No, you can’t manufacture that like Apple does (2014)
beneinstein.medium.com
beneinstein.medium.com
Having a manufacturing back ground, I did extended workbench management early in my career, this is just wrong. Sure, at high volumes casting and injection moulding is unbeatable at per unit costs. At smaller volumes, not so much. Casting and moulding tools cost a fortune, and those upfront NRC are just prohibitive for small volumes. So smart CNC machining it is. Added benefit, there are tons of small CNC machining ships around that are really good at that.
Nothing to add on ejector pin marks, if you see those on the outside of your part, your part design and manufacturing / tool design sucks. No need to be a billionaire genius, just a decent plastics engineer.
But since the target audiamce of this article are start-ups, it kind of makes sense, doesn't it? Brcause how many start-up founders have a solid grasp of hardware manufacturing, design, tooling and manufacturing processes? Especially those that want "Apple" without understanding what that means deffinitely don't.
I think this guy was just musing about Apple's quality and has watched a lot of AvE and This Old Tony on YouTube and wanted to share his opinions with the world. They're not all entirely correct.
For anyone who wants to go on the whole manufacturing journey, don't miss the Linus Tech Tips video about why their screwdriver took 3 years to bring to market (https://www.youtube.com/watch?v=2K5Gqp1cEcM). And they based their design on a fully functional production-ready screwdriver, they did not start from scratch.
Going from idea to production is really, really hard. Even today, even with CNC and 3D printing and other prototyping tools. Even with all the money LTT has to spend on it (millions of dollars, as I understand it).
The guy started bolt.io. A VC for hardware based startups and got $100 million funding. So I think he knows a thing or two about manufacturing.
At a quick scan, Bolt.IO includes "Carbon Robotics - Building intelligent robots for agriculture-related applications", "Denizen - Personal prefab offices...", and "Spyce - An original restaurant concept featuring a robotic kitchen that cooks and serves delicious meals". I guarantee those robots, office furniture, and kitchens contain CNCed parts, if not hand-fitted parts.
On the other hand, he also wrote "What happened when Apple wanted to CNC machine a million MacBook bodies a year? They bought 10k CNC machines to do it." If a CNC machine can't make more than one piece every 3.5 days, you're either building aircraft or you're doing it wrong.
IDK what the author's qualifications are but the ones you list would make him at least as qualified as whoever's "simple but wrong" take on a problem is thrice guided on Reddit or at the top of HN at any particular minute.
The reality is that if you wanna write "popular" context you gotta cater to a low common denominator, maybe not the lowest but you will have to stare deep into the abyss of ignorance and hubris to find it. A nuanced take that will pass the muster of a subject matter grey beard is going to be too nuanced, too informed by 1st/2nd hand experience and physical reality to resonate with the hordes of casual readers it needs to resonate with in order to gain any traction. Regardless of the qualifications of the author the content MUST be dumbed down to the point of being useless to anyone with the problems discussed therein in order to resonate with the audience it needs to resonate with for it to gain enough traction for us to be talking about it here (and frankly it's a lot easier to write that kind of resonating content from within the abyss than looking in above it).
I'm not really in the mood to write a sufficiently internet-nitpicker-proof wall of text but this sort of content where people opine on things of which their depth of knowledge is only a hair deeper than that of the casual or create content from that perspective despite being smarter than that is highly, highly incentivized by how we have set up the platforms and information flows that control where/how entertainment content (like a blog article or a video) gets dolled out to people seeking to consume such content.
"3 years, and millions of dollars" — is very typical when you keep going from factory, to factory, leaving tooling investment every time they screw up.
Instead, you have to pick at least a sane, and understanding factory owner, and wait for them to improve.
It depends very heavily on the complexity and size of the part.
Earlier in my career I worked in Medical/BioTech product development for a consultancy working with startups. We found the most cost effective route (in the UK) was to have tooling manufactured in China, but then shipped to the UK. We worked with moulding companies who had in house experience building and modifying tools, so they could make alterations to the tools on site.
Tooling for small plastic "clam shell" type products would be in the £20k-£50k region by doing it that way.
Being able to visit your manufacturing partner, as a small startup, is essential when launching a product. Too often physical product startups try to run before they can walk. You can always ship a tool back to China once you are up and running, but realistically at that point you know what v2 should be and retooling is required.
