The optimization can include default & user supplied similarity metrics that apply to plans & sub-plan parts. These can be made compositional, but we don’t assume that they behave as metric distances by default.

Note, for example, a 4W processor can easily be enclosed in a casing that blocks radio signals & provides any required cooling/stable power. But what is the recommended steps to antennae?

Background point to emphasize where it applies:

If multiple sources of uncertainty play out on the same dimension(s) then their sum/avg. tend towards univariate (multivariate) gaussian distributions – the rate of convergence is the Central Limit Theorem, which is very fast – 12-15 flips of a bernoulli are basically a normal for calculations.

If the choice is really a mix of this or that, without summation, then it’s a mixture distribution, which doesn’t share any of the ez properties of a gaussian.

So which is which where often matters & practice should reflect that.

It would be helpful to set up a forum for criticizing & defending features of various plans, with some sort of voting process for provisionally deciding which set of arguments carry the day.

The cognitive level drill down for first level probabilities & also for Wishart distributions & other higher order interactions are interesting partitions of the event spaces. The interfaces should make it easy to add new partitions of interest that potentially increase the insights & intelligence of the estimations without messing it up or adding extra noise.

(e.g. how much will the salaray cap go up by over the next few seasons & where will the smartest GMs be working???)

Consider a different, but related top level view:

IoT + Comm + Public Phone/email + Work Related Computing System + Systems of Identitification

For this system level, we can conceptually ask what is the value of various information security architectures & within that question, what is the value of various information security protocols, and within that question, what is the value of fresh silicon contributions vs. reworking of existing mfg. devices.

That’s very compicated. But there should be good analysis to where important needs are not being met & how to work around it. The FBI reading all the glass screens and all the big private CERT keys is, for example, one big issue among many.