Prompts

An open learning resource covering prompting, context engineering, RAG, and AI agents.

A hand-curated index of prompt and context engineering resources, covering papers, tools, models, benchmarks, courses, and security material.

Builds the strongest case for both directions, compares conditions, benefits, risks, and hard objections, and asks one question most likely to change the conclusion.

Removes domain jargon to identify the problem structure, searches history and at least three distant fields for analogous mechanisms, and ends with a reversible experiment.

Reduces a problem to unavoidable facts, unverified assumptions, the real goal, and practical constraints, then derives a feasible path without relying on convention.

Selects three genuinely complementary professional perspectives, compares their paths and risks, and uses cross-examination to identify consensus, disagreement, and assumptions.

Separates a claim into verifiable facts, inferred conclusions, and value judgments, checks sources and context, and then audits the reasoning chain.

Studies how a subject developed over time, compares it with relevant peers using consistent dimensions, and connects historical capabilities and constraints to future paths.

Reverse-engineers a strong product, page, process, or finished work into its audience, structure, key decisions, definition of done, reusable principles, and a practice exercise.

A multi-turn prompt that delays advice, asks one decision-changing question at a time, separates facts from interpretations and goals, and ends with a more precise question.

A two-turn prompt that delays the final answer, asks the model to identify assumptions, missing information, and a common mistake, and then asks one key question before responding.