LearnAI
Новинки AI · «Думающие» модели · Lesson 3/10
← To the catalog
Task

How to pose a deep reasoning problem

Rewrite the query so that the model reasons deeply. Return JSON {"task":"clear statement of the problem","context":"introductory and restrictions (budget, habits, what is important)","steps_request":"please reason step by step and take into account the options","output_format":"in what form the answer is needed"}

Unlock access to submit solutions for instant AI review.
Formulate a complex problem
💡 A piece of theory

The mechanics are simple: the more specific the input, the better the model reasons. Garbage in, general answer out. A good request is made up of four parts: the task itself; context and limitations; please reason step by step and take into account extreme cases; response format. An analogy: it’s like setting a task for a contractor - “do it beautifully” versus a normal technical specification with dimensions and budget. For the tariff, give: how many people, how many minutes and gigabytes they spend, the ceiling for money, what is more important - communication or the Internet. The first trick: clearly state the priorities (“budget up to X, stable connection is more important than speed”) - the model builds reasoning around them, rather than guessing. The second technique is not obvious: ask her to first write down the assumptions that she makes about your situation - this way you can immediately see where the reasoning may go wrong, and correct the introductory ones. A typical mistake: a “one-liner” question without context, after which it is strange to expect an accurate analysis.

How is it assessed · passing 70

  • 1Valid JSON without text around it25%
  • 2The required fields are there30%
  • 3The request has become specific and causes reasoning45%
Your promptClaude ⌄
🔒

Unlock access to submit solutions for instant AI review.