The classic example, internal corners: Dimension them with convenient big radius and your feature needs a single pass with a big end-mill. Inadvertantly specify them “sharp” and suddenly it’s off for broaching (if you’re lucky) or maybe a custom EDM sinker tool.
Injection moulding tools have two sides, a static and a moving side. The static side will contain the "gate" where the molten plastic is injected. You would almost always expect the moving side to be the visible side of the component, it tends to be simpler in features and polished. The static side will have many mechanical features such as ribs for strengthening, or snaps/bosses for securing other components.
When the tool opens the part will stay in the static half, then be ejected by the ejector pins. All these should be hidden on the inner side of the component.
The exception to this is components with no inside/outside, the sort of thing that may be a visible component in a hinge.
What is much more likely to be an issue for someone trying to imitate Apple are visible seams or witness marks on side actions on the tool.
A side action is where you have a part of the tool move at 90deg (or some other angle) to the main opening action. This lets you build in undercuts, or have holes in the side. (There are also collapsible cores, but these tend to be on the inside of a part)
The witness marks where the side actions fit against the rest of the tool are almost impossible to hide, you even sometimes see them on some Apple products. You need very very tight tolerances through the whole stack to ensure they are so small you hardly notice them.
Well designed plastic component are created when the engineer/designer has a thorough understanding of the manufacturing process and are thinking about how the tooling will work while they are designing the part. The issues arise when you design a part, then hand it over to someone else to design the tool with no input into the design process.
Edit:
Another one I used to have customers ask for was to have "no seam" between two plastic parts "like Apple". Fitting two plastic parts together with no seam is basically impossible unless you are Apple. It's why you almost always see a ~1mm uniform gap between the parts, it helps to hide any misalignment and gives you a closing tolerance between the parts.
How's Apple do it? Any tricks aside from very tight tolerances
(eg. Not applicable to all geometries but in 3D printed parts I've leveraged chamfers that bring a mating feature to a "knife" edge to essentially "bite in" to the other surface when I want a tight fit without gaps and without post-processing)
My guess is that the problem is more likely to be finding a vendor that can print/cut to your specification at the scale you need and, critically, who can produce it on your schedule. Any capable printer is likely already booked solid for months.
The scale makes it hard.
Better to keep profit margins up than actually make revenue.
The box contains smaller boxes of the medicine and they all come with the same description paper - all 30 boxes! I have read that paper 100 times already, there is no need for me to read it anymore.
I need to know when/how to take the drug, for how long, and what side-effects to look out for. Basically the rest is irrelevant to me.
A brief search alerts me to the presence of Cat 6A to Cat 8 shielded pass-through design with triple prong gold plated contacts, supporting POE++. Does your connector have integrated magnetics? Led indicators? Are they ruggedised? No? You’re losing out.
Here you go:
https://www.truecable.com/blogs/cable-academy/selecting-the-...
https://www.cuidevices.com/blog/the-ultimate-guide-to-rj45-c...
They're super.
People be proud as hell of their box businesses!
The title, "You can't..." starts right. The subtitle... "startups can't" takes it off course. It's not about startups, small companies or whatnot. I mean, some things require scale... but this is not really about specific design features. It's about "good enough" vs "really good."
Imagine "AverageCo." It's the 19th largest industrial cleaning product supplier in the world. 2nd largest in west belgium. They have decided to invest in a custom ERP system that will handle all their business problems. Stock. Accounting. Manufacturing. Sales. Everything. You are advising them.
AverageCo needs to aim for adequate. If they aim for excellent, they'll fail. It's not a good idea to demand UIs be slick like netflix, or clever like google. The brutal reality is that designing and project managing a good Salesforce or Oracle implementation is likely a stretch for AverageCo. They are not about to outdo amazon. It does not matter how many consultants they hire. Outside consultants get you to adequate, not excellent.
OTOH, a lean, agile software startup may well outperform amazon in some specific area. The way they will do this is by being clever, tactical, and knowledgeable in little areas. They know what's easy or hard to solve. They'll pick good goals, make clever trade offs and slick compromises. They'll choose a stack that makes the hard parts easy. They'll avoid getting bogged down, changing the plan dynamically to avoid difficulties. Those tactics are only possible when you are close to the process.
