How people are using LLMs for development

Quite a few people writing about it at the moment - my thoughts are still forming, but there are some good bits, and common themes, in these.

Simon Willison - Here's how I use LLMs to help me write code

Quite a long article, and its very much worth reading all of it, but this sticks out at the top:

Using LLMs to write code is difficult and unintuitive. It takes significant effort to figure out the sharp and soft edges of using them in this way, and there’s precious little guidance to help people figure out how best to apply them.

If someone tells you that coding with LLMs is easy they are (probably unintentionally) misleading you. They may well have stumbled on to patterns that work, but those patterns do not come naturally to everyone.

Sean Geodecke - How I use LLMs as a staff engineer

• Autocomplete for things he knows, but makes sure its properly reviewed by someone with domain/language knowledge for areas he doesn't
• Asks the LLM "is this idiomatic C?" as he's not a C dev.
• Very useful for throwaway code ("vibe coding")
• As a proof reader

Phillip Carter / Honeycomb Blog - How I code with LLMs these days

• Use Claude, and pay for it (the former part of this might not hold over time, but it does right now!)
• Ask for small changes, not big ones

Fahim Zaman / Honeycomb Blog - Does AI help write better software, or just... more code?

However, this can be a subtle trap. LLMs can be shockingly good at tasks that humans find difficult, and downright awful at tasks that humans find easy as well! This can lead to more overall time spent trying to get the LLM to do “easy” things than simply doing them yourself.