Moderndescartes

> I came up with a nifty implementation in Python that outperforms the naive impl by 30x, allowing a pure python MCTS/NN interop implementation. See https://www.moderndescartes.com/essays/deep_dive_mcts/ Great post! Chasing pointers in the MCTS tree is definitely a slow approach. Although typically there are < 900 "considerations" per move for alphazero. I've found getting value/policy predictions from a neural network (or GBDT[1]) for the node expansions during those considerations is at least an order of magnitude slower than the MCTS tree-hopping logic. [1] https://github.com/cgreer/alpha-zero-boosted

You may wish to search for "readability at Google". Here is one article: https://www.moderndescartes.com/essays/readability/ (I have not read this article closely, but it is about the right concept, so I provide it as a starting point since "readability" writ large can be an ambiguous term.)

Taste is the field behind the goalposts, and we as humans are constantly expanding that field by "having taste" and developing taste as a community. https://www.moderndescartes.com/essays/taste/

Could be a good candidate for factobattery. Overbuild the system, run them at full speed at peak solar generation, then underclock them at night. https://www.moderndescartes.com/essays/factobattery/

Even pumped hydro is pretty small for seasonal storage. Probably the only economical approach will be some sort of factobattery[1], which can be thought of as a flow battery which operates "once through" to avoid round-trip losses. Instead of outputting electricity, the beneficiated material itself is the economic output. This eliminates the entire chemistry-to-electricity side of the process, which may require dissimilar equipment under some approaches. [1] https://www.moderndescartes.com/essays/factobattery/

For non-Googlers like me, here's some background about "readability": https://www.moderndescartes.com/essays/readability/

This report, which is often quoted, https://www.eia.gov/analysis/studies/powerplants/capitalcost/ gives a crazy low cost for a solar + battery plant that assumes storage for an hour and a half which is certainly too little. When I split out their generation and storage numbers and put in the assumption that 12 hours of storage gets you through the night the price is getting in the same range as gas turbine power plants. There's the seasonal problem too, the answer to that is some combination of building more solar capacity or adding huge amounts of storage. I'd estimate that the daily insolation varies by a factor of 2 or so in NY https://www.solarenergylocal.com/states/new-york/new-york/ so you could build maybe twice the solar capacity and have enough generation in the winter. Judged that way the system cost is creeping in the direction of what nuclear energy costs, though you've got a lot of "free" electricity in the summer although that could be "free as in puppy". Hypothetically you could do something like desalinate seawater and pump it uphill into reservoirs but operating any kind of industrial factory intermittently is going to be murder for capital and operating costs. There is this idea https://www.moderndescartes.com/essays/factobattery/ where you could smooth out diurnal variation in a "hydrogen economy" factory by overbuilding electrolyzers, but to take advantage of "free" summer electricity you might have to lay off all your workers half the…