What people actually care about w.r.t. build times is usually the end-to-end time. Talking with @skevy, we were saying a typical production build pipeline will transpile and minify, which previously entailed running Babel (ex: with es2015 preset) and then Uglify. But now a pipeline could run just Babel (ex: with es2015 + babili).
The perf benchmarks currently make Uglify look way better than Babili to people who prefer 1.5-2x faster build times for roughly the same end result, but if the actual end-to-end build times that comprise transpilation and minification are faster, that'd change a lot of minds.
If you want a good before and after you could compare lodash-es as it has hundreds of modules that can be transpiled to cjs+uglify compared to just babili minified.
I gave it a try in a local branch. For the 619 lodash-es JavaScript files, I calculated the Transpile + Minification time and got the following averages:

Considering that:
babel.transform with the es2015 and babili presetsbabel.transform with the es2015 preset + uglifyAlso, I haven't (yet) managed to get it work with Google Closure Compiler since ES6 import and export are a bit tricky (cf this issue). But I'll keep working on it !
Any feedbacks so far ?
Most helpful comment
I gave it a try in a local branch. For the 619
lodash-esJavaScript files, I calculated the Transpile + Minification time and got the following averages:Considering that:
babel.transformwith thees2015andbabilipresetsbabel.transformwith thees2015preset + uglifyAlso, I haven't (yet) managed to get it work with Google Closure Compiler since ES6
importandexportare a bit tricky (cf this issue). But I'll keep working on it !Any feedbacks so far ?