AI Thinking System
The AI Questions Canon.
A living map of the questions that define artificial intelligence — from beginner foundations to frontier models, jobs, safety, geopolitics, energy, consciousness, and civilisation-scale risk.
AI is too large to understand through definitions alone. The better way is to learn the questions: what is changing, what is bottlenecked, what is misunderstood, what could break, what could become valuable, and what humanity still does not know.
How to use this page
Five lenses, one canon.
Most AI pages give answers. This page teaches you which questions are worth asking. Pick a lens that matches where you are — vocabulary, building, research, strategy, or frontier uncertainty.
Daily question
One beginner. One frontier. One civilisation-scale.
Three questions seeded from today's date. Refresh tomorrow for a new set.
The map of AI questions
Twenty-five domains. One stack.
From foundations and chips to safety and philosophy. Tap a domain to filter the question bank below.
Question bank
Browse, flashcard, quiz, or follow a path.
Filter by difficulty, domain, type or flag. Toggle on mind-bending, technical, quiz-eligible or flashcard-only questions. Click any card to expand the full answer.
Questions that change how you see AI
Twenty-five reframings.
These do not have right answers. They reset your priors about value, jobs, trust, persuasion, taste, accountability, and the shape of civilisation under cheap intelligence.
Model selection
Which model should I use for this application?
A practical framework, not a leaderboard. Twenty workloads — for each, a default model class, when to reach for a frontier closed model, when to use open-weight, when to use a small specialist, how to evaluate, and the failure mode to watch.
Model choice changes quickly. The durable skill is learning how to evaluate models for a workload.
Architecture comparison
Transformers, Mamba, MoE, RAG, agents — which idea matters where?
Strengths, limits and use-when for the patterns most relevant to building AI systems today. Treat each row as a tool, not a religion.
Geopolitics · the AI race
Is the AI race rational?
Labour · jobs · education
Jobs: replacement, augmentation, or labour-market shock?
Sources & methodology
Where this canon draws from.
Foundational papers, AI Index trends, energy and chip sources, governance frameworks, and operator playbooks.
AI changes quickly. This page is a thinking system, not a static leaderboard.
The point of the page.
AI is not one question. It is a stack of questions. Some are technical, some economic, some political, some philosophical, and some still unanswered. The goal is not to memorise every answer. The goal is to build the judgment to ask better questions than everyone else.