Post

This is where my opinion has landed as well. Most of the “typing” is done by LLMs for my own code these days, but I’m still in my IDE, reading, watching the diffs go by, interrupting to tell it to do something differently, refactoring regularly to make sure the code is legible.

This aligns completely with where I’m at too. Code may never have been the hard part, but it’s still an important part.

I think it’s going to follow a similar cycle to the discourse around dynamic languages. - “Developer time is the expensive bit, write in a more expressive language like Ruby or Python.” - “Actually this is really slow and annoying, let’s make it faster.” - “Maybe we actually rewrite in Rust.”

Gastown to me was like throwing 100 monkeys in a room and telling it to produce Shakespeare. We could hope that battling agents eventually produce something performant. Or we could explicitly direct the LLM on what can be improved because we have an understanding of the underlying code.

At some point the economics of have an understanding of a code base and being able to directly request changes via LLM will be cheaper than them consistently churning.

I think Gas Town’s primary value is as a form of entertainment and upvotes on Hacker News for whatever Steve Yegge is working on atm. And to be fair to it, it _excells_ at this.

The real challenge I’ve been thinking through is what’s the right fidelity of decision to delegate. I tend to be more granular than most and wanting more control over the decisions (which generally means more pre planning and keeping requests small so I can thoroughly review)