Hacker News Reader

Comment by codeulike | Hacker News Reader

There is an interview somewhere where Musk admits that their 100% automation plan was a bad idea. And the interviewer says 'well everyone told you it was a bad idea'. And Musk says (something like) 'everyone tells us everything we do is a bad idea'.

edit: interview is here: https://www.bloomberg.com/news/features/2018-07-13/-the-last...

100% automation was a gamble that didn't pay off. They scrabbled around a bit, and got to more-or-less where they needed to be via a different route.

View on HN

From watching some videos on the workings of their factory, it still seems like they've automated significantly more than their competitors. They reached for the stars and fell short, but that short is still ground-breaking.

And resulted in production hell, didn't it?

This is only going off of what I've read from private interviews with Tesla employees (I'm sure there are some lurking on HN that can give a better idea), but it seems the production hell came mostly from the constant product updates that were rolling through. A connector on a wire harness becomes obsolete, and all of a sudden your camera that picks out that connector from a bucket needs updating, the arm that grips it needs changing, the milling machine that machines the hole for the connector needs changing etc. One up-issue knocks on to the entire production process. This isn't an issue with humans. You tell them the new part they need to use and that's it.

building cars in a tent is called disruption

That's called dilettantism at best combined with desperation. How would call manufacturing chips in a warehouse instead of a chip fab?

Yep and scrappy, a very good attribute for a startup or even for a maturing silicon valley company.

Tesla's manufacturing capabilities are child's play in the automotive industry. Nothing Tesla has done here is revolutionary. Their quality control is abysmal compared to other major OEMs. They make the Model 3 in a tent. There are regular garbage fires at their Fremont plant.

https://boingboing.net/2018/08/24/fire-breaks-out-at-tesla-f...

And there was another one in February this year.

How are fires at a paint shop indicative of abysmal quality control?

It is a sign for abysmal factory operations. And the first time pass rate is really bad compared to other OEMs.

> It is a sign for abysmal factory operations.

Not really. It's just a sign that their paint shop needs some work.

> And the first time pass rate is really bad compared to other OEMs.

For a few weeks in June. Great work taking isolated incidents and generalizing across the board.

The fires aren't just in the paint shop. They are all around the factory property. Inside and outside. Garbage fires on the lot happen about once a year.

Regardless, you are trying to say that paint shop fires mean they just need a new paint shop. What I'm saying is that their entire manufacturing operation is amateurish compared to the rest of the auto industry.

I mean, the BMW plant near here has fires about once a year two, if the last couple of years is a baseline. I'm not convinced that you are proving anything.

It wasn't just those few weeks in June. Their first time pass rate, on weekly, monthly, and quarterly basis is really bad compared to other OEMs. A large part of the reason they have a repair parts shortage is because spare parts that would go to existing customers are instead being used to repair newly-built cars before they leave the factory.

Tesla's gotten better at managing their first-time pass rates, but they're still close to the bottom compared to their peers.

> really bad compared to other OEM

Compared to companies that have been in operation for decades, and in some cases a century? Sure, let's make sure to compare all start-ups to entrenched incumbents.

Tesla is not a startup! It's been in business for 18 years.

Depends on your definition. I think they qualify as a startup because they're still in their hockey-stick growth phase.

In the context of the auto industry, it absolutely is.

  Peugeot - 1889
  Daimler - 1890
  Fiat - 1899
  Renault - 1899
  Buick - 1903
  Ford - 1903
  Rolls-Royce - 1904
  General Motors - 1908
  Alfa Romeo - 1910
  Dodge - 1910
  Chevrolet - 1911
  Aston Martin - 1913
  Maserati - 1914
  Mitsubishi - 1917
  Bentley - 1919
  Citroen - 1919
  Chrysler - 1925
  Mercedes-Benz - 1926 (1883 if you include Benz & Cie.)
  Nissan - 1934
  Jaguar - 1935
  Toyota - 1937
  Jeep - 1941
  Ferrari - 1947
  Honda - 1948
  Land Rover - 1948
  Lotus - 1952
  Suzuki - 1956
  Subaru - 1958
  Mini - 1959
  Mazda - 1960
  Tesla - 2003

And this is explaining why Tesla ignored more than century worth of automotive mass production knowledge again? Because they are a start-up?

