Right before I left Samsung Austin Semiconductor (Samsung’s fab in the U.S.), in 2019, they were phasing out local share drives in favor of a self-hosted cloud that Samsung created. The supposed reason was better security, though it’s unclear to me why they couldn’t globally apply whatever rules they wanted to enforce to all office locations, instead of forcing everyone to use a remote endpoint. The throughput was absolutely terrible, like < 1 MBps. My department had some large files, so that was fun.
One such file was an Excel file that was more script than anything else. We had to have labels in a specific format on every machine we owned in the fab, which was something like 250 of them. The normal stuff like its id number, and also which points of contact for a technician and engineer, as well as their photos and phone numbers. Manually balancing and re-balancing every time a shift gained or lost an employee would’ve been obnoxious, so naturally instead countless hours were spent coercing VBA and ODBC to query a DB containing employee info, extracting and resizing their headshot, applying all of this to a template label, queuing a print job, and repeating. It was pretty fun to watch, honestly. I think I also had created a floor plan map somehow, and it would do its best to group a given technician’s assigned machines such that they minimized distance traveled during inspections. Anyway, the large file size was due to it caching the headshots (might have made a hidden tab for each? I don’t remember) for better performance, as that had proven to be a bottleneck.
It's always worth remembering there's exactly one reason Excel is abused -- IT isn't giving "non-developers" access to tools they need to automate their work.
I can only imagine the thousands of lines of code I saved by using Excel '97 for a scientific tabular calculation effort alone.
Because I had already done about the same thing in HP basic before PC's existed, when there wasn't a spreadsheet app anyway.
And it was on my mind the whole time.
It didn't take many more years before clients wanted their results in XLS on one tab and the invoice on another tab. Rather than the DOC files they had been wanting, which were reasonable facsimiles of the top linen typewritten paperwork I used to fax then mail by USPS or Fedex at the beginning. With the original invoice attached using a "gold" paperclip, nothing ever stapled or folded.
Oh how low I have sunk since then :\
Anyway it was pretty easy to switch from Word to Excel using the same fonts and similar formatting, even though direct integration was provided for, I just switched. When clients printed it they looked almost indistinguishable from my old typewritten multi-font documents.
And that just happened to put the paperwork onto the same spreadsheet as the data. Left one blank tab for future AI use, and the two tabs to send to the clients are in place to the right of that.
Eventually there were quite a few tabs in between the blank placeholder and the final two tabs that the clients would receive. Each with its own function to further leverage built-in Excel capabilities and make further progress toward a "paperless" office.
Saved gobs of code there too, but it gets "worse" from here ;)
It all came together by writing more VBA modules until there was an automation "shell" around the working system. Once complete it would stop so you could look at the data, decide what to type into the blank tab ("naturally", in the absence of AI) then hit the button.
while I don't doubt that such situations also exist, that wasn't the reason for any of the many "Excel abusers" I've encountered in different positions. Quite to the contrary, these people all had access to the appropriate tools, but their whole thinking was totally formatted and fixated on Excel as their go-to tool for everything:
be it things better done with a database, a word processor, a diagramming program, a label generator, a Form editor, a markup language, a web page, anything: they had all the tools at their disposal but no, no, they felt the odd compulsive need to do it with only Excel instead…
…often leading to problems down the line when the limitations of Excel for the use case (for which it wasn't made) would show more and more but they wasted already so much time and (needless) effort doing it in Excel that they would be even more reluctant to the possibility of switching to any more appropriate tool for the task.
Another reason I did so (at a different job, actually, the one immediately prior to Samsung) was that the only other language I knew to any degree was Python, and while I did compile one Python program for Windows to more easily distribute it, Excel Just Worked.
That one estimated nominal and peak loading on electrical distribution circuits by querying a DB for consumption on a given branch, and comparing it to a table containing current handling capabilities of various kinds of ACSR (overhead power lines). It was easier to do that in Excel because people were already used to it for various calculations.
Also if you’re wondering, yes, we had some bespoke software whose name escapes me that was specifically created to model electrical circuits; unfortunately no one knew how to use it, but one guy thought he did, and convinced management that we didn’t need to pay for training because he would train everyone else. He did not.
In hotel industry they need to calculate and set a price for each room/night. Typically that's done by solving convex optimization (e.g. simplex method) and using "shadow values" from it as the price per room.
Lo and behold, not every math package provides "shadow values". So you either buy a specialized math tool, or... use Excel, that has it built-in.
all about the joys and problems developers encounter using kubernetes/k8.
