I've been thinking about different applications of large language models in developer workflows outside of code creation. I wrote down 3 main ideas in my newsletter, but this is the TL;DR for this blog:
The future of DSLs 💻 The reason why we don't have more domain-specific languages is that it's hard to get adoption and grow a community for them. If every developer in the world could now translate their code to use it, what would that enable?
LLMs as "Reviewer Zero" 🔎 A language model can be a great rubber duck, and it can also generate tests for you. Can we scale senior developers' time (and open source maintainers) time by adding a first layer of automated testing + code review powered by a LLM?
Docs generation for open source ✍️ LLMs are good at summarizing non-ambiguous information provided to them. Every open source project could have all their docs and readmes automatically generated by a model, saving maintainers a lot of time.