English

Microsoft has made a new 3D-generation system available as open source. This means that researchers, developers, and studios can access the model, its code, and its training information. They can run it locally and adapt it for different projects. The system can create three-dimensional objects from images and export them for use in programs such as Blender and Unreal Engine. The system is also very fast. Microsoft reports that it can create a 5123 3D object (a highly detailed 3D model made using a grid of 512 × 512 × 512 small points, called voxels)  in about three seconds when using a powerful NVIDIA H100 computer. Larger and more detailed objects may take between 17 and 60 seconds. However, this technology requires expensive equipment, Linux, and an NVIDIA graphics card with at least 24 GB of memory. For therapists, 3D generation may provide new ways to create visual and interactive materials. An occupational therapist, for example, could create a model of a chair, kitchen tool, wheelchair, or bathroom. This model could help patients understand safety issues, practise daily activities, or discuss changes to their home environment. Three-dimensional models may be especially helpful for patients who learn better through pictures and objects than through spoken explanations. Therapists could use them to demonstrate how to hold an object, move through a room, or use an assistive device. This may support children and people with communication, cognitive, or learning difficulties. The technology could also support education and research. Students and professionals might create virtual homes, workplaces, or community spaces to study accessibility, balance, mobility, attention, and problem-solving. However, a realistic-looking model is not always accurate. AI may create objects with incorrect sizes, shapes, or details, so therapists must check every model before using it. Open-source access may allow researchers to study the system more carefully and explain how it was used. Nevertheless, open source does not mean that the technology is automatically safe or clinically tested. Results depend on the image, instructions, and editing process. Privacy is also important when using patient photographs, home images, or body scans. Therapists remain responsible for all professional decisions. AI should support clinical judgment, not replace it. Professionals should record how each model was created, explain AI use when appropriate, protect patient information, and make sure the final material is understandable and suitable for the patient. Microsoft’s 3D-generation system may become useful in therapy, education, and research. Its value lies in helping professionals create personalised and interactive materials more easily. Safe use will require careful checking, privacy protection, clear documentation, and further research into its benefits and limitations.

Artificial intelligence can now create very realistic images, videos, and written text. In the past, people looked for obvious signs, such as strange hands, unusual faces, spelling mistakes, or robotic writing. These signs may still appear, but they are becoming less reliable. AI systems are improving quickly, and people can also edit AI-generated content before sharing it. Many people believe they are good at identifying AI content. However, research shows that this is often difficult. People may perform only slightly better than chance when deciding whether an image, video, or piece of writing was created by AI. This means that therapists, teachers, and researchers may feel confident about their decision and still be wrong. Important decisions should therefore never be based only on personal impressions. It is useful to understand that unusual content does not always mean that it was made by AI. Writing may sound different because a person is using English as a second language, translating their ideas, or using a formal style. An image may look strange because of poor lighting, low quality, or heavy editing. These possibilities should be considered before making an assumption. Therapists and researchers should be aware of two main risks. The first is trusting AI content too quickly because it looks professional or sounds confident. The second is becoming suspicious of genuine material simply because it looks polished or unusual. A false accusation may harm a client, student, colleague, or research participant, especially when it affects treatment, education, employment, or professional reputation. A safer approach is to ask where the material came from and how it was created. When possible, professionals can review the original file, the date it was produced, and any records showing how it was edited. It may also help to compare the material with clinical notes, previous work, direct observations, or other reliable information. No single clue can prove that something is real or AI-generated. AI detection tools can sometimes be useful, but they should not be treated as final judges. These tools can incorrectly label human writing as AI-generated, especially when the writing is formal, simple, translated, or written by someone who uses English as an additional language. They can also fail to identify AI writing that has been changed by a person. A detector’s result should be treated as a reason to ask further questions, not as proof. This issue is especially important in therapy. A client may use an AI-generated image, story, or character to communicate feelings that are difficult to express directly. The material may not describe a real event, but it may still represent a real emotional experience. However, if the material is being used to document abuse, injury, safeguarding concerns, or a legal matter, the therapist should seek additional evidence and follow professional procedures. AI can also produce clinical notes, summaries, referral letters, and research drafts. Although the writing may be fluent, the information may not be correct. AI can leave out important details, change the meaning of events, or present uncertain information as fact. Professionals should check important information against original records, assessment results, and their own observations before using it. The most important skill is not becoming perfect at spotting AI. Instead, therapists and researchers need to learn how to manage uncertainty. They should consider the purpose of the material, the possible consequences of being wrong, and the quality of the available evidence. Professionals should be transparent about their use of AI tools and remain responsible for their decisions. Careful questioning, respect, and professional judgment are more reliable than quick conclusions.

