He mentions that re-writing a script from memory produces a better script. I discovered the very same thing in high school. The best essays I turned out where the ones were we had to write a rough draft at home, and type up a final draft due the next day, and I forgot the rough draft in my locker[1]. I always attributed that to the fact that the first draft is an act of discovery, and inevitably carries cruft from wandering paths and dead ends. Starting anew is much better at removing the cruft than editing the original.
Then I became a mathematician in college. In the maths, it's pretty well accepted that every problem and every proof, except for the most trivial, will take at least two drafts. One for discovery that is exploratory and meandering, and one that drives from the hypothesis to the conclusion like an arrow from Athena's bow. Published proofs are never in raw form, even a single page can hide a year (or several lifetimes) of refinements and incremental improvements on partial results.
Don't be too afraid to scrap your current work and start over, no matter what you're doing. The second round is much faster, and much cleaner. Such an undertaking shouldn't be done lightly, but it's often not as onerous as most people assume it will be.
[1] Yes, this happened several times. Don't mess with a winning formula.
In mathematics it's hugely damaging though, the conciseness of published proofs makes it much harder for other mathematicians to see how they reached that result. And in mathematics how you got there is as important as the end results.
Especially for students studying proofs, the step from being students to being researchers is greatly impeded by not being able to see how researchers reached the end results they did.
The space in mathematics journals is limited. They really can't afford to publish anything that isn't of the form "Theorem-Proof."
What is useful is contacting people in the community surrounding a particular result, including the person (or people) that proved it. Most are more than willing to share every aspect of the problem and how their solution evolved over the course of time.
Then I became a mathematician in college. In the maths, it's pretty well accepted that every problem and every proof, except for the most trivial, will take at least two drafts. One for discovery that is exploratory and meandering, and one that drives from the hypothesis to the conclusion like an arrow from Athena's bow. Published proofs are never in raw form, even a single page can hide a year (or several lifetimes) of refinements and incremental improvements on partial results.
Don't be too afraid to scrap your current work and start over, no matter what you're doing. The second round is much faster, and much cleaner. Such an undertaking shouldn't be done lightly, but it's often not as onerous as most people assume it will be.
[1] Yes, this happened several times. Don't mess with a winning formula.