Fallbacks are worse than we think they are

Nate Meyvis argues that fallback mechanisms in generative AI tools are overrated, as reliable primary paths are more achievable than assumed and crashing or alarming is often preferable to falling back.

Here https://www.natemeyvis.com/generative-ai-tools-really-love-fallbacks/ is a post I made a while ago about fallbacks. I stand by it, but models are a lot better now at not inserting bad fallbacks. I don't know whether that's a product of my guidance, the models' improvement, or both. Now I'd like to make explicit why I like fallbacks a lot less than most of my peers. In short: extremely reliable functioning of the main, non-fallback path tends to be a bit more achievable than we think, and this is getting more and more true over time. And when the main path fails, crashing and/or alarming is somewhat underrated relative to falling back. None of this means that you should never use fallbacks, just that they're overrated. Inserting fallback paths feels to me like one of those things like writing too many integration tests https://www.natemeyvis.com/notes-on-integration-testing/ that, for sociological and psychological reasons, has become associated with wise professionalism and is overrated.