Post

Gergely Orosz
@gergely.pragmaticengineer.com

I like to think of the AI tool / agent in my IDE as some kind of weird alien, and avoid mental models like "intern", "junior developer" or similar. This is because an intern would not do the same dumb thing twice after figuring out that it was dumb. This thing does.

6 reposts122 likes

Manu
@techashuman.com

I started calling it my rubber duck alter ego. It helps me discuss, plan, iterate and improve things I am doing or refactoring. Cannot think of coding without it.

Gergely Orosz
@gergely.pragmaticengineer.com

Another huge difference is how it always carries out instructions. A human would be able to say "I don't know", "I'm not sure" or something similar. This thing doesn't - unless it is set up with a prompt to get more information (such as Deep Research by OpenAI is instructed!)

Stephen Bateman
@iambateman.com

Including “ask further questions if needed” sometimes helps that. But the sinking feeling that it’s not going to be able to figure out the task is so irritating, for sure.

Gergely Orosz
@gergely.pragmaticengineer.com

And there is no point in denying to me it has capabilities that are super-human: output speed, digesting massive amounts of information in seconds (e.g. feeding an hour-long YouTube video via NotebookLM and it can analyze it in seconds, really, try it) etc Why I use this "weird alien" mental model.