Scientific research often requires many different tools. Researchers may need to search medical databases, analyse data, write computer code and use powerful computers. Managing these tasks can be difficult and time-consuming. Anthropic’s Claude Science is a new AI workbench designed to bring many of these activities together in one place. Claude Science can help scientists and research teams search the literature, organise information, analyse data and create scientific figures. It supports areas such as genomics, proteomics, neuroscience, drug research and structural biology. In simple terms, it works like a digital research assistant that can support several stages of a scientific project. A key feature is that Claude Science can save the code and computer environment used to create its results. For example, when it produces a graph, it can also save the instructions behind that graph. This allows researchers to check the work and repeat the analysis later. Such transparency is important because scientific findings should be understandable and reproducible. The system can also display complex information, including three-dimensional protein structures, genome data and chemical structures. These visual tools may help researchers understand difficult material. However, attractive figures should not automatically be treated as reliable evidence. Researchers must still check the data, methods and conclusions. Claude Science can also help manage computer resources. It may prepare an analysis plan and send tasks to a laboratory’s computer system or to online GPU services. This could help research teams complete complex analyses more efficiently. However, faster computer processing does not guarantee that the research question, data or statistical method is correct. Early users have reported time savings in drug design, neuroscience and glioma genetics research. These reports are promising, but they mainly come from early users and the company itself. Independent studies are still needed to determine how often the system makes mistakes and whether it improves the quality of scientific work. For clinicians and therapists, Claude Science could help locate research articles, summarise evidence, organise datasets and prepare early drafts of reports. It should support professional reasoning, not replace it. Clinicians must still decide whether the information is accurate, relevant and appropriate for a specific patient or clinical setting. The platform includes a separate reviewer agent that checks citations, calculations and possible errors. This may be useful, but it cannot replace experienced human review. AI may identify a missing reference while failing to recognise problems such as biased sampling or an incorrect interpretation of cause and effect. Privacy is another important concern. Claude Science is designed to keep sensitive research data within the user’s own systems or institutional infrastructure. Even so, organisations must follow privacy laws, research ethics rules and local policies. Confidential patient information should never be entered into an AI system without proper approval and protection. AI systems can produce incorrect information, repeat biases and make uncertain findings appear more definite. Users may also trust a well-written answer without checking the original evidence. Clinicians and researchers therefore need to question AI outputs, verify sources and understand the system’s limitations. Claude Science may help researchers spend less time managing technical tools and more time focusing on important questions. Its value, however, will depend on careful testing, responsible use and continued human oversight. It should be viewed as an assistant, not as an expert making final decisions. Professional judgment, ethical awareness and attention to evidence must remain central.