If you hire a prototype company, followed by an industrial design consultant, then outsource manufacturing... you are not close enough to the process to be brilliant. You'll need to live with adequate. This isn't because excellence is impossible, it's because you are not excellent at this.
Big companies are famously incapable of admitting they are anything but excellent at anything. Startups can be conceited too. In this case, a lot of startups need to realize that they are suits. You are not the hacker who can write his own JS framework or patch the OS. You are the guy who has no idea how to design the manufacturing process yourself.
The one you get from Salesforce or Oracle can land on either of those categories. The same applies to the one you develop in house.
Proper ERPs come from the likes of MS (MS Dynamics, an aquisition), Siebel (or whatever they are called...), INFOR and, obviously, SAP. The implementation is make or break so, regardless of vendor.
You just have to hire a SWE team to custom make one for you and then fulltime maintain it.
Companies and lines of business aren't eternal, they merge and split, and it's a big question whether the investment for a custom ERP will pay back over its inevitably limited lifetime, including the migration to it and afterwards from it.
Strong vibes from Borat's sister here, tbh.
I hate this whenever I buy a consumer product, I look at the box and think, that's a £10er I'm just sticking in the bin.
I like functional brutalist packaging, not only is it cheaper but usually also more durable and reusable due to not being specifically shaped. Whenever I buy a piece of gear that comes in such a box It makes me happy that more of my money went into the actual device, or at least funding the company to invest more in developing the device, rather than a pretty piece of cardboard.
I’d gladly buy all my tech in the ugliest refurbished bare cardboard boxes, but few companies package their goods as efficiently as Apple these days.
They have also shrunk the overall size of the boxes. A 2023 14” MacBook Pro is about the same size as a white plastic iBook G4 from 2005, but the box is about half the volume. This also lowers the per-machine shipping costs because they can fit more of them in the plane.
WF-1000MX4 and WH-1000MX5 products come to mind
Alernate title: No, you can't manufacture that like Apple does, even if you literally worked with Steve Jobs to build the iPod, iPad and iPhone, hire your best colleagues from Apple and work with the exact same factories as Apple uses
https://www.linkedin.com/pulse/what-founder-nest-thermostat-...
https://venturebeat.com/business/apple-laser-manufacturing/
Things like this provide clarity and precision, and reinforce (or nuance) the original point.
There are a couple of past discussions on this as well.
That can’t be right. 1M / 10K = 100 MacBooks per CNC machine per year. Surely it’s at least an order of magnitude or two off?
I however recall that they claimed in a keynote it took 10 hours to machine a single unibody MacBook, which would make the capacity for a single machine about 700 cases a year.
2k machines to make 200k laptop bodies for an initial sales rush over a month is 3.3 laptops per day per machine but each laptop uses multiple parts front and backplate for top and bottom. Assuming 4 major components per laptop and 5 different CNC machines per component you might be at say 3.3 * 4 * 5 = 66 parts per CNC machine per day. Which doesn’t sound that far off when they are discarding such a large percentage of the initial blank, though it’s likely some stages are much faster than others.
The equipment would be extremely expensive, but it’s also going to last for a long time.
The real problem with the point in the article is that it’s easy to get CNC work done
Edit: The tact time per casing is somewhere around 30 seconds (assuming 24/7 operations). Which is absolutely doable, not just for apple but for everyone with a demand of 1 million units per year for the same part.
Presumably they wanted to create rumors to improve peoples perception of quality. On the other hand it’s several fairly large complicated parts and their removing a lot of material from the blanks.
Also, most likely they don’t run them for a full year. They need 1mm Macbooks in a window of a few months. In which case, 10,000 machines making 100 macbooks each makes sense.
Edit: also, this is assuming every machine is the same machine doing the same thing. Which most likely is not (unless the unibodies are made in one go on a 5-axis mill)
You are not Amazon, Google, Apple or Microsoft.
It's fine, it works until it doesn't. It's not like they really know what to do with the extra VC money anyway, might as well spruce up the company to appear premium.
https://www.youtube.com/watch?v=6ktvE2vfxSQ
"You think I have time to ask a man why he giving me money, where he gets his money from? I'll take any motherfuckers money if he giving it away" - Senator Clay Thompson
VCs gave money away, so we spent it. These companies usually need 1/10th of whatever money they get for the core product (in all sincerity, a lot of them need no money, but that's a whole different story few would cop to). Money spent on everything else is, well, everything else.