And reading now, today here about Excel/VBA, it struck me that something was missing: couldn't I eliminate 90% of computer software and ease 80% of software devops simply by building/selling a system that ran purely Excel atop kubernetes?
I mean, isn't that all there is to computing nowadays? Plus a little social media? And some AI for fun and maybe a little Excel/k8 work as frosting on the cake.
EDIT: Upon some further thought and prompting by helpful anonymous others, I have found a slightly different path: a FORTH/kubernetes system that will be used to bootstrap versions of VBA & Excel to the cloud. This will be made available to the great unwashed masses of humanity (and their AI friends) for implementing social media for all.
Largely, yes. Another department built a Python script that mostly was a driver for Selenium, because while they weren’t allowed direct access to some DB, nothing stopped them from automating a script to click various buttons in the shitty IE6 web app and scrape the results. I also recall they found that while passwords for said web app needed to be changed every N days, there was no lookback, so they had the script rotate between two passwords.
In my case, another not-insignificant factor was that multiple people on the team were quite good at abusing Excel and VBA, but only two others knew any other programming languages. One of them was an engineer (not SWE, engineer as in fab tool engineer - I think he was an EE) who occasionally wrote stuff in C for fun, and the other was also an engineer, but one who had convinced management that we should create an in-house SWE team. Great idea, and we weren’t the first to do so; unfortunately our manager was abruptly replaced, and his successor saw no value in the proposition.
Bingo. Working in an automotive OEM the entire company works on excel specifically excel vb scripts. No one knows why and no one has figured out how we got here. But it’s slow and steady abuse excel.
Also how the simplest, fastest and most secure way to send someone from another corporation a file for collaboration is an encrypted zip and phone in the password.
Roko's Basilisk or whatever it is should more properly be called Excel's Visual Basicilisk - once Excel becomes self-aware it is going to punish everyone who tortured it for decades.
It just sounds like Mail Merge to me. Which if you were using a computer for work in the 90s, you might know something about. (It was also convoluted) :)
> it’s unclear to me why they couldn’t globally apply whatever rules they wanted to enforce to all office locations, instead of forcing everyone to use a remote endpoint
My guess is that they're worried that you'll download data and then copy it out of the device while the device is offline. An employee could even "lose" the device, giving an attacker unbound time to extract data from it.
Another equally likely explanation is that the exec in charge of their cloud services gains more prestige due to his solution being universally adopted internally, or some other crap along this line.
It is great that particular tools enable employees automate stuff and make their work more effective.
From developer point of view I see that the effort would most certainly be diverted in another kind of solution.
But yeah, "citizen developer" stuff is a thing that microsoft pushes especially in Power Platform / Canvas Apps - one programs with WYSIWIG and Excel-like formulas (PowerFX)
But then again I wonder who are the people that can program in VBA and chooses excel. Is it the constraint around software they can use? An excel being a GUI which you don't have to implement? Anyways, a net positive for business.
I'm sure the free GUI is the gateway drug, but at this point practically everybody in finance uses Excel. You can pass files and scripts around and be pretty confident that the external auditor will be able to use them.
This is a pathology of megacorps who don't continually reinvest in process and support infrastructure and instead abuse backoffice desktop software inefficiently. In 20/20 retrospect, it should've been converted from Excel to a proper app as soon as it became a bottleneck.
A common and innovative solution to this problem now, that I have seen in other fabs (not semis, but other industries), is to put QR codes on each machine. That way the info behind it can be dynamic and maintained
At least they didn’t make you use Sharepoint and OneDrive. How fun using VBA and ODBC in 2019, proving the mighty Excel will go on as the new MS Access.
Just knowing how the thing is built doesn't seem to be enough. Comac still sources its jet engines from Pratt and Whitney for instance, despite many years of trying no local manufacturer has been able to build them to the same spec.
The product I oversee at my job is something that can only be built by people who are intimately aware of the process and have a strong understanding of the underlying engineering.
We could hand the full project file to a competitor and they almost certainly would not be able to build functional units. The failure points are fractal, so you need a strong intuition about what part you are installing, what qualities an ideal part has, what qualities the one in your hand has, how you might install it differently because of those qualities, and/or how you might change a later process to accommodate it. Or if the part should just be junked. The process is fraught with seemingly good intuitions that will ultimately lead you to failure as well.
These units also cannot be reworked, reused, or repaired, so any mistake before finalizing the build junks the entire thing.
For extremely low-entropy products, mother nature is incredibly unforgiving.
This is a bigger issue than most people appreciate, and a huge problem for the USA.
