What are LLMs good for?

I’ve been thinking about where LLMs/agents have had a genuine impact on my job as a technical writer, and it’s basically two things:

1. Markdown formatting. My writing style is to start by dumping as much as I can onto a page in stream-of-consciousness, then start crafting that into a tight, coherent document. It’s truly helpful to be able to take 300 words of brain vomit and turn it into a table, or a list, to quickly split one table column into two, to rapidly iterate to find the most effective way to present something.

2. Research. No, I don’t mean asking an LLM for information. I mean using an agent to help me find internal documents, Jira issues, support tickets, and Slack threads (via various CLI tools) that might be relevant to a particular PR or code snippet so that I can better understand the impact of a feature or bug fix on users. This saves me many hours each week and prevents my brain from being totally spent before I even start writing.

I don’t want to understate number 2. This is not a “5% better at my job” kind of thing. It’s easily doubled or tripled my output and raised the quality of the resulting documents while improving my own understanding of the product.

Surely there’s some way to develop tools for these use cases that, you know, isn’t an ethical mess?