I wonder what an apple car would look like.
I have used android auto and car play in the same vehicle, and car play seems noticably worse.
The top of mind issues I've seen:
Car play requires unlock when the phone is reattached via USB. With android auto, I can simply reconnect after initial configuration.
Music playback stutters fairly often when streaming spotify.
Car play and the iphone more often get into a state where the phone is connected physically but the phone "isn't connected"
I thought this was just me. I used to use Android Auto on the same car I have now, and never had this issue. Got a iPhone some year/s ago and sometimes this happens, maybe once every month or so, but thought the issue was my car or the phone itself. Good to know it happens to others.
I don't think this is true. CarPlay still covers my entire display (which I use for navigation/GPS normally) when someone calls. How this pass any sort of QA/user testing is beyond me. At least Android Auto displays a little message about it, rather than covering the entire display together with the navigation.
A fancy windshield with rain sensors and a powered rear view mirror and blah blah will cost much more.
Just went to Safelite’s site and got a quote for a few basic Honda/Toyota models and they’re all in that range.
I would argue that LEGO hides them just as well if not better than Apple.
The "not too loose, not too tight" thing is part of it, but isn't actually the key reason which is "accumulation of tolerance". Imagine two rows of 100 bricks. The difference in length between the two rows is 100 times that 1um tolerance. If the tolerance was less good, things wouldn't fit together well.
Lego model design guidelines actually have a maximum size of a "fully" connected block of bricks because eventually even this precise manufacturing can't cope. Really big models have purposefully designed decoupling between sections.
Doesn't tolerance work both ways? One brick is 1um longer than it should, another one is 1um shorter - in the row of 100 bricks those differences will mostly cancel each other out, won't they?
Also, it's not just about variance between different bricks (1x1 vs 2x1) it's about variance between different parts out of the same tool (the correct term for a mould is a "tool"), and variance between different tools producing the same brick.
Some really subtle effects affect the size of parts if you care down to the level of 1um - all sorts of temperature variation, pressure, humidity, colour mix etc.. Lego's skill is the ability to control or compensate for all of these confounding factors and get consistent results, regardless of age of mould tool, manufacturer of machine, ambient conditions etc.
However it also magnified any scratches on the surface (they would leave shadows on the white layer that made them much more noticeable) and I think that's why they dropped it for the G4 range. That just had plain white plastic.
But Huawei is nothing like a startup.
And ASML cannot manufacture that like Zeiss does.
Most companies have suppliers for a reason. I don't see anything wrong with that.
What they do, is work with their contract manufacturers (like Foxconn or TSMC) in an unusually deep and collaborative way to design custom parts and the manufacturing processes to produce them.
They are not a startup, but they are rather small.
No, a monopoly buying out all of the production capacity for the foreseeable future is not a good thing at all.
Conclusion: at a certain scale or level of execution, challenges change from a quantitative to a qualitative nature.
They built things just good enough, and just built out from there as they grew, fixing things as required for the usage they had. I think another example is writing their own PHP interpreter (or something like that) to speed up the backend when that became a legit concern at massive scale.
I think this serves as an existence proof that 'will it scale' is often a meaningless question in the early days of a company. It doesn't matter. If it can't, deal with it then in an appropriate manner with all the specific context of how it's being used - regardless it's a good problem to have either way!
Part 1, don't do things that you should know won't survive a year of growth when you could do something that you think will at around the same cost. Sometimes people interpret "don't optimize prematurely" to somehow mean don't do things right the first time; if you can do it right the first time, you'll save some headache down the road.
Part 2, if you build a complex product, it's hard to predict which parts of the product will see the most use, even if you can predict your overall growth rate (which is hard too). The longer you wait to make your product scale, the more information you have on what is actually used, and what you actually need to scale.
Part 3, if you make some reasonable choices, with today's server hardware, you can do an awful lot of work on a single machine. Current generation Epyc is 96 cores per socket, and AMD says 6TB per socket as well, although 3TB per socket might be more realistically achievable. You can do an awful lot of computing with a two-socket, 192 core, 6TB machine, it will be expensive, but writing a check is a lot easier than sharding a relational database. Get a database server, maybe a big one, and horizontally scale the frontend is a sensible basic architecture that gets you pretty far these days.
