciplinary collaboration is driving discovery and creation at a breathless pace. In the coming year, a transition will begin to a new era in which humans stand at the centre of the cycle and AI enters it. That cycle of change will create a major opportunity to solve problems that truly matter. The starting point is addressing one unintended consequence of the hyperconnected society: loneliness and the absence of companionship. It means turning the forces that created the problem into part of the solution.

Redefining “companions” for those who need them most

Loneliness has reached the level of an epidemic, affecting 1 in 6 people worldwide, and the World Health Organization has defined it as a public health crisis. Social isolation increases the risk of death by 32 percent, with an impact comparable to smoking, while loneliness raises the risk of dementia by 31 percent and stroke by 30 percent. The crisis is particularly pronounced among older people, with 43 percent of adults aged 60 and older reporting loneliness, and the impact worsening among those aged 80 and older. As ageing populations place a major burden on care systems worldwide, we are standing at the threshold of a fundamental shift in the human-technology relationship, one that seeks to confront the loneliness crisis head-on through genuine emotional connection.

Just 10 years ago, forming a meaningful emotional relationship with a robot was the stuff of science fiction. Today, rising older populations, advances in cutting-edge AI and a global loneliness crisis have combined to create the conditions for a “companion revolution.” We are witnessing a move away from transactional interactions with simple devices toward forming relationships with physical AI that shows increasingly nuanced emotional intelligence and responsive behaviour.

Clinical evidence that companion robots are effective in easing loneliness is very strong. Long-term care facilities and hospitals in Canada have already adopted robots such as Pepper, Paro and Lovot to support mental health and well-being. In clinical studies of Paro, 95 percent of dementia patients who interacted with the robot regularly showed noticeable positive changes, and anxiety, depression and loneliness fell by measurable amounts. Reduced medication use and improved sleep patterns were also observed.

The therapeutic effects of companion robots are not limited to elderly patients. In research on the “Huggable” social robot at Boston Children’s Hospital, paediatric patients showed a much stronger willingness to emotionally connect and interact with the robot than with on-screen virtual characters or medical staff. In one notable case, a child who showed extreme stress during drug administration stayed calm and focused when with Huggable, and a procedure that had previously been close to traumatic went much more smoothly.

Why are these robots so effective at easing loneliness and forming bonds with us? Biologically, humans are designed to project intention and vitality onto self-moving objects in the space around them. As shown in research by Kate Darling, a researcher at the MIT Media Lab, people treat robots more like animals than like devices. They name them, feel protective instincts and form real emotional bonds. This is not limited to sophisticated humanoid robots. The fact that 50 to 80 percent of Roomba users give their vacuum cleaners names like family members shows as much. When something moves freely in our space and seems to have a purpose and personality, we instinctively respond by trying to form a relationship. That biological trait underpins how companion robots can ease human loneliness in a way traditional devices cannot, by providing consistent emotional presence.

Amazon’s Astro team has observed cases in which people, over time, form emotional relationships with companion robots beyond simple interactions. Unlike existing smart home devices, Astro has mobility, an expressive visual interface and proactive functions that enable it to move around the home on its own to find users, such as for medication reminders or checking on family members, fostering natural attachment. Emotional expression through head movements and facial expressions creates an anthropomorphised presence that strongly resonates with users. Many families name Astro and treat it like a family member, and observations also show clear longing when the robot is briefly not at home. That shows a fundamental shift in perception toward accepting robots not as simple tools but as companions.

One case we observed involved a family with a disabled child that bought Astro as a companion for hours when professional caregiving services were not available. The robot provided consistent presence and interaction to fill an important care gap and reduced the family’s emotional and economic burden. Companion robots have now advanced to the point where they can offer practical care support as well as meaningful emotional bonds that ease feelings of isolation.

This companion revolution does not replace human caregivers. It creates a new collaborative model in which technology and people provide care together and respond to loneliness. Robots handle routine monitoring and provide steady emotional presence, serving as stable and unbiased companions that ease isolation. That allows humans to focus more on complex decision-making and deeper relationship-building.