In inclusive classrooms, teachers and therapists carefully consider how to support each student. Some students need simpler language, more time or help organising information. However, support should also give students opportunities to ask questions, make choices and take an active role in learning. Generative artificial intelligence, or GenAI, is a technology that can create text and respond to questions. ChatGPT is one example. It may help students understand difficult ideas, find information and explore topics that interest them. However, its value depends not only on the answers it provides, but also on whether it helps students become more involved in their learning. A recent study published in the British Journal of Educational Technology by Rappa et al. (2026), titled “Can Generative AI Support the Learning Agency of Students with Disability? A Case Study of an Australian Secondary School,” explored this issue in a secondary science classroom. Three students with disability used ChatGPT while learning about gravity, friction, car brakes and motion. The researchers examined how students used the tool, what choices they made and what challenges they experienced. The study focused on learning agency. In simple terms, this means having the opportunity and ability to take part in decisions about learning. A student may choose to ask a question or request a simpler explanation, but may still need support with communication, attention, memory or planning. The students used ChatGPT in different ways. Arthur used it to learn more about gravity because he was already interested in science. Inez connected science ideas with real-life concerns, especially road safety. Louis asked science questions but also treated ChatGPT as a conversation partner. These examples show that students may use AI to make learning more meaningful and personal. However, choosing to use ChatGPT did not always mean that students could use it independently. Some found it difficult to write clear questions, understand long answers or decide whether information was correct. Teachers therefore modelled how to ask questions, request simpler explanations and ask for examples. This kind of support is similar to the scaffolding used in therapy and education. The findings suggest that GenAI may support students’ participation, but it does not replace human support. Teachers, therapists and education assistants may help students create questions, understand responses and connect AI information with classroom activities. The goal should be to build independence gradually, rather than simply giving students access to technology. There are also important limitations. Only three students from one Australian school were included, so the findings cannot represent all students with disability. More research is needed with larger groups, different subjects and students with more significant communication or learning needs. Ethically, professionals must consider privacy, accuracy and bias. Students should not enter personal information into AI systems. ChatGPT can provide incorrect or confusing answers, so adults must check its responses carefully. Students should also learn that AI can be useful, but it can make mistakes. In the end, GenAI does not create learning independence by itself. It may support learning agency when it helps students ask questions, explore interests and make choices about learning. For therapists and educators, AI should be used as an additional support, not as a replacement for professional judgement, relationships or communication. Reference Rappa, N. A., Nonis, K. P., Tang, K.-S., Cooper, G., Cooper, M., & Sims, C. (2026). Can generative AI support the learning agency of students with disability? A case study of an Australian secondary school. British Journal of Educational Technology. https://doi.org/10.1111/bjet.70048

In clinical work, we often think before we speak. A therapist may listen, consider a few possibilities, and only then respond. Much of this thinking remains invisible. Researchers at Anthropic recently described something similar in their AI system, Claude: an internal space called “J-space,” where the model briefly holds and works with ideas before producing an answer. J-space can be understood as a small mental workspace. It is not physical, but a pattern where important concepts are gathered and used. This idea is similar to what psychologists call a “global workspace,” where selected information becomes available for reasoning. In both humans and AI, thinking depends not just on knowledge, but on how that knowledge is organized in the moment. A simple example helps clarify this. Claude was asked about “the animal that spins webs.” Internally, the concept “spider” became active, leading to the correct answer: eight legs. When researchers replaced that internal concept with “ant,” the answer changed to six. This suggests that the AI forms an internal idea first, which then guides its response. For therapists, this process feels familiar. We often hold a working hypothesis in mind before speaking. These internal steps are essential, even if they are not visible. Importantly, when researchers reduced the influence of J-space, Claude still produced fluent language, but its reasoning became less accurate. This reminds us that clear or confident speech does not always reflect sound thinking. At the same time, we should be careful with this comparison. AI systems do not understand in the human sense. Their internal patterns come from data, not lived experience. While the idea of a “workspace” is helpful, it is still a simplified description of a complex system. For research, J-space offers a new way to study how AI reaches its answers. It may help identify errors, biases, or hidden signals, such as when a model detects misleading input. In clinical contexts, this could support decision-making, especially in complex cases—but it does not replace professional judgment. There are also important limits. These findings are early and mostly tested in specific models. We do not yet know how consistent or stable this feature is. Future systems may also become harder to interpret, not easier. Ethically, this raises questions about trust and responsibility. If clinicians use AI tools, they remain accountable for decisions. Internal signals like those in J-space should not be treated as fully reliable evidence. Transparency must include recognizing uncertainty and avoiding overconfidence in systems we do not fully understand. In the end, J-space does not prove that AI “thinks” like humans, but it highlights something important: reasoning often depends on hidden steps. For clinicians, this is already familiar. What is new is the chance to observe a similar process in machines, which may deepen how we reflect on both human and artificial thinking.

