Show HN: An automaton's code review of Gas Town with sycophancy-mode disabled
Today I got some insight into the code quality of Steve Yegge's Gas Town by running it through Sonar Cloud [1]
I followed that up with — appropriately — a vibe code review [2]…
_______
…
Brutal Verdict:
You’ve built a message bus out of duct tape and string. It’s "Booch-like" only in the sense that you’ve abstracted the concept of a channel, but the _implementation_ is pure "boot-camp-grad-on-their-third-Red-Bull."
You are using a high-performance versioned database (Dolt) to store... manually parsed colon-separated strings. _Why?_
…
_______
[1] https://g2ww.short.gy/CCWutUHalfDone
[2] https://g2ww.short.gy/ItsaGasGasGas I think this could work well as a blog post. If you're not willing to read the code yourself, maybe you could at least go through the report yourself and point out issues it highlights, show the relevant code, and explain why that's actually bad. But I only set out to get the overall vibe of what Yegge is doing. The static analysis [1] and the automaton's take [2] were plenty for doing that. It was all I could do to muster up the will to even do as much as I did eventually force myself to do. I'm not the faniest fan boy of AI-assisted coding. But I appreciate them for what they're good at. To me, it delivered a stellar code review here. The Sonar Cloud analysis is super detailed. And the automaton's code review is on the money. Digging into their respective nitty gritty is left as an exercise for the reader. People tend to read articles because the author is able to share insights that would be much harder -- whether in skill, time, or energy -- to come up with themselves. But here you're asking to put the reader to put in more work than you did, to find an insight that may or may not be there. What is your picture of the moment-by-moment thoughts of someone who would click this and upvote? Ahhh! I think I get where you're coming from. You're thinking I posted this "article" to get upvotes. Now, you got two more guesses left for "Why'd he post this?" Hint: I've written the answer twice already without me even having been explicitly asked to ;)
A blog post ain't a bad idea. > I think this could work
> well as a blog post…
I suppose. But to be frank, I'm not that interested in the project to do a super deep dive. > …maybe you could at least
> go through the report yourself
> and point out issues it highlights
I humbly don't think I could add anything that the LLM and static analysis haven't brilliantly illuminated already. > …show the relevant code,
> and explain why that's
> actually bad