As people build deeper trust with robot companions, companies developing the robots must put strong controls in place to prevent that trust from being exploited to influence users’ decisions or manipulate beliefs. When developed responsibly on the basis of such safeguards, technology can be used in its most desirable form. It expands our capacity to support those who need help most while keeping humans at the centre of care.

Dawn of the renaissance developer

Tools change, but fundamentals do not. As generative AI reshapes how software is made, the familiar claim is resurfacing that developers will soon become useless. But as history shows, this does not signal the end of developers. It marks the dawn of a new kind of person: the “renaissance developer.”

You will have heard these rumours. You will have read provocative headlines saying AI will make developers useless. The claims say anyone can code now, that you only need to describe what you want and the tools will do the rest, and therefore the era of professional developers is over.

We have heard these claims many times. Early assembly programmers were warned that compilers would make them useless. Compilers instead raised the level of abstraction and enabled far more people to participate in software development. Work that once required deep hardware expertise shifted into a realm of logic and creativity. As software became something more people could build, new industries emerged. Companies, research institutes and universities suddenly gained the ability to make their own tools.

When cloud computing emerged in the 2000s, operations engineers voiced similar concerns. They feared automation might make them useless. Cloud computing instead lowered barriers to experimentation and drove an explosion of new projects, new companies and new engineering roles. Each time simplification happened, demand grew even larger.

Technological leaps always follow a similar pattern. Tools evolve, workflows change and complexity keeps increasing, but the core qualities that make great developers do not change. Creativity, curiosity and systems thinking still define the essence of the field.

We have repeatedly seen that lowering barriers to entry does not weaken human expertise but strengthens it. Generative AI can produce code in seconds, but the “garbage in, garbage out” principle still holds, and we must remember that the output can now be “plausible-looking garbage.” AI does not sit in budget meetings where executives debate whether to prioritise cost or performance. It cannot understand context such as a customer service system needing 99.999 percent uptime while an internal reporting dashboard can be briefly down during a revenue peak. It also cannot read that when a stakeholder says “build it fast,” it may carry the hidden meaning of “build it cheaply.”

The political factors, constraints and implicit priorities that drive technical decisions are highly subtle, and developers who understand why they matter to the people who use what is built and the people who pay for it are essential.

Before painting the Mona Lisa, Leonardo da Vinci dissected cadavers to understand muscle structure, studied the flow of water to design canals and observed birds’ movements to imagine flying machines. His “Vitruvian Man” was not merely an artwork. It was a diagram explaining human proportions and a philosophical declaration exploring humanity’s place in the world. Like the Renaissance masters who combined art, science and engineering, developers in a world assisted by AI must become versatile modern “renaissance developers” to succeed.

They understand that systems are living, dynamic environments that ripple change across everything from services, APIs, databases and infrastructure to people. They communicate clearly in ways both humans and machines can understand. Especially as AI grows increasingly confident in its own errors, they take responsibility for the quality, safety and intent of what they build. They have domain knowledge that AI cannot replace, including an understanding of business, customers and real-world constraints. They keep learning.

The basic principles that have made great developers do not change. But like the great Renaissance thinkers who refused to be confined to a single field, developers can no longer remain isolated. They must draw the bigger picture. This is that moment. A new era is opening for developers, and their value is higher than ever. Developers’ creativity is also needed more than ever. Keep building, stay curious and keep solving the world’s hardest problems.

An era when only quantum-safe is truly safe

Malicious actors seeking the arrival of the quantum computing era are already collecting personal information, financial records and state secrets. Most organisations have assumed they would have years to prepare, but that assumption is no longer valid. Advances in error correction and algorithmic efficiency have shortened response timelines, and the window for proactive defence is closing quickly. In the coming year, a post-quantum mindset will be required. It must cover everything from cryptographic systems that protect the most sensitive communications to education that trains quantum engineers to handle them.