In therapy, much of our work happens between knowledge and presence. We are not only applying what we know, we are listening, observing, and adjusting in real time. Often, the challenge is not a lack of knowledge, but making sense of it in a way that fits the person in front of us. As MedGPT expands across Europe, many therapists are beginning to wonder how a tool designed for medical use might support this process. MedGPT first gained attention in France, where it reached more than 200,000 users in a short time. It is now being introduced in several European countries, including Germany, Italy, Spain, Portugal, Belgium, the United Kingdom, Ireland, the Netherlands, Poland, and Switzerland. The rollout starts with smaller groups of users in each country, allowing the system to adapt to different healthcare systems, languages, and clinical cultures. This gradual approach is important, because clinical practice is never one-size-fits-all. Although MedGPT was developed for physicians, it can still be meaningful for therapists. Our work often involves connecting physical, psychological, and social aspects of a person’s experience. We move between different models and perspectives, sometimes within a single session. In this context, a tool that helps organize and revisit these layers can support our thinking without replacing it. When we explored MedGPT, what stood out was not that it provided perfect answers, but that it helped structure reflection. For example, when considering a patient with chronic pain and fear of movement, the system brought together different contributing factors, such as physical sensitization, beliefs about pain, and avoidance behaviors, in a clear and organized way. It did not make decisions for us, but it helped us slow down and see connections more easily. In everyday practice, this kind of support can take simple forms. MedGPT can help rephrase complex information into clearer language for patients, summarize research when time is limited, or assist in organizing clinical notes. It can also support preparation for communication with other healthcare professionals. Used in this way, it becomes less of an authority and more of a thinking partner. At the same time, therapy remains deeply human. Our decisions are guided not only by knowledge, but by experience, intuition, and the relationship we build with each patient. No system can fully capture these elements. This makes it important to use AI carefully, protecting patient data, staying aware of possible biases, and remembering that responsibility for clinical decisions always remains with the therapist. There is also a broader question of how these tools influence our thinking over time. If used without reflection, they may lead us to rely too quickly on structured answers. But if used thoughtfully, they can help us become more aware of our reasoning, more curious, and more precise in our work. In the end, MedGPT’s expansion across Europe is not only about technology, but about how clinicians choose to engage with it. For therapists, it offers a way to support clinical thinking while staying grounded in the human side of care. Its value may lie less in the answers it gives, and more in how it helps us reflect, organize, and approach our work with greater clarity.

Many clinicians and educators are now expected to create video content, whether for patient education or teaching. For many, editing has always been a barrier—time-consuming, technical, and sometimes overwhelming. New tools built around systems like Claude are starting to make this process feel more manageable. Recent updates have made this even more noticeable. Claude’s AI agents can now take on larger parts of the editing process, helping structure content and refine videos with less manual effort. For those already using tools like ChatGPT Plus/Pro or Claude Pro, this opens up more advanced and practical ways to work with video. Another shift comes from tools like Palmier, designed specifically for AI-driven editing. With a single prompt, users can guide the entire video process, trimming clips, reorganizing scenes, adding B-roll, and working directly within a timeline. It feels less like traditional editing and more like guiding the process through conversation. This changes the workflow in a meaningful way. Instead of focusing on technical steps, you describe what you want to communicate. The system helps shape the structure, visuals, and pacing, and you can refine it through feedback. The process becomes more iterative and less dependent on technical expertise. From a cognitive perspective, this reduces the load of managing multiple small tasks at once. The detailed mechanics of editing move into the background, allowing more attention to stay on the message itself. For clinicians and researchers, this can make content creation feel more approachable. In practice, this can be especially helpful. A therapist creating a psychoeducational video can focus on explaining a concept clearly, while the system supports how it is presented. This can ease time pressure and make it easier to produce consistent, high-quality content. At the same time, these tools don’t replace professional judgment. Decisions about what to include, how to frame information, and what fits a specific audience still rely on human expertise. AI can assist with execution, but it doesn’t fully understand clinical nuance. There is also a tendency to trust polished outputs too quickly. Even when something looks complete, it still needs careful review. In clinical and educational contexts, small details matter, and accuracy remains essential. For teams, these tools can support faster collaboration and content development. Ideas can move more quickly from concept to final product. Still, speed shouldn’t come at the expense of reflection or quality. Limitations are also worth keeping in mind. AI systems are shaped by existing data, which can include gaps or biases. This means outputs may not always reflect diverse perspectives or fully accurate information, making review and adjustment necessary. Ethically, transparency and responsibility remain key. Understanding how these tools contribute to the final output helps maintain trust and accountability, especially in clinical contexts. Overall, tools like Claude are making video creation more accessible and less technical. They open new possibilities for communication, while still relying on thoughtful use and professional oversight. Their real value lies in supporting clearer, more effective ways to share knowledge—not replacing the expertise behind it.

