The "sub-second response loop flow state" he mentions here is "the doherty threshold"[1] just applied to programing as a particular user experience.
After working at a CDN that delivered a couple hundred kb objects in milliseconds around the world I thought "code is data, so why cant code be updated in milliseconds?" I tried starting a startup that could do this, treat code as data, and therefore achieve sub-second round-trip trial-and-error loops.
Did not pan out, but to be honest we never really got to testing that thesis. I still think it could be amazing, just not sure how much of the lang/ide/build/test/deploy/validate cycle you could integrate and how much you'd have to build.
IMHO this is the main thing that made PHP successful. The "edit a file -> alt-tab -> click refresh" test loop being faster than you could click.
After working at a CDN that delivered a couple hundred kb objects in milliseconds around the world I thought "code is data, so why cant code be updated in milliseconds?" I tried starting a startup that could do this, treat code as data, and therefore achieve sub-second round-trip trial-and-error loops.
Did not pan out, but to be honest we never really got to testing that thesis. I still think it could be amazing, just not sure how much of the lang/ide/build/test/deploy/validate cycle you could integrate and how much you'd have to build.
IMHO this is the main thing that made PHP successful. The "edit a file -> alt-tab -> click refresh" test loop being faster than you could click.
[1] - https://lawsofux.com/doherty-threshold/