How to use AI as a thinking partner, not just a task shortcut

Executive overview

Most people underuse AI because they treat it like a 2022 assistant — giving it narrow problems and simple prompts. The capability gap between how people use AI and what it can actually do is enormous.

The shift required is treating AI as a collaborator: bringing it your messy, open-ended problems with rich context rather than asking for pre-decided solutions. This is increasingly a social skill, not a technical one.

The top 1% of AI users give AI more context, raise their ambition continuously, and engage with it to get smarter — not just to get tasks done faster.

The AI-native mindset

• People who grew up with AI don't underestimate its current capabilities
• Veterans of 2022-era AI tools often still treat the model as a basic assistant
• Those with no prior AI exposure sometimes have an advantage — they see the tool for what it is today
• Constantly raise your ambition: the problem type you gave AI last year is below what it can handle now
• Treat AI like a colleague who doubles in capability every month — not one frozen at last month's level

Context is the biggest differentiator

• The single largest gap between strong and weak AI users is how much context they provide
• Before asking a question, load in: relevant documents, company context, prior thinking, stream-of-consciousness notes
• AI can synthesise large amounts of context well; it cannot reason well from a vacuum
• Come with the problem, not the solution — open-ended problems unlock far better responses

How you engage determines what you learn

• An Anthropic study split students into AI-allowed and AI-not-allowed groups for a coding assignment
• In a subsequent AI-free assessment, the no-AI group scored 17% higher on conceptual understanding
• However, students who used AI in an inquiry mode — probing and questioning rather than extracting answers — performed just as well
• The lesson: transactional use atrophies skills; inquiry use builds them
• Define your goal before opening the tool: fast task completion vs. getting smarter while doing the task

Treating AI collaboration as a social skill

• The prompting era is over; the collaboration era has started
• Effective AI use now resembles the skill of working with a senior colleague: understanding their strengths, communicating clearly, giving useful feedback
• Practice and repetitions matter — spend time with AI the way you would building any professional relationship
• Allocate a fraction of your time to experimentation that may temporarily slow you down; it builds long-term fluency
• The future involves an inversion: AI handles high-level strategic thinking and delegates human-taste decisions to you

AI in education: risks and the right model

• Students predominantly use AI transactionally — asking for homework answers — which is the low-value, skill-atrophying mode
• One-on-one tutoring is one of the most effective learning formats; AI can scale this, but only if used in inquiry mode
• The interface for learning AI won't just be chatbots — richer, more contextual formats are emerging
• Teachers in Drew's WhatsApp network build custom flashcard apps, formative assessments, and lesson tools with Claude daily
• The goal for 2030: AI that knows your curriculum, your learning history, and your progress — a real learning companion, not a generic chatbot

The human element remains central

• AI tutoring scales access; human tutors provide accountability and care
• Schools exist not just to teach concepts but to build human-to-human skills — AI should strengthen those connections, not replace them
• Peer-to-peer tutoring platforms like Schoolhouse show that global learning communities surface shared curiosity across cultures
• The ideal 2030 classroom: AI operating invisibly behind the scenes, freeing teachers to build richer human environments