Until recently, people were not even sure quantum computers could really exist. Just 3 years ago, when I spoke with John Preskill, it still seemed that developing quantum hardware capable of solving hard problems would take decades more. Since then, we have been witnessing the schedule begin to shorten.

Important advances are continuing across quantum hardware and architecture. AWS unveiled the Ocelot chip, which implemented quantum error correction with hardware efficiency that reduces overhead by up to 90 percent compared with existing approaches. Google’s Willow chip demonstrated that the error rate decreases exponentially as code distance increases. IBM said it would build a fault-tolerant quantum computing framework by 2029.

Error correction has long been the core challenge in building scalable quantum computers, and the pace is now accelerating rapidly. There is an expectation that quantum computing will drive innovation across fields ranging from medical research to investment. But the area that must be considered most seriously right now is security.

The problem lies in how we protect data today. Malicious actors have been collecting encrypted data for years and waiting for computing power capable of decrypting it to emerge. Much of our digital security relies on public-key cryptography, and mathematical problems such as RSA and elliptic curve cryptography that are hard for classical computers become trivial for quantum computers running Shor’s algorithm. Unlike symmetric cryptography, which can be strengthened by using longer keys, public-key systems need an entirely new mathematical foundation to survive in the quantum era.

A study published in May this year found that a 2,048-bit RSA integer can be factored with fewer than 1,000,000 noisy qubits, down by nearly 95 percent from estimates 6 years ago of about 20,000,000. Within about 5 years, quantum computers could emerge that can neutralise most RSA and ECC used to protect internet communications, financial transactions and sensitive personal data.

Preparation can no longer be delayed. Action must begin now, and organisations must move on three fronts. They should adopt post-quantum cryptography, or PQC, where possible. Where they cannot, they should plan to update or replace physical infrastructure. They should also develop quantum-ready talent to support the transition.

The good news is that PQC solutions already exist and can be deployed immediately at the operating system level, browser level and in cloud environments. Major technology companies are adopting NIST standards such as ML-KEM to ensure interoperability and security. Microsoft released PQC tools for Windows and Linux, and Apple integrated quantum-safe protocols into the latest iOS and macOS versions. Google switched Chrome to quantum-resistant encryption. AWS also deployed these standards across KMS (Key Management Service), ACM (Certificate Manager), CloudFront, Secrets Manager and AWS-LC. Detailed migration plans are also already in place. But this is only the beginning.

The transition becomes most complex in the physical world. Think of devices connected to home networks such as smart TVs, thermostats and connected refrigerators. Many devices around us rely on cryptography, like the access key systems at hotels where we have stayed. Power companies have already deployed smart meters in the millions that use current cryptographic standards but lack the processing power to run post-quantum algorithms. Power grids, water treatment systems and transport infrastructure face similar constraints because of embedded devices that are difficult to upgrade. Scaling up the number of such devices requiring physical updates into the millions makes the size of the problem clear.

These constraints will pressure companies to find more creative solutions. New approaches will emerge, such as hybrid methods that place quantum-safe gateways in front of existing devices or deployment models that replace hardware sequentially without interrupting critical services. This is no longer a simple IT security project. It is an enterprise-wide transition spanning engineering, logistics, manufacturing and operations.

Finally, there is the talent issue. A report by the UK Quantum Technology Taskforce forecasts 250,000 new quantum computing jobs by 2030, rising sharply to 840,000 by 2035. As I noted 2 years ago, “higher education alone cannot keep up with the pace of technological change.” Organisations that invest now in quantum education and training will gain a competitive advantage that will not be easily matched. The quantum era demands new combinations of expertise that are rare now but will become essential in the coming years. The challenge for companies is creating incentives that draw people into the quantum field, whether through university education or other learning pathways.

