Peer Programming with LLMs, For Senior+ Engineers
Programming with LLMs is both promising and frustrating. While these AI assistants can help with coding and debugging, they often waste time too. Yet for senior engineers, pair peer programming with LLMs shows real potential.
This article is a collection of blog posts written by other senior or staff+ engineers exploring the use of LLM in their work, without the usual hype or buzzwords from the usual suspects. I hope you find them useful and inspiring.
Articles and Resources
-
Practical AI techniques for daily engineering work by Sean Goedecke: Sean shares two major ways he uses AI in his daily work — The “Second opinion” technique and The “Throwaway debugging scripts” technique. I’ve been using the “second opinion” technique in recent weeks, so it resonated with me. Given the right context/problem, I’d like to try the “throwaway debugging scripts” technique as well. Sean also shares some other tips that can help you get started with using LLM tools for your work.
-
Harper Reed’s LLM Codegen Workflow: Harper shares his workflow for using LLMs for code generation. Here’s a summary of how he uses it: Brainstorm spec, then co-plan a plan, then execute using LLM codegen. REPEAT. I like the idea of using LLMs to brainstorm and co-plan before executing it. I’ve not been successful using this technique to build a complete feature or prototype. Maybe I’m doing it slightly differently, and it ends up being a lot of hours chatting with an LLM. However, I’m grateful for the times it helps me quit a project early, given that I realize it’s going to require more time and resources than I imagined. Dig into the blog post and steal this technique for your next feature or weekend project.
-
Documenting Your Prompts a Best Practice for Success by Lee Boonstra: This is a short and concise post about the importance of documenting your prompts when using LLMs. It’s the only post I’ve seen about documenting your prompts, and I think it’s a great idea. I think it’s a great idea because I’m exploring ways to use LLM tools for various kinds of work, and I found myself saving prompts with the goal of reviewing the ones that worked best. My approach isn’t structured and this post gave me a few ideas on how to do it better.
-
ChatGPT is dumber than it looks by Seth Godin: This is a short and somewhat philosophical post that reminds us that LLMs are not as smart as many think they are. Although he sings the praise for Claude, I think the approach he described is useful for any LLM. He suggest that you should “figure out how to create patterns and processes where you can use it as the useful tool it’s becoming”.
That’s A Wrap
There you go, a small collection of resources I found useful and think can help you explore the use of LLMs in your work. If you have any other resources or articles that you think should be included in this list or that I should read, please let me know using any of my social handles at the end of the article.
I’ll finish with an excerpt from Seth’s blog — When you get stuck, first ask Claude, then ask a human.
What do you think?