February 12, 2025

Use LLMs for Research and First Drafts

There have been many posts about AI and large language models (LLMs) and how they can be either a threat or a blessing. I've found them to be useful tools, and this blog post captures why in a detailed, succinct way.

If you want an even shorter version, my rule of thumb for AI and LLMs is: Use them for research and first drafts, never straight to production. The technology will write pull requests and bug fixes outright one day. For now, it's best used as a tool to take fast first steps, learn topics, or answer questions too niche for Stack Overflow.

If the internet is a newspaper, tools like ChatGPT are freshmen staff writers, and humans are the senior editors. The staff writers are eager to produce new content and help. That's great, but they're still new—we need to edit their work vigilantly before it gets into the paper.

Read "How I use LLMs as a staff engineer" by Sean Goedecke.