Crosscut

The most recent change I was making to the previous prototype, was to add a static type system with full type inference. Which turned out to be too ambitious. I got the type inference 95% working, but those last 5% turned out to be very challenging, with no end in sight.

To be clear, this alone would have been easy to fix: Just roll back the most recent changes and require explicit type annotations. Or switch to a dynamic type system. Or not have any type system for a while longer. So this work, while being the trigger that got me thinking about a new prototype, wasn't actually the reason I started working on that.

But it helped me recognize all the other ways in which the previous prototype had been overly ambitious. And more importantly, how I could address those.