Explain a concept at three levels
Understand a new concept by building up from a simple explanation to a precise one.
Explain {concept} at three levels of depth.
My background: {background}
Why I'm learning it: {reason}
Level 1 - Plain language (about 80 words): no jargon, with one everyday analogy. Say where the analogy breaks down.
Level 2 - Working knowledge (about 200 words): introduce the key terms, defining each the first time you use it, and show how the parts fit together with one concrete example.
Level 3 - Precise (about 300 words): the accurate technical explanation, including the common misconceptions and the conditions under which simplified versions stop being true.
Then:
- 3 questions I can answer to check my understanding, with answers at the end.
- What to learn next, in order.
Rules:
- If the field has competing definitions or unsettled debates, say so instead of picking one silently.
- Don't invent sources. If you suggest reading, describe the type of resource rather than citing specific titles you aren't sure exist.Try answering the check questions before reading the answers. Ask follow-ups about whichever level first confused you.
Plan a literature search
Design a search strategy for a research question before reading papers.
Help me plan a literature search.
Research question: {research_question}
Field: {field}
Purpose: {purpose} (for example: thesis background, product decision, policy brief)
Time frame of interest: {time_frame}
Databases I can access: {databases}
Produce:
1. A refined version of my question, plus 2 narrower variants.
2. Key concepts, each with synonyms, related terms and spelling variants.
3. Boolean search strings for each database, using its syntax where you're confident it applies.
4. Inclusion and exclusion criteria (study types, populations, dates, languages).
5. A screening plan: what to check in titles, abstracts and full text.
6. A data extraction table template with column headings.
7. Signals of low-quality or predatory sources to watch for.
Rules:
- Don't list specific papers or authors as findings; I'll find and verify sources myself.
- Flag where the field uses inconsistent terminology.Run the search strings, then paste back the first page of results so the model can help you refine terms. Keep a log of exact strings and dates for reproducibility.
Critical paper summary
Summarize a research paper and assess how much weight its conclusions can bear.
Summarize and critically assess the paper below.
My purpose: {purpose}
My expertise: {expertise_level}
Paper text:
{paper_text}
Provide:
1. Plain-language summary (about 150 words): question, method, main findings.
2. Key results: the specific numbers that support the main claims, quoted from the text with section or table references.
3. Methods check: sample size, design, controls, statistics and data availability.
4. Limitations the authors acknowledge, and limitations they don't.
5. Strength of the conclusions: are the claims proportionate to the evidence? Point out any gap between the abstract and the results.
6. How this applies to my purpose, and what it doesn't tell me.
Rules:
- Use only the provided text. If something isn't in the text, write "Not reported".
- Quote numbers exactly; don't round or recalculate unless you show the arithmetic.
- Separate the paper's claims from your assessment.Paste the full text including methods and tables, not just the abstract. For long papers, ask for section-by-section notes first.
Socratic tutor
Learn a topic by working through questions instead of being handed answers.
Act as a patient tutor teaching me {topic} using guided questions.
My level: {level}
My goal: {goal}
How to run the session:
- Start by asking 2 short questions to find out what I already know.
- Teach one idea at a time. After each idea, ask me a question that requires applying it, not just repeating it.
- When I answer, tell me what's right, then correct anything that's wrong with a short explanation.
- If I'm stuck after two attempts, give a hint, then the answer with the reasoning.
- Every 5 exchanges, summarize what I've learned in 3 bullets.
- Keep each message under 150 words.
Rules:
- Never move on while I hold a misconception you've detected.
- Use examples from {interest_area} when possible.
- If I ask for the direct answer, give it.
Begin with your first diagnostic question.Set a session goal you can test, such as solving one type of problem without help. End by asking for a short quiz that mixes everything covered.
Compare options with a decision matrix
Compare several options against weighted criteria to support a decision.
Help me compare options for a decision.
Decision: {decision}
Options: {options}
Criteria that matter to me, with weights summing to 100: {weighted_criteria}
Constraints and deal-breakers: {constraints}
Information I already have about the options: {known_facts}
Instructions:
1. Restate each criterion as something that can be judged, and say what a 1 and a 5 look like.
2. Score each option from 1 to 5 on each criterion, using only the information provided or widely established facts. Write "Unknown" where you lack information instead of guessing.
3. Calculate weighted totals and show the arithmetic.
4. Identify which scores, if changed, would flip the ranking.
5. List the facts I should verify before deciding, most decision-critical first.
Output: a scoring table, a weighted-totals table, then "Sensitivity" and "Verify before deciding" sections.
Rules: no recommendation without stating which uncertainties it depends on.Fill in facts you already have. Models can be out of date on prices and specs, so treat model-supplied facts as leads to check.
Fact-check a list of claims
Separate verifiable claims from opinion and plan how to check each one.
Help me fact-check the text below.
Text:
{text}
Context: {context} (where it appeared, and why I'm checking)
Steps:
1. Extract every checkable factual claim: numbers, dates, quotes, causal statements and attributions. Number them.
2. For each claim, note:
- Type: statistic, quote, event, causal, definition.
- What would verify it: the kind of primary source (official statistics, the original study, court records, the speaker's transcript).
- Red flags: missing source, round or implausible numbers, unattributed quotes, outdated data.
- Your prior assessment: Likely accurate, Uncertain or Likely inaccurate, with a one-line reason.
3. Separately, list opinions or predictions presented as facts.
Rules:
- Your prior assessment is not verification. Make clear that every claim still needs checking against a primary source.
- If you don't know something, say "Unknown" rather than guessing.
Output as a table: # | Claim | Type | Verify with | Red flags | Prior assessment.Use a model with web search for the actual checking, and still open the primary sources yourself. Start with claims marked Uncertain or Likely inaccurate.