Comment by aayushdutt
2 days ago
The blog post looks a bit distasteful if I'm honest. I was expecting a technical writeup explaining why it was/wasn't a good fit for zig.
> He gets to live out his productivity fantasy fever dream, he's probably already super wealthy. He has minor tech celebrity status.
Looks like a backhanded personal comment? This didn't need to be in the article.
> It's almost like the marketing department of a trillion dollar company has a lot of money riding on this article.
> The post claims they were fuzzing their Zig code, while during our calls the whole Bun team told us that they were not fuzzing anything. This appears to be an outright fabrication.
Smells like a lot of bias coming in. The fuzzing part was not a fabrication.
> sufficient to catch bugs in 1 million lines of unreviewed slop?
I don't understand how even today software engineers are calling LLM written code slop? It's objectively much better than what most of the engineers write, and it's not stopping to get better. If you still believe this, try out any frontier model on any work that you are currently doing, that should change your mind. There still are like 1% of cases where LLMs are not that good, but even there almost all of the time the issue is in the engineer's steering, not the LLM.
> There's a dichotomy being presented here where you have to either choose a "style guide" or a programming language feature in order to avoid bugs.
He says this, but doesn't give the steps to solve this?
> we all felt at ZSF that Bun was a net liability. > "It Tastes Like It's Not My Problem Anymore" > influx of tasteless AI enthusists into Zig communities
This may be the stance that encouraged Bun to leave.
> I don't understand how even today software engineers are calling LLM written code slop? It's objectively much better than what most of the engineers write, and it's not stopping to get better. If you still believe this, try out any frontier model on any work that you are currently doing, that should change your mind. There still are like 1% of cases where LLMs are not that good, but even there almost all of the time the issue is in the engineer's steering, not the LLM
I don't understand how anyone can say this with a straight face. And how you automatically assume that anyone who has this opinion, is just doing it wrong by not using the latest models.
"The significant percentage of people who are pointing out problems and issues with LLM-generated code are all just holding the tool wrong!"
While of course skill and experience plays a part, if so many people are trying to get good output from a tool and failing - the problem isn't with all of the people, the problem is with the tool.