Quantum technology is much closer than we imagined. Companies that adopt comprehensive quantum readiness strategies, including PQC adoption, quantum talent development and plans to transition physical infrastructure, will protect data and gain new capabilities in secure computing, privacy-preserving AI and trust-based data sharing. Cloud-native organisations can transition smoothly through updates managed by cloud providers. Infrastructure-dependent companies can also survive if they start planning the transition now. But companies that fall behind now will face vulnerabilities without a response when quantum computers mature. The era when only quantum-safe is the only safe is not far off.

Defense technology that changes the world

War has changed dramatically. Hand-to-hand combat is now a last resort, and wars are fought from hundreds, sometimes thousands, of miles away, through screens with controllers, keyboards and mouse clicks. Investment in military technology is surging in both the public and private sectors, and the pace of innovation is accelerating sharply. In the next few years, the time from battlefield to civilian applications will shorten, fundamentally reshaping global infrastructure, emergency response systems and medical systems.

The lineage of innovative civilian technologies born from military needs is striking. The pioneering work of Admiral Grace Hopper, who developed the Mark I computer for the navy, later led to COBOL, which supported business systems for decades. DARPA research gave birth to the internet and GPS, technologies now so essential to our lives that we almost forget their military origins. Radar technology developed by Britain in the 1930s evolved into air traffic control systems and also led to an unexpected consumer technology: the microwave oven. The EpiPen also began with Cold War-era research into antidotes for nerve agents, and millions of people worldwide now rely on it every day.

But these transitions have unfolded gradually over a long time. Successfully transferring technology from the battlefield to civilian domains requires significant cost reductions, improvements in manufacturing processes and clear market validation before commercial viability is secured. Historically, that process took 10 to 20 years. But change is beginning at this point.

What has changed is not the scale of investment but the fundamental approach to innovation. Companies such as Anduril Industries, which posted $1.0 billion in revenue in 2024 and grew 138 percent year-on-year, and Shield AI, which had $267 million in revenue the same year, operate more like technology startups than traditional defence contractors. They design from the outset with dual-use in mind and treat civilian applications not as an afterthought but as a core business model. This shift effectively removes the traditional adaptation steps that have extended technology transfer by years.

Think of what is happening in conflict zones around the world. Under extreme pressure, technology is becoming more sophisticated. Software updates for autonomous systems happen weekly, not annually. AI algorithms train on real data and can improve overnight. This creates feedback loops that operate on days, not decades. A scene of a Ukrainian farmer using a consumer drone for reconnaissance and sharing information through encrypted messaging apps shows military and civilian technologies merging in real time.

Beyond conflict zones, night-vision goggles once used only by special forces now help navigation for rescue helicopters and are used in wildlife conservation. Tactical edge computing, refined to operate in disconnected environments, enables remote medical clinics and industrial operations in regions with limited infrastructure. Autonomous systems developed for military logistics are applied to address agricultural labour shortages and make food production more efficient and sustainable, and are also used immediately in power plants, wind farms, search and rescue work and port security. Innovation in military robotics is also driving new solutions that can be used for urgent humanitarian purposes across industries that affect billions of lives.

Medical systems, emergency services and infrastructure operators must prepare for capabilities arising from today’s defence investment not in 20 years but within 2 years. Organisations that understand the accelerated timeline will gain a significant advantage in solving major problems, from disaster response and food security to access to healthcare in remote areas.

Technologies being refined today under extreme pressure will spread to the public in both wartime and peacetime. These technologies, designed from the beginning to meet both military and civilian needs, are arriving even now. The old model of a decades-long adaptation cycle is being replaced by direct diffusion. Organisations that recognise this as not “evolution” but “disruptive innovation” will solve problems that affect billions of lives.

Personalised learning meets limitless curiosity

Every student deserves to meet educators who understand how they learn best, spark curiosity, respect individuality and nurture creativity. For most of human history, private tutors were a privilege reserved for the wealthy. That era is ending.

When I look back on my own education, the most meaningful moments were not lectures in classrooms filled with students. They were conversations with teachers who tried to understand how I think, identified where I was confused and willingly spent time explaining in ways that fit me. Those teachers were rare.

