Meta-Research

When you enter an unfamiliar field, it is tempting to ask an agent for the answer and begin working from whatever it says first.

That starts one level too low.

Before solving the task, you can ask the agent to discover how excellent work in that field is already done: the standards experts use, the best examples, the tools and methods that already exist, the disagreements that matter, and the failures beginners repeatedly make.

Then you can fit that knowledge to your own goal and learn it by making something real.

Map the field → choose what fits → practice on real work → evaluate → keep what you learned.

This is meta-research: research that helps you acquire the methods of a field, not merely facts about a topic. With an agent, you can run this loop whenever a new kind of work appears—writing a grant, conducting customer research, producing an event, analyzing a market, designing a curriculum, or building software.

It will not give you years of expert judgment instantly. It can give you a strong starting map, protect you from predictable beginner mistakes, and help you practice against a much better quality bar.

Use the prompts below in the same conversation so each step can build on the research before it.

Map the best of the field

Begin with a real task or capability you want to learn. This prompt will find the standards, exemplars, tools, practitioners, disagreements, and failure modes that define good work before recommending what you should do.

Prompt

I want to become capable at this kind of work:

[DESCRIBE THE TOPIC, TASK, OR CAPABILITY]

Before advising me how to do it, map what excellent work in this field currently looks like. Research current primary sources and working artifacts online. Do not rely on a generic overview or the first popular framework you find.

Establish:

  1. 1. The quality bar: how strong practitioners recognize good work.
  2. 2. The canonical standards, methods, or bodies of knowledge that still matter.
  3. 3. Three excellent real examples I can inspect, with direct links.
  4. 4. The practitioners or organizations whose work is unusually useful to learn from—and the evidence for choosing them.
  5. 5. The best existing tools, templates, checklists, courses, open-source projects, or reference implementations.
  6. 6. The major live disagreements or schools of thought.
  7. 7. The mistakes and failure modes that repeatedly catch intelligent beginners.
  8. 8. Which parts are durable practice and which are changing quickly.

Prefer primary sources, inspectable artifacts, and practitioners who show their work. Separate broad consensus from one person's opinion. End with a compact field map and the five sources or artifacts I should actually study first.

Fit the field to your goal

The best-known method is not automatically the best method for you. Your constraints, desired outcome, existing skills, appetite for maintenance, and definition of quality determine what is appropriate.

This prompt will use the field map and your actual context to recommend a path without pretending that every best practice belongs in your life.

Prompt

Now translate the field map into a recommendation for me.

Inspect the goals, projects, files, prior work, tools, and constraints you can access in my Personal OS and relevant past conversations. Infer what I am trying to accomplish and state that interpretation so I can correct it later without pausing the analysis now.

Compare the three strongest approaches for my situation. For each, show:

  • - the outcome it is optimized for;
  • - the evidence that it works;
  • - what knowledge, time, money, access, or maintenance it requires;
  • - where it conflicts with my actual constraints;
  • - and what I would give up by choosing it.

Recommend one approach and the smallest useful slice of expert practice I should adopt first. Also name the advanced practices I should deliberately ignore for now. Keep the quality bar high while removing complexity that does not serve my goal.

Learn it by doing one real rep

Reading creates recognition; a rep creates capability. The next move is to use the chosen method on a real piece of work, compare the result with an exemplar, and get feedback while the experience is still fresh.

This prompt will turn the research into a small practice project with an artifact, a quality bar, and a feedback loop—and begin the first step.

Prompt

Turn the recommended approach into one real learning rep using work I already care about.

Design the smallest project that will produce a useful artifact while forcing me to practice the field's most important method. It should be finishable soon enough to receive feedback, but real enough that quality matters.

Give me:

  1. 1. The artifact we will produce and who it is for.
  2. 2. One excellent exemplar to keep open while working.
  3. 3. The three to five criteria we will use to evaluate the result.
  4. 4. A sequence of short stages, with the specific expert move practiced in each.
  5. 5. What the agent should do, what I should do, and where my judgment should remain in the loop.
  6. 6. A feedback source: a person, test, rubric, real user, or observable outcome.
  7. 7. A short reflection that will turn what worked into a reusable checklist, template, skill, or project instruction.

Then begin the first stage using the context you already have. Ask me only for information or permission that you genuinely cannot obtain safely yourself.