For Edison:

>What looked like prescience was the result of having seen more of nature.

That says it pretty well, plus determination.

I say Burbank had both, myself. This doesn't do him justice:

>was terrible at documenting and communicating his experiments, and therefore cannot be understood as a scientist.

The shortage is in understanding, not scientific accomplishments.

Too bad there was not somebody dedicated enough to follow him around with a clipboard his whole life, but I imagine he had the choice to come up with 400 of the most well-documented strains versus 800 overall which is often the exact tradeoff so much of the time. Even if the scientist themself can produce documentation that is consistently above-average, the right person can still be far more effective at experimentation by orders of magnitude, and for real progress that is where the tasks should be concentrated.

It's actually quite rare when there is any documentation at all when experimentation has been as active as possible.

IOW almost all scientific progress has been lost since like forever, and it's likely to continue.

Now Ford is another one I studied since childhood, he really put things through the roof with efficiency, turning something that "everybody" wanted but was out-of-reach, into an affordable commodity that could fly off the shelf by itself. And that was before the actual "car salesman" was invented, drawn by the incentive of a once-in-a-lifetime big-ticket purchase, it put the whole thing on steroids ever since.

But Gibson was already mass-producing mandolins and guitars in Kalamazoo before Ford built his first assembly line, and I can only imagine there were seasoned instrument wheelers & dealers who really perked up when they saw cars coming down the pike :)

>The Bitter Lesson teaches us that innovation in complex domains emerges from systematic exploration of combinatorial complexity, not brilliant insights or domain expertise.

I feel this way too, even as a pre-teen, many decades ago, but if you already "invent" something every day anyhow you're probably going to want a sweeter lesson to go forward with.

How about taking the domain expertise which can stand on its own and often embodies brilliant insights as part of the core function, and combining that with a tool that can leverage the expertise?

Now you've built a Stradivarius for a virtuoso, but you need a virtuoso to step up to the plate and be leveraged by the tool in such a way that they become better at leveraging the tool.

This could be like an endless loop that goes on forever but I have to think you end up in a way different place when there is a human in the loop than without.

I figure when you leverage human intuition rather than be constrained by it, the bitterness is mainly somebody else's problem :)