Regardless of our disagreements on Tesla -- this is a great list and really interesting to see all these OEMs, in order. Thanks for this good stuff. Upvoted.

I didn't know Peugeot was the oldest of the gang!

    Detroit Electric - 1907

The granddaddy of electric cars which out-produced everyone.

Nissan finally caught up in 2010.

Tesla didn't outproduce it until 2012.

Manufacturing capabilities =/= % of process automated. I'm not saying they're effective at production output, or quality control, or a number of other things. I'm just saying that as far as I can tell as an interested outsider looking in, their automation processes are impressive compared to their competitors.

Tesla admits that large portions of their cars--even the Model 3--are essentially built or assembled by hand.

Meanwhile, Ford and Toyota have videos of almost an entire car being constructed by robots on their production factory lines (their are parts where humans assist the robots but otherwise most of the work is actually done by the robots). There'a a famous Ford video from a decade or two ago showing one of their (now-older/discontinued) model sedans from being built and finished--entirely by robots-from frame to final painting touchups. And if you don't believe the videos, you can tour their factories and actually see their factory lines in action, live.

Unfortunately it is not looks that count but results. And there Tesla is far behind. Which kind proves the incumbents right on the way they are manufacturing cars. And that is my whole point. You can have a killer product, if you are unable to produce and distribute it that doesn't mean shit.

Ideas are a dime dozen, right? Execution is what matters is the start-up advice, right? And on that front Tesla is not looking good right now.

The 'tent' was an improvisation when they realised their automation plan wasn't going to work. It was 1 of 3 lines I think, a way to expand capacity quickly. You could scoff at the 'tent', or you could think about what it takes to come up with a rapidly implemented and physically massive plan B involving hundreds of people and huge amounts of hardware when there's a huge amount at stake.

edit: Here's some pics of the tent, and description of the large scale scrambling it took https://www.bloomberg.com/news/articles/2018-06-25/the-futur...

Or you could ask yourself what happened that you needed a tent in the first place. Again, incumbents or any of their final assembly contractors know exactly how to launch and scale assembly lines for new plants and car models. Why Tesla choose to deviate from all industry best practices in that case still puzzles me.

It's curious to read this sort of reaction on HN, because it maps pretty well to process differences between startups and big, established (some would say "fossilized") software development firms.

"That startup has no idea what they're doing. Just last week they had to hire a bunch of new developers and they went all-remote and picked up a bunch of new management SaaS to do it. Look at IBM, they always know exactly how many developers they'll need for a project."

Tesla's got some significant problems, sure, but most of their problems seem to be the sort that a rapidly growing startup would have (including the lightning-rod CEO).

Software =|= Manufacturing, comparing the two looks like a fallacy to me.

EDIT: That Tesla is still treated as a typical Silicon Valley start up and not an automotive OEM might be one of Elons greatest achievements.

Innovation is a gamble, and Tesla take big gambles. That's my point. The Tent happened because one of their big gambles failed.

If you think Tesla's factory is amazing, you need to check out Toyota's factories. It's like the difference between a Fisher Price playset and Disneyland.

Tesla talks big about automating...but Toyota actually did it, without fanfare, and managed to do so in a manner consistent with the kanban manufacturing philosophy that has defined their productivity and quality achievements of the past several decades

Toyota used to produce ~440k cars out of that plant. Tesla does ~360k. I'm not sure that "fisher price playset" vs disneyland is the analogy that I'd use for that difference.

NUMMI was Toyota's oldest and least productive US factory, and the cars were produced on contract for GM, which sold them under a GM brand. This is one of the reasons Toyota closed the plant and sold it to Tesla.

NUMMI is also not the factory that Toyota shows off in its videos demonstrating the awesomeness of its manufacturing processes.

There's more to plant quality than output alone. I'd expect a "disneyland" factory to have other metrics, like low rework, safety, worker satisfaction/retention, responsible procurement & disposal of raw materials, etc.