In a recent supervision session, a therapist shared her experience using AI to write notes and summaries. The output was quick and well-written. But when asked if it truly made her work feel easier, she paused. “It saves time,” she said, “but I still have to think just as much.” That hesitation reflects something many clinicians are starting to notice. We often hear that AI will dramatically improve productivity. Some even suggest it could make work ten times faster. Yet in everyday clinical settings, the change feels more subtle. Speed may improve, but the depth of the work, understanding, deciding, staying present, remains just as demanding. Part of this comes down to how we think about productivity. In clinical practice, it goes beyond getting things done quickly. It involves making sense of complex situations and applying thoughtful judgment. AI can support parts of the process, but it doesn’t replace the thinking behind it. Sometimes, it even creates a sense of rushing through decisions rather than sitting with them. Engagement also plays a role. Gallup (a global analytics and advisory firm specializing in workplace research; State of the Global Workplace Report, 2023) reports that only 20% of employees worldwide feel engaged in their work. When many people are already feeling disconnected, introducing faster tools doesn’t necessarily improve the experience. It may simply make it easier to move through tasks without feeling more connected to them. Motivation matters here. People tend to engage more when they feel a sense of purpose and ownership in what they do. When AI is introduced as a support for thinking, it can feel helpful. When it feels like another demand or shortcut, it can have the opposite effect. Another important shift is in the type of effort required. Writing may take less time, but reviewing, checking, and adjusting often take more attention. The work doesn’t disappear, it changes shape. The cognitive load is still present, just redistributed. The surrounding environment also makes a difference. When teams have space to reflect, ask questions, and learn how to use AI thoughtfully, the benefits are clearer. Without that support, AI can feel like an extra layer rather than a helpful tool. There’s also the question of what happens with the time saved. In many settings, it quickly fills up with more tasks. This can leave people feeling just as rushed as before. A more intentional approach might use some of that time for reflection, learning, or improving care. Ethical considerations remain central. AI can produce outputs that sound confident but may contain inaccuracies or bias. This makes careful review essential. Responsibility for decisions and content always stays with the clinician. Overall, the impact of AI depends less on the tool itself and more on how it’s used. When approached thoughtfully, it can support clearer thinking and better workflows. The goal isn’t just to move faster, but to work with more clarity, care, and intention. Gallup. (2023). State of the global workplace: 2023 report. Gallup, Inc.

