Hacker News Reader

Comment by FartyMcFarter | Hacker News Reader

> Musk said he wants all parts of the company ready to prepare 6,000 Model 3 cars per week by the end of June, triple the rate Tesla has achieved in the recent weeks.

This sounds like madness, and I wonder if anyone believes them.

Has anyone made a timeline or visualisation of Tesla's promised production rates vs reality? They appear to be constantly off, and now they're apparently doubling down and promising a production target that seems even less realistic than the ones they already failed to achieve.

View on HN

Auto final assembly lines make about a car a minute. Some run a little faster, some a little slower. About 500 cars per shift. 2 shifts a day, 6 days a week will get you to 6,000. Beyond that, you need another assembly line. Going to 3 production shifts is rare; you need some maintenance downtime.

This doesn't work like a growth curve. Once you get it working, an assembly line produces at a constant rate. Tesla's low output indicates serious downtime problems. They've shut down for some line rework, which is probably a good move.

There's still no discussion of the Fremont plant's problems from someone who has both access and the knowledge to understand what they're seeing. So everybody is still guessing. There's some CBS video, but it doesn't tell you much.[1] You need to see a video of a full cycle at the stations with problems, and they're careful not to show that. CBS shows 2-second clips of assembly stations but has detailed coverage of Musk's sleeping bag.

Robotic precision alignment by force feedback tends to be slow. Here are some DARPA videos of recent work they funded.[2] It's embarrassingly slow. (I look at that and think "ROS", which runs on Linux and is not hard real time, so you have to go slow.) Most of the stuff you see robots doing fast is preprogrammed motion. Sometimes, there's a vision system finding a target followed by preprogrammed motion to reach the target.

Apple used to have a highly automated factory in Fremont. Steve Jobs was very proud of it. Productivity wasn't very good.[3] They centered the factory around an automated storage and retrieval system, big racks with a mechanical system to put things in storage and take them out. This gets you a tidy plant, but not necessarily a productive one.

[1] https://www.cbs.com/shows/cbs_this_morning/video/FMN4XL5kYzi...

[2] https://www.youtube.com/watch?v=HPD5tUlKGMM

[3] https://www.youtube.com/watch?v=Dk306ZkNOuc

> Apple used to have a highly automated factory in Fremont.

> This gets you a tidy plant, but not necessarily a productive one.

Just build a 2nd plant. Then you'll get twice the productivity.

Or a 3rd plant, for 3 times the productivity.

The factory appears to be a very impressive semi-automated factory for 1980s to 1990s technology.

Has anyone made a timeline or visualisation of Tesla's promised production rates vs reality?

Yes. Bloomberg.[1]

[1] https://www.bloomberg.com/graphics/2018-tesla-tracker/

> Has anyone made a timeline or visualisation of Tesla's promised production rates vs reality?

Yes (search for "production target"): https://www.bloomberg.com/graphics/2018-tesla-tracker/

Their latest target seems less detached from reality than earlier ones, but 6000/wk by July 1 still seems unlikely.

Thanks!

Moreover, even if they achieve 6k/week, it'd take about 83 weeks to go through the 500k+ model 3 reservations.

That’s assuming they are going to plateau at 6k/week, which is certainly not their plan.

Only if you (strangely) assume they won't keep ramping up.

Are they planning on opening other assembly plants? Their facility is going to hit a hard limit at some point, even if they do get their act together.

Yes, they are.

Back-of-napkin math suggests that it'd take about 2-3mo, assuming that they'll need to triple capacity to make their target and that they can triple capacity twice again before exponential growth must slow to logistic growth.

> This sounds like madness, and I wonder if anyone believes them.

The people who worship him as they would a superhero.

Everyone else just sees it as Musk doing what Musk does best; promote himself.

It's a peak rate, not a sustained rate. I think their peak rate has been about 2500-3000/week over a day (so about 350-425/day), so 6000 peak rate isn't totally out of the question. It wasn't long ago that they were only doing 1000 per week for the Model 3, so why (fundamentally) couldn't they increase it to near 6000 in a few months, too?

Remember this is also an internal target from an internal email, not supposed to be a public statement. You don't reach your high external goals by setting the bar low for internal goals.

> It wasn't long ago that they were only doing 1000 per week for the Model 3, so why (fundamentally) couldn't they increase it to near 6000 in a few months, too?

It was almost 4 months ago that they hit a peak rate of 1000 per week if this article is correct:

https://seekingalpha.com/article/4143164-tesla-model-3-1000-...

Right, and they hit over 2000 a week about 2 and a half months after that (end of March). Setting an internal goal of 6000 per week (burst) in another 3 months from then seems appropriate if they want to actually achieve 4000-5000 per week in that time frame.

EDIT: If their production ramp doubling time really is 2.5 months, then based on the Bloomberg Model 3 tracker's estimate of 2600 Model 3s per week right now[0], they'll have about 4900/week by the end of June: https://www.google.com/search?&q=2600*e^(2.26*ln(2)/2.5)

I just don't understand this implicit idea that setting lower internal goals would actually help them achieve a higher production level sooner. That goes against common sense and experience.

[0] https://www.bloomberg.com/graphics/2018-tesla-tracker/

Or they could set a realistic and achievable goal, and not stress out their entire workforce (and not expose their workforce to excess health and safety risks) trying to reach a goal that they know is impossible based on their current process constraints.

Tesla’s existence itself was never realistic nor achievable, from the very start, a point that the author makes. Also, the goal isn’t impossible; in fact, the goal is close to their current ramp rate. Improbable is not a synonym for impossible, no matter how common that misunderstanding is.

As far as safety and health, I have seen no objective evidence that Tesla is worse than the industry standard range, although certainly they can improve. Media sensationalist coverage of anecdotes doesn’t count as objective evidence of being worse than the industry overall. Admittedly, this is a low bar.

They are only measuring physical injuries, not psychological burnout, then they are missing the point. You cant treat automotive manufacture like a AAA video game. Nobody dies if Arkham Asylum 5 is janky on playstation. But if you are missing bolts on your brake lines....

> As far as safety and health, I have seen no objective evidence that Tesla is worse than the industry standard range, although certainly they can improve.

You mean, apart from the investigative reporting at Reason that uncovered systematic under reporting of injuries.

This is a management technique perfected by Elon. No matter what date you set, it’s going to slip. There’s little reason to set conservative milestones.

This technique has its limits. If you squeeze enough people will be in a constant panic state that allows no forward thinking because everybody is in a rush to deliver now which then will predictably fail. I see that in my company. Every "6 month" project turns into 2 years but with the depth of thinking of a 6 month project.

I haven't been following the particulars of this storyline, but as I understand it Tesla is trying to metaprogram the automation of auto manufacturing. If it works out, everyone's estimates will be off like considering acceleration to be constant when improvements are being made at a higher level.

Reply on news.ycombinator.com