There is a specialised trade known as tool and die maker, or just die maker (https://en.m.wikipedia.org/wiki/Tool_and_die_maker) that is fundamental to a country’s ability to create industrial capacity. So far, no automation tool has been able to replace their expertise.
Without die makers, you can’t build tools that make things, you can’t build factories, and mass production in general is dependent on their skillset.
Right now, the USA only has 50k die makers, and the average age of a die maker, including apprentices, is 54. The average age of a master die maker is 73, and the average age of a journeyman is 62. A master die maker can teach about 5 apprentices to the journey man level in a work environment, after 2+ years of basic engineering school, in about 5-7 years. A journeyman may generally considered a master after 10-20 years of experience, depending on the nature of their experience.
We don’t have enough die makers to rebuild the industrial capacity of the USA, and we can’t teach the amount we need in less than 5-7decades without some kind of major change in the process of doing so. And since more than half of the master die makers are months from retirement or death, we are in an extremely precarious position as a viable industrial power.
This is why it is extremely difficult to build anything physical in the USA using only USA sourced parts and materials. It’s almost impossible to even get a decent variety of screws and fasteners made here in the USA, and we can’t easy build the machines to make screws because of the critical shortage of master die makers.
If we are to maintain the ability to build and maintain our machines, weapons, and critical infrastructure without being completely dependent on imported tools, supplies, and knowledge, we will have to reinvent the industrial process using automation or something similar to compensate for our foolish exportation and devaluation of strategic skills and capabilities.
If the average age is 73 that's the issue right there. People are refusing to retire at 62 and make room for juniors. Then when they finally do retire (73 average seems like "dies in office" levels) there are 2 decades of missing juniors who haven't been trained in the pipeline.
This is the shortcoming with most technical fields. No one is incentivized to see the big picture of the training pipeline that exists well outside the scope of their own company. No one likes juniors but that is their future.
The actual killer is the pay and working conditions, and the problem is fairly intractable.
If you are smart enough to be a good tool maker, you are likely smart enough to be a good 6-figure keyboard-all-day worker. Losing a finger (or three) and breathing VOCs all day for half the pay is not very enticing.
These industries aren't glamourous for investors either. The business proposition sucks, the cost and liabilities are intense, and the margins would need to be negative to be truly competitive.
And worse than anything, the stuff that comes from China is not only 1/10th the cost, it's also now better quality.
At this point, even if die makers started making 300k a year, there isn’t enough master die makers left in the USA to bootstrap the capacity in less than 5 decades. The only practical solution at this point is to max out apprenticeship and somehow incentivise the industry so that we could flood other countries with apprentices as well. Then, we might be able to start getting things balanced in as little as 2 decades.
Otherwise, completely homegrown manufacturing is essentially dead in the water.
If we want to regrow US domestic manufacturing, we need to throw about 100B at scholarships and incentives, or figure out how to capture remaining diemaker empirical knowledge into ML / robotics.
Or maybe we could spend 10B on immigration incentives for qualified Diemakers? Idk about the global state of affairs and how that would work. At any rate we need at least 100-300k new diemakers if we really want to rebuild 100 percent domestic capabilities overnight. And then we need a career path for them 20 years farther down the road.
It was really, really foolish to allow strategic industrial capability to wither on the vine.
Yes. Realistically speaking the value that a diemaker brings should easily be worth 2x or more their current scale.
The SmarterEveryDay piece was disheartening for sure, echoes what I have been singing from the rooftops ever since the government dropped the USA origin of manufacturing requirements for all purchases. Sure, it was a kind of corporate welfare, and it made for a tricky incentive misalignment, but it kept a lot of strategic industries on life support. We would be in a way, way better position now if it hadn’t ended.
The window has shifted as to what constitutes “us made” though. For most things “US assembled” is the norm, and when you get out into things like office supplies, vehicles, and construction materials it’s often far from true domestic manufacturing. For strategic military stuff it’s better, but often at the cost of using 1990s tech because we don’t make the newer stuff here. (It also has the effect of making things more repairable, but that argument gets weaker and weaker as time goes on)
At this point I think in practice it’s a preference for US manufacturers but a long way from robust enough to establish domestic supply chains.
Also -a lot- of military purchases are local, discretionary on a credit card, and there is no oversight on the origin there- it’s whatever they need from office max or the hardware store.
That may be the case in many industries, but I don’t think it applies in this case. The industry isn’t flooded with people looking to move up, it’s starving for interest. The problem is that you can farm out the work overseas for much less than you can do it locally, because our automation is decades behind China.
Chinese skilled labor wages aren’t that low anymore. The reason they can do most work for 1/10 of the local cost is advanced automation, government VC investment, and somewhat less c-suite greed. The incentive alignment of the government VC is also distinct from typical US VC.
