Where I just list a bunch of techniques and prompts.

One must possess the will to think in order to maximally leverage the effective use cases of LLMs.

Motivation is at least as important as method for the serious thinker. The essential element for successful work in any field was "the will to think". This was a phrase he learned from the nuclear physicist Enrico Fermi and never forgot. "In these four words," Shockley wrote later, "[Fermi] distilled the essence of a very significant insight: A competent thinker will be reluctant to commit himself to the effort that tedious and precise thinking demands -- he will lack 'the will to think' -- unless he has the conviction that something worthwhile will be done with the results of his efforts." The discipline of competent thinking is important throughout life.

— William Shockley


  • It's very helpful to have the right context such as giving multiple ideas for the LLM to use rather than use its own. LLMs tend to default to not-so-good frameworks.
  • Ask to search your past week of chats and have it quiz you over things you learned.
  • Give 2-4 specific ideas that might have connections and ask what those connections are.
  • Ask what first principles the last output was based on.
  • Thinking in writing feedback loops: present an idea you write, ask for feedback, write more, ask for feedback..... (especially helpful if you have uploaded files and want to specifically work with that knowledge)
  • Present an argument and for the strongest steelman against it.
  • Present an argument and ask it to identify weak spots and assumptions.
  • Present an argument and ask it to identify possible emotional motivations/blind-spots instead of pure logical errors.
  • Ask what mental models are most useful to deploy for the problem at hand and why
  • Ask for psycho-active questions to ponder about (and how it came about those questions)
  • Ask what things you would have to encounter to make your idea false
  • Ask for examples in different domains
  • Give it a text you are about to read and ask it to generate some psycho-active questions you should ponder before reading
  • Ask for ways to use the information in real life and actually do the things
  • Ask for many iterations of a request rather than just one good one
  • Describe the effect you want rather than the thing you want
  • Give examples of what you like and ask it to identify the underlying patterns, then apply those patterns to something new
  • Ask it to explain what's _not_ true that sounds like it should be true given what you just learned
  • Ask about edge cases
  • Describe a decision you're stuck on and ask what information would make the decision obvious
  • If you're not getting what you want, paste its response back and explain specifically what's missing or off
  • For complex tasks, ask it to first outline its approach and check with you before executing
  • Ask it to explore 3 different branches of the last output by tinkering with different variables
  • Ask it to imagine a future where the current idea or plan has failed catastrophically and why
  • Request the model to show its work or think step-by-step before providing the final answer
  • Ask to generate a thesis, anti-thesis, and synthesis of different arguments
  • Ask for multiple framings of the same problem from completely different angles
  • Ask what adjacent problems share similar structure
  • Request analogies from domains you'd never think to connect
  • Ask to identify second and third order effects of an idea
  • Request confidence levels on claims and what evidence would shift them
  • Request what the model is most uncertain about in its response
  • Ask to convert abstract concepts into specific implementations or experiments
  • Ask to translate between mediums (concept to diagram, text to code, framework to checklist, etc)
  • Request what downstream problems get easier if get this one right
  • Ask for open-ended quizzes forcing you to think in writing
  • Ask to identify the 20% of the problem that drives 80% of the outcome
  • Present your reasoning step-by-step and ask it to identify weak spots
  • Ask for it to give you a harder problem you have to solve using the same first principles
  • Request an explanation of why the code works, not just that it works
  • Paste unfamiliar code and ask to trace through execution with a specific input
  • Ask what logging or console statements would most quickly isolate the problem
  • Ask to review your code as if it were a pull request, focusing on maintainability not just correctness
  • Ask what mental model experienced developers use when thinking about [thing]
  • Paste an error message and ask to explain what the system is actually complaining about, not just how to fix it
  • Present a problem or concept and ask what first principles are needed to understand or solve the thing
  • Present a thing and ask what first principles are at play
  • Ask what concepts from other fields describe the same underlying dynamic
  • Ask what emotional or identity-based reasons might make certain conclusions feel more appealing to you
  • Ask for a concrete situation where you'd actually use this and what the trigger to remember it would be
  • Meta: ask what prompts from this list would be helpful in the current context of your conversation
  • Ask to generate a knowledge graph/skill tree you can systemically climb in the context of learning something new or developing a skill.
    • \
  • Export some relevent data and ask the LLM for any useful patterns/knowledge.
  • Describe a skill you're trying to build and ask what deliberate practice exercises would target your specific weak points.
  • Ask what questions a domain expert would ask about your idea that a beginner wouldn't think to ask.
  • Give it some info and ask it to generate a questionnaire or worksheet based on the info, then have it analyze it.
  • Describe a pattern you've noticed and ask whether it's a real regularity or a coincidence shaped by selection bias.
  • Ask it to generate a "what would have to be true" ladder for a goal, breaking the goal into the necessary conditions stacked underneath it.
  • State a goal and ask what specific, observable evidence would tell you you're making progress.
  • Ask what the load-bearing assumptions are in your worldview on [topic] and what happens if you remove each one.
  • Ask what part of your confusion is genuine uncertainty vs the mind refusing to look at something it already knows.
  • Describe a stuck situation and ask: "What if the stuckness is not a bug but the output of a system that's working exactly as designed? What is it optimizing for?"
  • Ask what the maintenance costs are of a belief or commitment you hold, not just the benefits.
  • Make sure you are aware of framing/anchoring biases in LLMs. The way you frame a question to it will shape its answer entirely. Ask it to not just provide confirmation bias for your prompt, but rather genuine whole thinking like a normal human.