Hacker News

points by _davidwarner 2 days ago | 0 comments

I think both can be true: as coding languages go from lower to higher level, they go from being more implementation-like (e.g. assembly, C) to being more spec-like (e.g. Rust, Haskell). It then follows that writing a good spec could be considered coding if your compiler (or agent) can reliably turn that spec into your desired output.

So the main challenge I see with this approach is getting the spec right. Too little detail and the output is uncontrolled, too much detail and you're adding too many constraints (and you may be better off just writing code). From my experience, agents haven't been the best at writing good specs. But maybe this project can help with that?

[–] memjay 1 day ago

Yes. That’s the problem I noticed with specs too. Agents aren’t really good at finding the right abstraction layer. They tend to write way too detailed specs (e.g. naming variables in specs) and when controlling for less details the output degrades.

But I worry the most about who will be able to define all of those specs (with all of their necessary detail and edge cases etc.) without understanding the systems that live below the specs. Maybe that won’t be a problem anymore if agents will be able to 100% accurately translate specs into code. But for now I imagine this to be really difficult.