There’s an interesting way to experiment with reasoning by using a language model. Start by breaking your thought process into clear points. For example, you can structure your reasoning with a beginning, middle, and conclusion. Then, remove one part—whether it’s the start, middle, or end. Once you’ve removed a section, you can test the model to see if it guesses or reconstructs the missing part.

The fun of this approach lies in seeing how the model handles incomplete reasoning. It might surprise you by filling in the blanks with unexpected ideas. You could end up gaining new insights or learning what the majority might think is the “correct” continuation based on patterns it has learned. This isn’t just about creativity; the results might even suggest what you should have thought, statistically speaking, by leveraging the shared logic of many thought processes.

If you’re curious about sharpening your reasoning or getting inspiration for new ways of thinking, this method is worth a try. Structure your ideas, remove a section, and let the model fill in the gaps. You never know—you might walk away with a better understanding of your own thought patterns or discover something entirely unexpected.