Simple systems scale better - how The Algorithm is taking over the world of hard tech

I’ve spent the last 10+ years observing Tesla and SpaceX grow from niche, risky hard tech bets into two of the biggest companies in the world. One thing that came out of that was a deep appreciation (plus a book that centers around that) for the now infamous 5-step Algorithm that is gospel within both companies:

1. Challenge every requirement that is not dictated by physics or is clearly linked to the central KPI for project success (and comes from a real human being instead of a department)
2. Delete parts or processes steps as much as possible (and sometimes even go beyond that, just to make sure you are biased towards deletion)
3. Only then optimize/simplify what is left 
4. Only then speed up cycle time (in order to increase iteration speed)
5. Eventually automate what works

This seemingly simple scalability heuristic now spread across an entire generation of (software-defined) hardware founders (often referred to as the SpaceX/Tesla mafia, because they often have prior operating experience in those companies), in the US but also in China. There are tons of books being written about it and podcasts that explore it.   

The seemingly crazy thing is that this stuff actually works - especially once you realize that basically almost everyone in every organization mostly has it backwards:

• First they automate the first version of their product or workflow or process
• then they increase their pace of operations (“it’s automated now, so let’s get more stuff done”), only to realize that it does not work properly and things are always late or brittle
• which forces them to optimize and simplify things
• only to realize that they have to delete things in order to unclog the system
• and eventually question many parts what they have built in the first place 

(which is how Tesla/SpaceX gave birth to the approach in the first place - by having it backwards, the painful way).

The reason why The Algorithm actually works is also not a surprising one, if you look at it from an information theory angle. It’s simply a repeatable approach to find the design with the minimum “information content” (see also the Axiomatic Design framework). Just think about it this way: There is a gap between what a machine/process actually produces vs. the tolerance or "target zone" you want it to hit (also called process variance). The larger the overlap, the better. So how to maximize the probability that your system stays within the required tolerances, consistently? Correct, by going for a simpler design: Less requirements, less design parameters to meet, less steps or parts to put together, less operational complexity, less things that can break, a less complex supply chain to organize and so on and so forth. Less complexity means a higher probability of success. Thermodynamics is simply in your favor here. 

The Algorithm is also a hidden rebranding of the Theory of Constraints (which in the 1980s applied the physics of flow to operational processes and always focuses on the one bottleneck in any system to remove in order to improve total system performance). Turns out it also works. There are many existing and proven frameworks baked into The Algorithm, but the main point is this: 

Since SpaceX will go public in about 2 months, and then 2 of the 10 most valuable companies in the world will operate with this simple 5-step rule as their core operating principle, you can either keep saying “those companies are a scam” - or you can ask what they have learned (within a decade) that others haven’t understood yet.

Either way, to quote the Mandalorian here:

So we probably need something like an “The Algorithm Meetup” in Munich, to build community around that?!