2026-03-10 20:15 UTC
@alan@ruby.social @nick@mastodon.nickcharlton.net I tried expressing this to a coworker some time back. I was trying to make the point that the choice between human-written and LLM-generated software isn’t ‘slow but high quality’ and ‘fast but worse quality’, because LLM-generated code demonstrably doesn’t even work.
I’m not sure they understood my point unfortunately, because they ended up effectively arguing that there’s no demand for code that actually works. 🙃
Replies (1)
-
@alan@ruby.social 2026-03-11 02:16
@benjamineskola@hachyderm.io Are you familiar with NASA’s transition from human to digital computers circa 1960? Dozens of mathematicians were laid off. Others thrived for decades. The key? Domain expertise. The computer could calculate. It could not formulate. How does that relate to today? AI does the typing faster. It does not know your domain. It also needs the same kind of code quality guardrails as humans. That’s where our domain expertise and standards come in. AI needs us.