We can’t even build the machines that would enable that kind of automation in the USA. We’d need to buy them from China, IF China was willing to sell them to us. Unlike the USA, China is probably not going to be stupid about diluting their advantages. I’m sure the CCP hasn’t forgotten the decades where we were unwilling to sell them high precision machine tools lol.
Unfortunately, we are more likely to see a slow deterioration of the US economy and infrastructure, until we hit about a 5-6x wage advantage over China which we could potentially use to copy the Chinese playbook and rebuild our presence.
A relative of mine worked at a medical devices company (brain sensors). She told me how small intricacies of the manufacturing process were critical to reach good enough yield or functioning devices, at all. The critical process steps were closely guarded and only a handful employees knew how to do them. The devil is often in the details - and the moat, too.
> Just knowing how the thing is built doesn't seem to be enough.
See perhaps:
> Tacit knowledge or implicit knowledge is knowledge that is difficult to extract or articulate—as opposed to conceptualized, formalized, codified, or explicit knowledge—and is therefore more difficult to convey to others through verbalization or writing. Examples of this include individual wisdom, experience, insight, motor skill, and intuition.[1] An example of "explicit" information that can be recorded, conveyed, and understood by the recipient is the knowledge that London is in the United Kingdom. Speaking a language, riding a bicycle, kneading dough, playing an instrument, or designing and operating sophisticated machinery, on the other hand, all require a variety of knowledge that is difficult or impossible to transfer to other people and is not always known "explicitly," even by skilled practitioners.
Building even basic things is difficult. Friend of mine tried like dozen companies from around the world to manufacture very basic milled aluminium part. None of them could make it. Only one company could make 1 or 2 right out of 100.
Slab of aluminium milled to a flat triangular-ish shape about 10mm thick, then inside hollowed out to half depth and 8 counter-sunk holes non-tapped.
Common issues:
- item is not flat
- nicks and dents on the edges
- non uniform depth of the counter-sunk holes
- holes in incorrect locations
- unapproved changes to design
- scratches
- pores (poor material quality)
- incomplete milling
Then parts have to be hard anodised and sand blasted.
- Sand blasting uneven
- Sharp edges remaining
- Streaks
- Contamination
- Exposed bare metal (it is only acceptable in the holes that could be used for hanging)
Companies also have virtually non existent QA. You can list them things that will cause rejection and they send it anyway.
One company told him after sending a batch that was complete carnage "We can give you $100 coupon for next order!" he insisted on rework. So they sent again, even worse.
I can't remember but I think Lance Air or Epic got split in a sale and a company bought the type design/blueprints but ran into issues actually manufacturing from them.
> Or for analogy, following a recipe doesn't make you a good cook.
For an analogy, no one cares if you're a good cook if you're able to make a passable burger. Most of the demand is not for the best burger money can buy, the just want a burger.
But it also depends on how precise the recipe is - if it's described down to the exact movements the cook needs to do, which may be replicated via a machine...
No recipe accounts for ambient temperature/humidity, very few for altitude differences, etc. etc. It still takes knowledgable tweaks to get just right.
At the end of the day, the map won't ever be the territory. Typically the properties that makes something successful are those which cannot be specified. If it were otherwise and those qualities could be specified in a reproducible manner, that thing would not be observably successful.
They exist in the industry, where a small variation of one parameter could affect tons of products. Same for individual restaurants, cooks know how to set up their equipment. For amateur cooks at home it does not matter. They do not need that level of consistency, and it would get quite expensive.
No, but they provide the best efficiency/performance for the buck. China can produce its own jet turbines, but they have to trade off performance or longevity to do it.
Obviously, but I would imagine that most of the “magic” isn’t in the supplied components, but in the finished product. Otherwise reproduction would be easier.
Yes. Their engines are mostly for military applications, tough, so they are less well known of the general public. Other than that, they are half of CFM International, but again they are not very visible.
That's the OP's point: COMAC is using CFM LEAP 1-C engines on the C919.
To be fair, they have taken the effort to build the CJ 1000A engine - which is on wing testing should the tangerine fellow cut them off. But its Plan B at best.
There is a very nice chapter in the somewhat dated but classic book Business Adventures [1] on trade-secrets and what happens when employees of one company move to another. In chapter 11, "A man, his knowledge, and his job", there is a story of a "space-suit" manufacturer Goodrich suing an employee for moving to its rival Latex for stealing trade-secrets. The story is timely in context of Meta hiring researchers from open-ai, deepmind etc for 100s of millions for the knowledge in their heads of the recipes which work for making superior LLMs - the knowledge of which is empirical and may take years to discover.