Are there decent videos of both factories to compare them? It might be interesting to see the state of the practice for an established player like Toyota.

I also remember way back when steve jobs tried to automate the production of NeXT cubes and broke his company doing it. I think a lot of folks do this sort of "follow one philosophy" into ruin. They want to be amazing and lose sight of what's doable/affordable/realistic.

Often that's what it takes to achieve significant progress, even as an individual. Shoot for the stars and you'll likely see modest progress, but if you aim for mediocrity, you might not even attain that.

100% automation was a gamble that didn't pay off

For now.

When they have a solid product, and a lot more money in the bank, I'd expect them to be at the forefront of automation, and then it will pay dividends, they just tried to do it too soon.

There are plenty of car companies out there with solid products and very deep pockets. Nobody is going this direction. All it will do is increasing the production issues ten fold if Tesla is facing the first facelift or model change.

Nobody went in the direction of seamless internet payments before PayPal, despite banking existing for centuries.

Nobody went in the direction of landing boosters, despite rocketry existing for decades.

'Nobody is doing it', is more often than not, an argument for 'it' instead of against.

It's not enough to have, 'Nobody is doing it.' If you want the full Musk analysis, you also have to do a 1st principles analysis of the true cost. Then you have to identify the factors holding things up across the entire industry. Is it that government regulation and meddling is distorting the pricing and scaring away competitors? Is it an entrenched industry with no motivation to innovate?

Often nobody's doing it, because "it" is truly stupid. What Musk is doing, is applying a formula for discovering when the majority are mistaken.

http://paulgraham.com/say.html

True at first glance. But the better allogy would be a better way to process cheques in the case of PayPal. And in the case of producing cars ignoring the Toyota Production System, ignoring decades of experience in high volume car manufacturing is at best incredibly arrogant.

> Nobody went in the direction of landing boosters

Of course they did, but they only wanted to do it on DoD money. When that wasn't further forthcoming they pivoted to other projects that were funded.

Landing rocket as a stalled idea wasn't due to lack of technical foresight, it wad a result of big companies lacking financial foresight. I give credit to Space X for making that business leap, even though landing their boosters was actually an ad hoc response to their initial failure to secure cheap Russian engines.

I once found a study from EASA, conducted in the early 2000s if I remember well, on reusing boosters. The result was, in a nutshell, technically feasible financially not so much due the low number of launches, costs to refurbish and such.

And now Arianne Space is having a serious look at it with Arianne 6.

Remember Nokia?

They had decades of experience in the industry and the software guys (Apple and Google) drove them to bankruptcy.

Also look at SpaceX - he’s not a complete novice at complex manufacturing, and knows how to assemble and motivate a good team, and is used to competing against even more entrenched incumbents.

I’ll be interested to see how this works out but I think Musk was just too early and trying to do too much at once; he’s not wrong that automation is the future.

And Nokia lost because the new products were better. Not because Apple or Google were better at manufacturing. After all Apple went to Samsung for screens and sub contracted manufacturing to Foxconn. Google designed and had the likes of LG build their Pixels and Nexus phones. And that is the difference I'm arguing here, not the product.

Also, building mostly single use rockets, reused boosters are pretty new even for SpaceX, in comparatively small number is totally different from building and running factories that spit out hundreds of thousands of cars. But I can understand why Elon might have thought he knew better. But drawing lessons from an industry like aerospace and rockets which had only a hand full of defacto state subsidized players to one as competitive as automotive was, it seems, a bad idea.

Yes, I fully agree that Elon wanted too early and too fast. And having the timing wrong still means you're wrong. And in that case this mistake and the underlying culture and way of thinking can very much doom the whole company. And that would really be a shame, wouldn't it?

The Model 3 is better, it just needs to get cheaper - it’s already the best-selling EV in many markets.

In fact others seem to go just the opposite direction. For example Mercedes is reducing the amount of automation as it makes the change too slow [1].

[1] https://www.theguardian.com/technology/2016/feb/26/mercedes-...

Reply on news.ycombinator.com