Grove, Andrew, High Output Management

I'm a chapter into High Output Management, by Andrew S. Grove – this is supposed to be the hook chapter – and I hate it. Let's have a goal, he says, to serve a breakfast. The stylized breakfast consists of a piece of toast, a coffee, and a three-minute egg. According to him, you time the preparation and assembly of each thing by working backwards from the limiting step, i.e. the thing that is going to take the longest. Naively, this may be the 3-minute egg, but if, say, the toaster has a queue in front of it, it becomes the toast.

Why? The naive version of this makes sense – make the whole process take the time of the longest step, and do everything else concurrently to that. Why does the toaster queue make it the "longest step"? There's frame-of-reference shenanigans happening here – the algorithm from the perspective of the "make one breakfast" computer is not the same as the algorithm from the perspective of the whole kitchen. Why is there a toaster queue? Ah, but perhaps we don't need to care about the toaster queue at all. Perhaps it isn't our business. We just need to get our part of this done – one breakfast.

I hate the book because at the end of the chapter he says that the limiting factor of the US prison system should obviously be the conviction, not cell availability. Therefore we should construct more prison cells to avoid compromising the quality of the only point of the whole process. Sometimes, with strong opinions encoded into the text of this kind, I can be like, okay, this is dated, whatever, the material still adapts well. People are wrong. People disagree. I don't have to let this book be about this disagreement. But in this case, this example in this place illustrates a few things too neatly. So far, this just-so story has relied entirely on rhyme as reason, and it was working on me. I was bored, but nodding along. Sure,

• we need to fix problems at the lowest-value stage possible.
• you want tests at intake steps to prevent late failures.
• you want not to compromise on the quality of your output, i.e. you need to pass the final test.
• you should consider maintaining an inventory of resources you need at each stage to prevent unavailability from cascading down the process you designed.

Wait what do you mean

• the quality of the justice system is determined by the courts
• convictions are the most expensive output, i.e. the limiting factor
• our current limiting factor is prison cells, which is compromising output quality
• build more cells because the relative cost is cheap

??? How did we get to the American prison-industrial complex from spherical McMuffins?

But it makes sense, see, because this process is goal-agnostic. It actively factors out whether an output is or isn't an output you want. It says, you have an allocation of resources Y with which to solve for X, go on, write the equation. Clean.

How do you surface information in this system? You do it via tests. What if your tests don't capture something? I'd imagine, at this stage of reading, that he's about to bolt on an error handling mechanism. This is about the kind of management that assists in the upward flow of power; in centralization of knowledge and decisionmaking. It's a book about how to be a good sensorimotor ganglion. When a structure like this is built, it's at risk of getting co-opted.