Link: Writing as necessary friction
Sections: - Writing as fitness exercise - Writing as synchronizing with reality - Writing as theory-building - Writing as aliveness (not posted yet)
Why should we write any more? Generative AI has made it possible to generate pages and pages of writing, seemingly making the written word worthless.
However, while writing is useful for communicating ideas to others, it also has tremendous value for the writer himself. This is true even if the writer never reads the writing again.
Writing as a fitness exercise
To build muscle, you go to the gym and work out. Similarly, if you want to build mental muscle, you write. Writing helps you practice thinking and communicating. It encourages you to organize your thoughts in a cohesive way that the reader can understand. It also encourages you to shift your perspective from someone who knows the material to someone who is learning it. We also need to consider arguments against our ideas and support our ideas with more facts and references.
This is not to say that AI can’t be useful in writing. It can be used as a tool that helps us to identify weak spots in our argument or to gather research resources faster.
Writing as synchronizing with reality
When you engage in a sustained intellectual activity it compels you to pay attention to the world in a different way.
When you are writing about something, you are thinking about that thing. And often, you aren’t just thinking about it while you are writing. You are also thinking about it between writing sessions.
Writing helps you to notice details about reality and to learn more in-depth. A simple observation, when written down, may explode into a full-blown discovery over time. Once I started writing down the similarities that I noticed between Rhombus macros and ASTs, my mind kept thinking about it over time until it turned into a mind-blowing revelation.
Reality has a lot of detail, and we often brush over it. Writing helps us to stop and synchronize with reality, noticing more of those details.
Writing as theory building
No one understands the world perfectly, and we are constantly forming theories about how it works. These theories are kept in our heads.
In order to communicate these theories, we can write them down. Understanding these theories is essential to producing useful results. Therefore, when writing, we need to elaborate all of the concepts in the theory so that a new person (including future us) can pick up the theory. The moment nobody understands theories is the moment that the theory is dead.
When we use AI to produce theories, whether code, mathematical proofs, or logical arguments, we introduce cognitive debt. Cognitive debt is the gap between our understanding of the system and the actual functioning of the system. The more cognitive debt that exists in a system, the harder it is to do anything useful with the system. If a software developer doesn’t understand the code, they can’t add features, fix bugs, or otherwise improve the code. If a mathematician doesn’t understand the proof, they can’t generalize it or specialize it, nor can they extend it. Cognitive debt decreases the value of a system.
In addition, the less we understand a theory, the less likely we are to apply it. If I don’t understand git, then I probably won’t reach for it (or even think about it) when I need to track versions of a file.
Writing helps us to construct, reinforce, and communicate theories.