He would have spent a lot of money on scale he ultimately would not need.
See also "the slashdot effect".
Yeah, I remember when that was a thing, so many years ago....
Does anyone still use that site now? I abandoned it at least a decade ago, because it had gotten so awful.
Molded plastic packaging
4-color, double-walled, matte boxes + HD foam inserts
Apple packaging looks pretty but it’s a pain. It’s so tightly packed it can be very hard to open, and all that plastic seems massively wasteful. I’ll admit it is kind of fun peeling off those perfectly-positioned plastic wrappers on absolutely everything, but that again seems very wasteful, and you can’t re-pack it in the same way.
I love products with simple but cleverly-designed cardboard packaging and recyclable or compostable filling. Apple doesn’t do that, but they should.
Even the cables in the box have a paper rings to hold them in shape, and everything has tabs to allow for easily removing from the boxes.
The exterior brown shipping boxes also have clever mechanisms which "raise" the retail box gently out when opening the flaps.
I bought an m2 laptop a few months back, and a Pencil recently.
Both boxes are double-walled and double-nested. I find them very hard to open, and they’re close to indestructible which makes them very annoying to break down for recycling. The boxes are glued rather than folded.
The insert trays do seem to be some kind of paper composite now rather than plastic, that’s good. The Pencil one is glued in place, that’s bad. (I still have the box because it’s so annoying to recycle! I guess it’s small enough that I could just toss it in without breaking it down.)
Absolutely every component has a little individual wrapping. Again, they do seem to be mostly paper now (though with some plastic I think) which is an improvement.
It's very roughly 4-5 years since Apple stopped using moulded plastic trays in the cardboard boxes, but they do deserve special criticism for them as they were the only company to tightly glue the tray to the box.
Double-walled outer box with one end glued shut
Double-walled inner box solidly glued together
Pencil in a molded paper(?) tray glued into the inner box (on top of a supporting frame of cheaper cardboard)
Instruction manual in its own slim double-walled box
The outer box is still fairly small, but it’s super dense packaging!
I think the Pencil itself had a paper-y protective wrapping too.
The whole outer box was also wrapped in plastic, if I remember right (rather than just a sticker holding it closed).
Last year needed a new laptop, iPad, iPhone, and a whole load of accessories (from work switch) - none of them had plastic packaging.
I've wondered what Apple-level funding and equipment could do in other fields. Like manufacturing those fusion pellets, which requires extreme precision. They're looking at automating some of it using computer vision*. Apple probably has the smartest mechanical engineers in the world because of $$; what would happen if their energies were diverted to traditional industries facing bottlenecks?
* https://phys.org/news/2022-12-giant-laser-star-trek-fusion.h...
Also, Apple does a few other things right besides engineering: Understand which problems are worth solving. Pick problems that they are certain they can solve. Understand the users / customers enough to know what a good solution even looks like.
But then, traditional industries could use a dose of all of that, too.
That's because that is precisely their purpose. Non-SpaceX NASA using Russian engines was not a technical decision, it was to keep Russians busy enough to not sell out to NK/Iran. Stuff being sent all over the country and back again in Boeing, ULA, EADS, Airbus et al. is a waste of money but is to get the votes from politicians wanting pork.
It is like saying nobody can ship B2C like Amazon does, while ignoring the fact a very big junk of shipments is handled by the likes of DHL, and ignoring all the 3PLs Amazon is using for warehousing.
Also, most smaller players cannot afford it because of economies of scale makes the first ten thousand holes way more expensive than the ten millionth.
If... and the result's dependent on the if... the worker saves at a high rate, i.e. dormitory and meals are free, and there really isn't a lot to do in these factory mini cities, and saves for 5 years, they'll have enough after those 5 years to pay outright for a new 40-50m2 ish apartment in many T3 cities, so that'll be by around age 23-25. A couple, double the size place, or a place further out in many T2s. It's not a high wage, but not a terrible wage as long as the benefits are included.
City tiers: https://en.m.wikipedia.org/wiki/Chinese_city_tier_system