For most students worldwide, personalised teaching remains a luxury. Schools are designed around efficiency, not diversity. We have standardised education, including what students learn, when they learn it and how we measure success. Education researcher Ken Robinson documented for decades how traditional education systems are run around uniformity over diversity and compliance over curiosity. He pointed out that in parts of the United States, 60 percent of high school students drop out. But the dropout crisis is only the tip of the iceberg. Children who attend school but are not engaged, do not enjoy it and gain no real benefits are not included in these statistics.

Artificial intelligence has the power to fundamentally change how we approach education. Children are natural learners. They keep asking questions until adults surrender. The only thing that limits their curiosity is whether they can access someone and tools to answer those questions. AI therefore adapts to each child’s way of thinking rather than forcing every student into the same system and learning path. It answers “why” as many times as a student asks, explores related topics that spark interest and adjusts explanations until understanding clicks. This creates a safe space where students can fail, try again and ask questions without judgment. It is not limited to STEM. AI enables students to explore arts, languages, music and the humanities. Most importantly, AI performs what great teachers have long done: drawing out students’ natural passion for learning rather than suppressing it.

Students can now access educational services from AI systems for $4 a month. Khan Academy’s Khanmigo attracted 1.4 million students in its first year, exceeding all forecasts by 1,400 percent. Anthropic started the world’s first nationwide AI education pilot programme in Iceland. A UK survey by UCAS found the share of students who said they use AI tools surged to 92 percent this year from 66 percent last year. This is not an experiment. It is a large-scale system in real operation. The change is also spreading quickly across India, Brazil and Africa. Physics Wallah serves 46 million students and grew revenue 250 percent. UNESCO’s CogLabs operates in 35 countries using the smartphones students already have. Amazon also launched a $100 million Education Equity Initiative to help underserved students gain AI skills.

Generation Alpha already views AI differently from us. Cultural anthropologist Rob Scotland, in a recent TEDx talk, described 16-year-old students who used ChatGPT and TikTok during a maths class to build their own curriculum. When asked why, they said, “We just wanted to try something different.”

For adults, AI is a tool. For Generation Alpha, AI is an extension of thinking. They have erased “impossible” from their operating system and replaced it with “not yet.” AI tutoring is effective because it nurtures that curiosity. When students use AI tools, their willingness to take on hard tasks rises by 65 percent. A Duke University study found AI-supported interventions raised the IQ scores of autistic children by up to 17 points. This is not just an improvement in test scores. Students who learn in an environment where “I don’t know yet” is a starting point, not a failure, approach difficulty itself differently.

To be clear, teachers are not disappearing. What changes is what teachers do. We face a global teacher shortage, and teachers should not spend most of their time on work that does not scale and can be automated, such as grading, administrative tasks and answering repetitive everyday questions. AI frees teachers from this heavy load so they can work more creatively, provide more personalised education and keep students engaged. Research supports this. Teachers who use AI tools save an average of 5.9 hours a week, equivalent to about 6 weeks per school year. It also helps reach more students in environments under severe financial constraints.

For example, a CTO fellow selected for AWS’s Now Go Build programme through NextGenU produced culturally contextualised textbooks at 1/100 of traditional costs and expanded lessons from 12 to 605 within 18 months. That scale would traditionally require a team of educators working for years, and it was impossible 5 years ago.

After 2026, personalised AI tutoring will become as ubiquitous as smartphones. Every student will be able to receive guidance tailored to their learning style, pace, language and needs. Education is fundamentally a human-centred system. There are conditions under which people can thrive and conditions under which they cannot. Ken Robinson compared education to rain falling in Death Valley. Death Valley, the hottest and driest region in the United States, looked like dead land where nothing grew. But after a single rainfall in 2004, the entire valley was covered with flowers in the spring of 2005. Death Valley was not dead. It was simply dormant, waiting for the right conditions.

When we use tools that spark curiosity instead of forcing compliance, and respect diversity instead of demanding uniformity, schools come back to life. And that changes everything.