# Remember yesterday's post about the cheating that AI brings with it to academia? Machine translation from the original post. Not reviewed by a human translator. Historical claims may no longer be current. Remember yesterday's post about the cheating that AI brings with it to academia? So here is the answer: a student at the University of San Francisco School of Law uploads the recording of his latest lecture to Claude. Instead of getting a ready-made summary, the model asks him: how would you build a legal argument? What evidence supports the position you chose? What would the other side do? In other words: he doesn't get an answer - he starts to think. As part of new collaborations with the Canvas, Panopto and Wiley systems, Anthropic's language model is beginning to take a permanent place in academic learning systems. But not as a convenience tool. On the contrary - as a means of encouraging thinking. Students upload files, articles and recordings of lessons, and Claude works them through open questions, alternative phrasings, and dialogic guidance reminiscent of a personal tutor. This is a new mode called "Learning Mode" - a mechanism in which the artificial intelligence asks the questions before it answers. In parallel, users report a new feature called "Study Together" that has popped up in experimental versions of ChatGPT - an interactive study tool that tries to stop giving answers, and start practicing understanding. This direction marks a deep change: from a model that supplies knowledge - to a model that awakens consciousness. The tools that flood us with answers are starting to rephrase the terms of the question. The assumption that AI is meant to save time or streamline a process gives way to a different approach - one in which it reshapes the thinking process, not only the product. The academic context emphasizes this especially: at universities like USF or Northumbria, it's not only about using technology, but about re-examining the way of learning itself. And it's not only a question of innovation. It's a question of responsibility. If a student learns to confront arguments, examine assumptions, ask questions before answering, then perhaps precisely in the age of AI, there's a chance critical thinking will not weaken, but sharpen. In a parallel move, in July 2025 OpenAI joined an initiative of the American Federation of Teachers as part of establishing the "National Academy for AI Instruction". Under this framework, over the next five years, 400,000 teachers will be trained in workshop days and practical courses, with an emphasis on tools for technology in the classroom, ethics, and internalizing critical thinking. OpenAI will contribute $10 million to it and allocate physical spaces and network schools under professional guidance This is a structural change in the education system's conception: from a role of "transferring information" to creating leaders of digital thinking. And it's necessary and needed and also quite urgent - yesterday I published a post about the existing failure in today's education system since the AI revolution, and how everyone is trying to learn it and get around it (and succeeding too). Fortunately the companies are not staying indifferent to the change and disruption, and are really working on creating a new learning environment for us. For years there were voices that dreamed of new learning methods and raised complaints against the system. Perhaps precisely now, with the artificial intelligence revolution, it can really happen. The post on Anthropic's blog: https://lnkd.in/dYhJQvDt The post on OpenAI's blog: https://lnkd.in/dUy5TWhc Original: https://ofershap.github.io/posts/ai-tutors-that-ask-before-they-answer/ Source: https://www.linkedin.com/posts/ofershap_%D7%96%D7%95%D7%9B%D7%A8%D7%99%D7%9D-%D7%90%D7%AA-%D7%94%D7%A4%D7%95%D7%A1%D7%98-%D7%9E%D7%90%D7%AA%D7%9E%D7%95%D7%9C-%D7%A2%D7%9D-%D7%94%D7%A8%D7%9E%D7%90%D7%95%D7%AA-%D7%A9-ai-%D7%9E%D7%91%D7%99%D7%90%D7%94-activity-7350781175109419008-OEHE Published: 2025-07-15T10:00:03.056000+03:00