▲ | ojosilva 4 days ago | |
Which is unfortunately a problem for AI trained on Zig, it makes some AI-assisted Zig coding more challenging, like Q&A and code-completion. It's sad that this glass-ceiling has been enacted for new languages and frameworks, not a deal-breaker at all, just that suddenly there's this penalty on time-to-deliver on anything Zig. But then... the same issue exists when hiring good programmers for lesser-known tech. There'll probably be a strategy (AEO?) for this in the future for newcomers and the underrepresented, like endless examples posted by a sane AI to their docs and github for instance so it gets picked up by training sets or live, tool calling, web-searches. | ||
▲ | rererereferred 3 days ago | parent [-] | |
Yes, I wouldn't train AI on Zig code just yet. But here's a radical idea, rename the language the moment it hits 1.0: all documentation, blog posts, discussions, SO answers and LLMs for older versions gets automatically voided. For future languages, maybe it's better to already have a dev name and a release name from the get go. |