LNSY.DEV

I have been using ChatGPT since it came out to write software. I think it is excellent for generating small snippets of prototypical code. However, I am concerned about its tendency to “hallucinate”, make odd architectural decisions and generally corrupt our codebases with a lot of verbose nonesense.

There is no software that will fix this problem – but I believe that its problems can be mitigated with process.

That is why I have a specific Obsidian.md notebook that I keep my AI notes in. With the plugin AI Commander I am capable of querying ChatGPT from my notetaking app, while ensuring no copyright-suspicious or factually incorrect information can corrupt my working notes and codebase.