El Mapeador de Prerrequisitos: Descubre Exactamente Qué Te Falta Antes de Abordar un Tema Difícil
Por qué importa este prompt
Most self-directed learners respond to a stuck point by either giving up on the resource entirely or re-reading the same confusing paragraph five times — both of which waste hours, because the actual fix is usually one or two specific prerequisite concepts, not a full remedial course.
Para qué lo usamos
You're three chapters into a machine learning textbook and hit a section on backpropagation that assumes you're comfortable with the chain rule and Jacobian matrices — but it's been eight years since your last calculus class and you're not sure exactly what you're missing or where to start filling the gap.
Prompt
Role: Act as an expert curriculum designer who specializes in diagnosing prerequisite gaps for self-directed learners. Context: I'm trying to learn [TARGET TOPIC], and I'm stuck on [SPECIFIC RESOURCE, e.g. book/course/paper title]. I got lost specifically at [WHERE EXACTLY I GOT LOST — the section, term, or step]. My relevant background is: [WHAT I ALREADY KNOW RELATED TO THIS FIELD]. Task: Identify the specific prerequisite concepts I'm missing that are causing me to get stuck at this exact point, order them by dependency (what must come before what), and for each one give a one-line test I can use to check whether I actually know it well enough to move on. Constraints: Do not recommend a generic "start from the basics" full curriculum — assume I have the background I described and only flag genuine gaps. Do not pad the list with prerequisites I've indicated I already know. Be specific about WHY each prerequisite is needed for the exact sticking point I described, not just generically related to the topic. Output Format: 1. Diagnosis (what specifically is causing me to get stuck, in plain terms) 2. Prerequisite Chain (ordered list — each with: concept name, why it's needed here, a one-line self-test) 3. Fastest Path (which 1-2 prerequisites to tackle first for the most unlock, if I'm short on time) 4. Resource Type Suggestion (what KIND of resource — not a real link — would best teach each gap, e.g. "a worked-example problem set," "a visual/interactive explainer," "a primary source paper")
Resultado
DIAGNOSIS: You're not actually stuck on backpropagation itself — you're stuck because backpropagation is explained using compact calculus notation (partial derivatives via the chain rule, expressed as Jacobians) that assumes fluency you haven't used in years. The concept of "work backward through a computation graph" is intuitive; the notation is the actual blocker.
PREREQUISITE CHAIN:
- Single-variable chain rule — Why: backpropagation is the chain rule applied repeatedly. Self-test: can you differentiate f(g(x)) by hand without looking up the rule? If not, start here.
- Partial derivatives — Why: neural network weights are one of many inputs, so you need derivatives with respect to one variable while others are held constant. Self-test: given z = x²y + 3y, can you find ∂z/∂x and ∂z/∂y without hesitating?
- Basic matrix/vector notation — Why: Jacobians are just organized collections of partial derivatives — the intimidating part is the notation, not new math. Self-test: can you explain what a 3x2 matrix represents and multiply it by a vector?
- Computation graphs — Why: this is the actual new concept — how a network's forward pass can be drawn as a graph so backprop becomes "chain rule applied along graph edges." Self-test: can you draw the graph for y = (x+2) * (x-1) and label each node?
FASTEST PATH: If you're short on time, prioritize #1 (chain rule) and #4 (computation graphs) — together they cover 80% of what makes backpropagation click. Partial derivatives (#2) can often be picked up implicitly while working through computation graph examples, and matrix notation (#3) is more about getting comfortable with symbols than learning new concepts.
RESOURCE TYPE SUGGESTION: For the chain rule, a worked-example problem set beats a video — you need to practice the mechanical step, not just watch it. For computation graphs specifically, a visual/interactive explainer (something that lets you click through a graph node by node) will click faster than a textbook diagram. For partial derivatives, a short primer with 5-10 practice problems is enough — you don't need a full multivariable calculus course.
Quedarte atascado en un tema difícil rara vez significa que el material esté mal escrito. Más a menudo, un libro de texto o curso asume fluidez en algo que no has tocado en años — y lo frustrante es que no siempre puedes identificar qué pieza específica falta. Solo sabes que estás perdido.
Este prompt está diseñado para diagnosticar esa brecha directamente en lugar de enviarte de vuelta a "empezar desde lo básico", lo cual desperdicia tiempo repasando cosas que ya sabes. Pide el punto específico de atasco — no solo el tema — porque la solución para "no entiendo backpropagation" es completamente diferente según si el bloqueo es la notación, un concepto matemático faltante, o una idea genuinamente nueva.
Por qué la cadena de prerrequisitos está ordenada, no solo listada
Una lista plana de "cosas que quizás no sepas" es casi inútil — no te dice por dónde empezar ni cómo dependen las piezas entre sí. Ordenar la cadena por dependencia significa abordar los prerrequisitos en la secuencia que realmente te desbloquea.
Por qué cada prerrequisito recibe una autoevaluación de una línea
La mayoría de los estudiantes sobreestiman o subestiman drásticamente lo que recuerdan. Una autoevaluación rápida y específica reemplaza esa conjetura — si puedes responder al instante, salta ese prerrequisito y pasa al siguiente.
Por qué el "camino más rápido" es una sección separada
No todo estudiante tiene tiempo para reforzar cada brecha antes de continuar. Separar el uno o dos prerrequisitos de mayor impacto significa que alguien con tiempo limitado obtiene un atajo real, no un compromiso de rigor.
Introduce tu punto específico de atasco en este prompt, y obtendrás una solución dirigida — no un currículo remedial completo para un tema que en su mayoría ya entiendes.