Hugo Tunius (software developer) has introduced a new workflow using Claude as a personal tutor to replace mindless social media browsing. By configuring Claude Projects with specific custom instructions, users can initiate a 'Teach me something' session that utilizes the Socratic method, a technique involving asking questions to guide discovery rather than direct explanation. The prompt instructs Claude to assess the user's existing knowledge in fields including programming, computer science, UX/UI, cybersecurity, machine learning, cooking, physics, economics, psychology, engineering, and music theory.

The system is designed to consult previous chats to ensure topic diversity and avoid repetition. When the user requests a session, Claude identifies a topic, gauges the user's baseline understanding through inquiry, and guides them toward insights via dialogue. The process aims to build upon user knowledge while filling gaps through personal reasoning. Sessions conclude with Claude recommending primary sources, such as websites, papers, podcasts, or books, to help verify information and further exploration.

Tunius reports that this approach avoids the 'information-dumpy' nature of standard LLM responses, noting that Claude successfully skips basics when the user demonstrates prior knowledge. Specific examples of topics learned include the Allais Paradox, the physics of consonance, and the chemistry of salt in cooking. One limitation identified is that Claude cannot rename chats via tool calls; users must manually rename them based on the model’s suggestions to effectively track progress across the project.

Hugo Tunius (software developer) has introduced a new workflow using Claude as a personal tutor to replace mindless social media browsing. By configuring Claude Projects with specific custom instructions, users can initiate a 'Teach me something' session that utilizes the Socratic method, a technique involving asking questions to guide discovery rather than direct explanation. The prompt instructs Claude to assess the user's existing knowledge in fields including programming, computer science, UX/UI, cybersecurity, machine learning, cooking, physics, economics, psychology, engineering, and music theory.

The system is designed to consult previous chats to ensure topic diversity and avoid repetition. When the user requests a session, Claude identifies a topic, gauges the user's baseline understanding through inquiry, and guides them toward insights via dialogue. The process aims to build upon user knowledge while filling gaps through personal reasoning. Sessions conclude with Claude recommending primary sources, such as websites, papers, podcasts, or books, to help verify information and further exploration.

Tunius reports that this approach avoids the 'information-dumpy' nature of standard LLM responses, noting that Claude successfully skips basics when the user demonstrates prior knowledge. Specific examples of topics learned include the Allais Paradox, the physics of consonance, and the chemistry of salt in cooking. One limitation identified is that Claude cannot rename chats via tool calls; users must manually rename them based on the model’s suggestions to effectively track progress across the project.