I elaborate on this more in this blog post where I explain why consistency and predictability are signs of bad engineering, but the basic idea is: - repeating yourself is bad - boilerplate is bad haskellforall.com/2024/07/soft...

Show Post

To understand why I’m so down on agentic coding I should explain what I think LLMs are good at LLMs are good if you want *consistency* because they are statistical models with extra steps. If you ask them what comes next they’ll bias to what came before … but consistency is *BAD* for programming

Show Post

I’m told this is more efficient

Show Post

It’s okay, we can fix agent-generated code with better tests *begins generating tests* It’s okay, we can fix the tests with formal specs *begins generating formal specs* It’s okay, we can fix the formal specs with better intent docs *begins generating intent docs* It’s okay, we can fix the int

Show Post

"Why are you writing code by hand? It's not very efficient, you could do that with AI." Oh yeah? Well living is not very efficient, you could save energy by being dead, why don't you go upload your brain to a GPU.

Show Post