Show HN: Jacquard, a programming language for AI-written, human-reviewed code

A fresh Show HN post introduced Jacquard, a new experimental programming language built for a future where AI writes code and humans double-check it. The pitch sounds almost sci-fi: the language is supposed to show, in plain view, what a program might touch in the outside world, let people test the same code in fake or real situations, and help reviewers trust machine-written software without reading every last line. In short, it’s trying to make AI-made code feel less like a black box and more like something a nervous human can actually inspect.

But the real fireworks were in the comments. The sharpest skepticism came from people basically asking: should we really trust AI people to design a language for AI code at all? One commenter dunked on the whole premise by saying models are already bad at writing prompts for other models. Another waved off parts of Jacquard’s big promise as something that should be handled by the operating system instead of the language itself. On the flip side, some readers were genuinely impressed, especially by the idea that a function can openly declare what it’s allowed to touch. That feature got a “Brilliant” from one commenter, while another compared Jacquard’s code identity tricks to ideas they’d long wanted to see spread more widely.

And yes, the thread absolutely delivered comedy. The standout drive-by insult was “Esperanto for Clankers,” which is so weirdly specific it practically deserves framing. Another commenter zoomed out and hit the whole AI hobby scene with a wallet joke, wondering how much money and computing power everyone has collectively torched on projects like this. So while Jacquard arrived as a serious research prototype, the crowd treated it like every great internet launch: part curiosity, part side-eye, part meme factory.