AI Policy, or How I Use Generative AI

I make wide use of LLMs and generative AI for this newsletter – but everything you read here is crafted by my mind and reflects what I think. For transparency, I believe it's better to provide a list of what I do with AI (and also what I refuse to do).

What AI do I use?

• NotebookLM – for information retrieval in large documents, audio or video, and transcripts of audio files such as interviews
• Claude – for more intensive tasks (see next section for details)
• Perplexity – for research
• ChatGPT and Google Gemini – for easy tasks like thesaurus, reformulating phrases, etc.

How do I use them?

• Better understand what I think through meaningful questions posed by AI
• Provide summaries of documents that I have read
• Find information in books or other materials
• Transcribe interviews and conversations
• Proofreading (English is not my primary language!)
• Improve text with suggestions: alternative phrasings, better word choices, clearer sentence structures, and style refinements
• Tagging, excerpts
• Translating

How I don't use AIs:

• Produce more text: I invest in quality, not quantity and I don't want to improve the enshittification of the web with more AI alop
• Write entire articles from scratch
• Publish text without supervision
• Bypass fact-checking

In short, AI is an assistant and, in rare cases, a co-creator – but even in these situations I remain the principal author.