[WheelsAtLarge]

Automation has its limits. I've tried to automate many processes at work but only the simple ones will work over time. Once you put a bit of complexity to it then it starts to fall apart. It's impossible to account for everything so eventually something happens and things fail.

We can take a hint from our current industrial automation where it's common for a process to stop and have to be reviewed by someone before it can be restarted. The stop-and-go's are worth it since the cost can be spread out over millions of units.

Also, it has been proven that the more automation you put in, the least likely it is for the whole process to work consistently.

The idea that you can put raw material in one end and have a finish product at the other end without human help is mostly a myth. But you can simplify the process and automate parts and have a great impact on the product's cost and speed of production.

There's a place for it but what gets automated has to be a very simple process and it has to impact a large number of units.

[lukemunn]

Yes, the promises of full automation have been around for at least a century, where we can go back to texts in the 20s, 30s, 40s, etc and see these claims of technical systems taking over work. In a lot of that rhetoric, the only problem is what happens to the superfluous human when that core block of identity - work - is removed. Of course, those promises have not materialized. There are many issues here in attempting to automate that huge constellation of tasks, gestures, roles, etc under the umbrella of 'work'. One growing area of work, as a quick example, is personal services - child care, elderly care, etc. This work is highly context sensitive, and often requires relational and affective labor, rather than repetitive, standardized, physical gestures. Many kinds of labor are non-trivial from a technical perspective.

[drewcoo]

Lucky for us then that good design makes complex problems simple.

Automation failures may be telling us we're trying to automate a something badly designed.

[imachine1980_]

i think you are a software guy like me, some process are to dynamic to fully automatize and make excessive division would make the process less resilient and more prone to accidents, container are cool but real container are make of real material, ship problems happens, manufacturer them isn't cheaper or easy and this real physical scale make them expensive to storage. TL;DR the world isn't kubernetes

[lukemunn]

That's a great line - the world isn't Kubernetes. Yes, in the real world, things break and problems arise. Things that worked in one context break down in another. In one of the chapters, I talk about the huge amount of maintenance work needed to keep these systems running, 'unsexy' work that never gets mentioned in the founder's keynote. There's often an immense amount of labor, knowledge, and expertise behind the scenes, some of it well paid, much of it precarious and under-paid.