no llm in writing pledge
I decided to avoid using LLMs or AI assistance for writing.
This applies to blogs, LinkedIn, design documents, peer performance reviews at work, or any place where the intent of the writing is to either communicate a message clearly or to use writing as a medium to clarify and shape thinking.
Exceptions are still possible for me in these cases:
- collecting secondary source references
- boring technical details of a reference nature
There is an excellent upcoming article from Troy Howard on the importance of an editorial mindset in today's information-overloaded age. Editors help authors preserve their voice while also connecting to the audience so that the intended purpose of the communication is accomplished. While Troy is actually actively exploring LLM-assisted writing that puts the author in an editorial role, I have concluded that my own purposes in communicating are better served preserving the raw verbal output as is. Including typos, confused grammar, and other trappings inevitable with my background.
This may be a strange position coming from someone who is very aggressively leaning into LLMs for production code. After exploring F#, and later Haskell and Scala and various advanced programming techniques I have concluded that production code is just not a good medium for ideas or preserving voice. It is fair game to outsource to chatbots. I am still experimenting with technique and may refine this with some nuanced exceptions possibly for hand-crafting data models and types.
Writing though is very different.
The world of work is increasingly noisy and lacking authenticity and a human touch. What about being original? What of having ideas that are worth having? Or maintaining the ability to manipulate advanced concepts and make connections. All these things hold intrinsic value for me and all of them are linked to writing. That is too precious to "touch up" and delegate to the machine.