How the Unprompted Deep Research Challenge Works

What “deep research” means in this challenge

In Unprompted, Deep research refers to a practice sequence. The site draws an unfamiliar concept and runs the timers. You choose the search tools, open the sources, compare the evidence, decide what to write down, and explain the result. The page does not browse, summarize, cite, or generate an answer for you.

“Deep” describes the quality of the questions you ask within the available time. A useful round separates the definition from the evidence, checks more than one source, notices disagreement, and states the limits of the conclusion. A ten-minute exercise remains an introduction. Its value comes from making the path from source to explanation visible.

This separation matters because collecting information can feel like learning. Search results, open tabs, and copied notes keep the answer visible. Closing the material creates a different test: can you rebuild the idea, connect the evidence, and give a concrete example without borrowing the source’s wording?

Turn the random topic into a researchable question

A topic label is a starting point, not yet a research plan. Before searching, translate it into three small questions. This prevents a broad query from producing a pile of unrelated facts.

Meaning

What does the term mean, and how is it distinguished from nearby ideas?

Mechanism or evidence

What process, data, or argument supports the central claim?

Boundary

When does the claim fail, change, or remain disputed?

For “network effects,” you might ask what counts as a network effect, how a larger user base changes value, and when congestion or multi-homing weakens the advantage. For “circadian rhythms,” ask what the rhythm regulates, what evidence connects light exposure to timing, and why individuals or studies may differ.

Write the three questions before opening another tab. They give you a stopping rule: when you can answer each one with an appropriate level of confidence, you have enough material for a first explanation.

Use a source ladder instead of collecting random links

Source quality depends on the question. An official definition may be strong for a policy term and weak for judging whether the policy works. A research paper may provide evidence and still need a clearer explainer to decode its method. Move through a small ladder and give each source a job.

1. Orient with a clear overview

   Find a university, museum, professional body, major reference work, or careful explainer that defines the topic in plain language. Use it to identify the standard term, related vocabulary, and the question that needs stronger evidence.

2. Move toward primary or authoritative evidence

   Depending on the subject, this could be a peer-reviewed study, official statistics, legislation, a standards document, a historical primary source, or a report that describes its data and method. Check what was actually measured before repeating the conclusion.

3. Add a credible counterweight

   Look for a review, replication, opposing analysis, known limitation, or later evidence. The purpose is not to manufacture a two-sided debate. It is to learn which part of the claim is stable and which part depends on definitions, samples, assumptions, or time periods.

Stop opening sources when new pages repeat the same facts without changing your answer. Under a short timer, two well-chosen sources and one limitation usually produce a better explanation than ten lightly scanned tabs.

Take notes you can turn into an explanation

Copying sentences preserves the source’s language and often hides whether you understand it. Write compact notes that record relationships. A useful note sheet can fit in four rows.

Keep source names or URLs beside the notes so you can verify them after the spoken round. If you record a number, include its unit, population, period, and comparison. “Risk increased” is hard to evaluate; “the study reported a higher rate in this sample over this period” preserves enough context to avoid turning a narrow result into a universal claim.

When sources disagree, write the disagreement itself. Ask whether they use different definitions, data, methods, populations, or dates. You may not resolve the conflict before the timer ends. A careful explanation can say what each source supports and what remains uncertain.

Choose a timer that fits the research question

Five minutes: orientation

Use a five-minute round to define a narrow term, locate one authoritative source, and produce a simple example. Treat the result as a map for later reading. Avoid strong claims about causation, effectiveness, or historical controversy when the evidence cannot be checked in the time available.

Ten minutes: the standard challenge

Ten minutes can support a definition, two-source comparison, example, and limitation for many well-bounded concepts. Spend roughly two minutes framing the question, five minutes reading, and three minutes selecting key points from your notes and checking the central claim.

Twenty minutes or more: context and disagreement

Add time when the topic needs historical sequence, policy details, multiple mechanisms, or contested evidence. A longer timer should change the depth of verification, not merely the number of links. Decide what the extra ten minutes will answer before extending the round.

Close the notes and explain what you learned

When research ends, stop editing. Close or minimize the sources, put the note sheet out of view, and take a short pause before starting the speaking timer. The ready state exists so the transition remains deliberate. You control when the explanation begins.

Start with the answer to the meaning question. Then connect evidence to the claim in plain language. Give the example and explain why it fits. End with the boundary: a limitation, uncertainty, tradeoff, or question that the round did not settle. If you forget a detail, state the level of confidence you do have instead of inventing precision.

A one-minute explanation needs a narrow claim. A three-minute explanation can add evidence and a counterpoint. Longer speaking time is useful only when the research produced a clear structure. Repeating the same definition in several forms does not add depth.

Worked example: Simpson’s paradox

Suppose the draw gives you “Simpson’s paradox.” Your meaning question asks what the paradox describes. Your evidence question asks how grouped and aggregated data can point in different directions. Your boundary question asks when separating groups clarifies a relationship and when it introduces a misleading comparison.

After speaking, reopen the sources. Check whether your example preserves the direction of the comparison, whether “paradox” was explained without implying a mathematical contradiction, and whether the caution about subgroup analysis matches the evidence you read.

Evaluate understanding without pretending to score it

Unprompted does not grade the response. You can still make the review concrete by comparing the explanation with the four-row note sheet and the sources. Look for errors that change meaning, omissions that break the logic, and wording that makes a limited result sound universal.

• Did the first two sentences define the topic and central claim?
• Could you explain why the evidence supports that claim?
• Did the example preserve the important conditions?
• Did you distinguish observed facts, the source’s interpretation, and your own inference?
• Did you name a meaningful limit instead of adding a generic disclaimer?
• Which sentence should be corrected after checking the source?

Finish by writing one correction and one next question. The correction protects accuracy. The next question gives you somewhere to take an interesting topic if you choose a longer research session later. Then return to the tool and draw again.