Venkatesh Rao

I’ve mostly been on the analysis/destruction/maps side as both an engineer (math/models over making) and writer (nonfiction over fiction). Now in midlife, I feel myself changing gears on both sides. And the biggest challenge is just the excruciating slowness of creation.

It’s not so much that it’s harder or different. It’s the same kind of thinking. It’s that everything takes 100-1000x longer because there’s a 100-1000x more interchangeable detail to lock down. And if you don’t make an active effort to make it interesting, it’s boring by default.

Sometimes a creative Aha! leads to a network effect in your head and vast trees of detail bind themselves down with low marginal effort, but that is draining and exhausting. There’s no way around the fundamentally higher work needed (time*power-output).

Simple example: my @threadapalooza thread this year is to do 99 jokes/micro-stories based on specific prompt words from people. It’s not *hard* per se, it’s just very slow going. An equivalent quality nonfiction thread would take me ~1h. This will take me weeks to finish.https://twitter.com/vgr/status/1339039499066609665 …

In each joke/story, coming up with the gag or idea takes 1% of the time. Binding it with the right kind of opinionated detail is the slow part. The more you’re a naturally abstract/analytical thinker, the more it will feel like brute force search rather than blithe inspiration.

When you read purely manually written fiction or review fully human generated design, you can tell which details the author/designer just chose arbitrarily vs deliberately. The proportion is much higher in genre fiction and generic engineering.

When you read literary fiction or review the design of extreme engineering, like the Mars rover, you are struck by how much more of it is consciously chosen or painstakingly optimized. The reason is trying to pack a lot more reality into fewer words/atoms.

This is not “realism.” It’s not that Ulysses is more “realistic” than say a pulp western or that the Mars rover is more “real” than a cheap Kia. It’s that there’s a lot more at stake, lot less room for error, and far higher cost of evolutionary learning through trial and error.