John McNeill

that we weren't going to violate again, and that is automate first.

We said, you can only automate last.

And by that, we mean you've got to do things manually, kind of third step of the algorithm.

You run a process manually, and you get it really, really efficient, and then you speed it up manually to show all the warts.

You add speed, and then finally you're at a process that works.

And so we did this in the Model 3.

We had to abandon this automated line that didn't work.

And the company needed the cash from Model 3 or else we weren't going to survive.

So we had to quickly get Model 3s out the door.

And so a tent was built in the factory parking lot.

And an assembly line was built in the tent by hand.

And we started to build these things by hand so we could figure out the most efficient way then to automate and then eventually build the automated line.

And we started to produce 100 cars a week and then 200 and then 400 and then 500 on our way to 5000.

And it literally saved the company to do that.

So step one, question requirement.

Step two, you delete steps.

Step three, now we're going to run the process manually.

We're going to get that as efficient as we can.

Then we're going to speed it up with a lot of cycle time.

So we'll try to double through.