Something subtle is changing in training environments. Tasks that used to belong to beginners, writing first drafts, organizing ideas, suggesting interventions, can now be done quickly with AI. The output often looks polished, even impressive. But it brings up a quiet, uncomfortable question: if these early steps fade away, where does real learning take place? For a long time, learning a clinical or research role meant going through imperfect stages. You tried, got things wrong, adjusted, and tried again. It could be frustrating at times, but that was part of the process. Those early efforts weren’t just practice, they were where clinical judgment began to form. Gradually, with enough exposure, you started to recognize patterns, sit with uncertainty, and think in more flexible ways. Now, many of those early tasks can be handled by AI. A study by Dell’Acqua and colleagues (2025), involving 776 employees at Procter & Gamble, showed that one person using AI could reach results similar to an entire team working without it. If this direction continues, it’s not hard to imagine workplaces hiring fewer junior staff and instead looking for people who can do a bit of everything with AI support. This shift touches something important about how we learn. Psychological research has been clear on this for a while: learning deepens when we are actively involved, especially when we struggle a little, make mistakes, and reflect on them. When that effort disappears, understanding can become thinner. You might arrive at the right answer, but without really grasping how you got there. In clinical work, that difference matters. It’s not only about reaching a correct conclusion—it’s about the thinking behind it. Take something like writing a case formulation. It asks you to bring together complex pieces, weigh possibilities, and make careful decisions. When an answer is readily available, there’s a risk of relying on it without fully developing that internal process. At the same time, it’s not all negative. Used thoughtfully, AI can support learning. It can offer alternative perspectives, point out things we might have missed, or even act as a kind of reflective partner. The difference lies in how we use it—whether we stay engaged, question what we see, and compare it with our own thinking, or whether we simply accept what is given. This also makes assessment more complicated. If someone produces strong work with AI, it becomes harder to tell what they actually understand. Training and supervision may need to shift focus, paying more attention to how people think rather than only what they produce. There’s a broader concern as well. If fewer entry-level opportunities exist, or if early tasks are largely automated, future clinicians and researchers may have fewer chances to build their skills step by step. Over time, this could shape not just individual careers, but the depth of expertise across the field. The ethical side remains just as important. Even with AI involved, responsibility does not shift away from the clinician. It still matters to question the output, to understand its limits, and to stay aware of potential biases. Being transparent about how these tools are used is part of maintaining trust. Similar questions appear in research and training. AI can help with writing or generating ideas, but it cannot replace critical thinking. Keeping that boundary clear is essential if we want to preserve the quality and integrity of the work. In the end, AI isn’t removing the need for expertise, it’s reshaping how it grows. The challenge now is to create spaces where learning stays active, where curiosity and effort still have a place. What may matter most going forward is not just what we can produce, but how deeply we are still able to think.

Think about the last time you sat with a complex clinical case, notes spread out, different possibilities in mind, trying to make sense of it all. That process usually asks for time, patience, and a kind of quiet reflection. With tools like NotebookLM, something is shifting. The tool now helps organize what we see, brings forward connections, and sometimes even nudges our reasoning forward. It starts to feel less like a tool we use and more like something we think alongside. As these systems become more capable, they move through information quickly and offer structured, clear insights. This can be genuinely helpful, especially when the workload is heavy. But it also changes our relationship with knowledge. When answers come faster and more neatly, we may find ourselves spending less time sitting with the uncertainty that often leads to deeper understanding. Some of the newer features make this even more noticeable. The ability to run analyses, process data, and generate results in one space removes many of the usual barriers. Tasks that once required effort and multiple steps now unfold smoothly. While this opens new possibilities, it can also create a bit of distance from the “how” behind the results. The same applies when it comes to producing reports or presentations. Turning ideas into something structured has become much easier, which can support communication in meaningful ways. Still, there is a subtle question that lingers, when something comes together so quickly, where did the deeper thinking take place? Even the way we begin our work is changing. Instead of carefully gathering and selecting sources, we can start with a simple question and let the system do the rest. It saves time, but it also touches on a core clinical and research skill: learning how to choose, question, and critically engage with information. This matters for learning as well. Growth often happens when we stay involved in the process, when we test ideas, reflect, and sometimes struggle a bit. If too much of that process is handled for us, we might still reach good conclusions, but without the same depth of understanding. In practice, this becomes very real. Working with patients asks for more than accurate answers. It asks for presence, sensitivity, and the ability to navigate uncertainty with care. AI can support parts of this journey, but it cannot step into the human side of the work. There are also responsibilities that remain firmly ours. Making sense of information, questioning its accuracy, and staying aware of biases are still essential parts of the role. Being clear about how decisions are made becomes even more important when these tools are involved. In the end, this shift is less about replacement and more about how we adapt. The challenge is to stay engaged in our own thinking, to remain curious, and to use these tools in a way that supports rather than replaces our clinical voice. When we hold onto that, technology can deepen our work without taking away what makes it meaningful.