You could also argue that they're hiring people for the general expertise they have in developing frontier LLMs. Distinguishing trade secrets from general knowledge is difficult and employees building the latter to make them more valuable to new employers is an explicit policy goal of US employment law.
Lots of space technology is classified as weapons subject to export control. ITAR has plenty of rules about who can see information. US immigration status generally has to be green card or citizen, and country of origin and any second citizenships are considered.
("export control" in this sense really doesn't have to do with moving a physical object out of the country but sharing information, to the extent that a conversation in an elevator could be an export violation. most export violations amount to emails being sent to the wrong person)
When I worked briefly in defense, for example, there would be regular random searches of my stuff as I exited the building and security would wander the building and look at what you left out on your desk while you went to lunch. Entirely seriously they told us not to wear our badge in public if we left the building and not to leave our laptops in our cars because someone might follow us and steal it. Had colleagues who were visiting a foreign country for work have their hotel rooms obviously thoroughly and messily searched while they were out.
They also do national security missions so there are folks there with high clearances.
Thing is that even if you did steal a bunch of information, that doesn't mean you could just copy and be successful. Any one of a million things can go wrong with a self-landing rocket that will cause it to explode, you can't just steal the whole system of operation that keeps these things from happening.
You couldn't steal all of the secrets of a circus performer and suddenly be able to juggle chainsaws while riding a unicycle.
I would say the main element of their success is not really in any specific close kept secrets - its in actually committing to reusable rockets and keep working until they had a working system.
While there are some really nice components and clever ideas (Merlin/Raptor engines & very good guidance tech) this all really has been doable for decades in less efficient form.
But so far no one other than Space X has been able to win against all the naysayers who were so sure only single use rockets are ever going to work, get enough funding to build a partial RLV & then operate it successfully as a business.
I don't think it depends on any single technology or a set of them only they have access to - rather that they have been able to persist and see it through, unlike all the other RLV projects that never got funding to go past the paper stage or very simply not viable (Space shuttle).
So, this is admittedly a little tinfoil, but I wouldn't be surprised if Musk is happy for some degree of espionage to happen. If it looked like there was a possibility of China getting this capability first, it would light a fire under the US government to give financial and regulatory assistance to the Starship program.
Musk does seem to think in terms of how much money he can get from the government for his companies. But to be fair, government subsidies are a successful strategy for entrepreneurs who want to make a lot of money.
Maybe they shouldn't be? And I think honest people can have that debate.
But you can't really argue against the effectiveness of government subsidy as a path to prosperity for the guy getting the money.
Despite his faults he does seem to care about the actual lofty mission statements of his companies more than just straight profit. Otherwise, you wouldn't see bizarre things like him directing Tesla to open source all their patents.
I imagine SpaceX having pretty fishbowl conference rooms for customers in the center of everything, with guest network ports just segmented off from the main but using the same hardware. Oh man that would stress me out if I was IT. And of course the '<Customer name> needs to print something off and needs access to the MFCs'. No, you print it out for them, like has been discussed and agreed to and keeps with ITAR. 'No, they need access, and now. Because I'm a sales guy and I won't tell them no' but if they get into/past the MFC, it's all on IT and IT being bad at their job/security, not the sales guy that demanded they get physical access.
Spacex rocketry tech is subject to ITAR regulations. That restricts who they’re allowed to contract with, data encryption and handling, but altogether those regulations are quite bare. It likely wouldn’t be enough to stop a state actor or rogue employees.
I think ITAR is mostly just to stop the outright sale of controlled items to foreign entities, not necessarily to prevent IP theft or corporate espionage.
I worked on an ITAR-controlled camera once and it was drilled into me than even allowing a non-US person to view the output images of the camera constituted an ITAR violation.
Right, which is all well and good if you're a benevolent actor, but ITAR really doesn't do anything at all to stop you from intentionally committing (corporate or actual) espionage if you want to. There should still be controls to prevent you from violating ITAR either accidentally or on purpose.
The controls are you go to jail if you get caught. At least this is what one is told in the mandatory ITAR compliance trainings. That seems to be how the law works for most crimes.
It's an incentive but it's certainly not a control. Controls stop something from happening, they don't respond to it after the fact. Controls would be things like security checkpoints, network monitoring, random searches, that sort of thing.
"You will go to jail if you kill someone" isn't a control against murder.
I don't think what SpaceX is doing is that hard to replicate. There's already competitors launching smaller payloads for smaller costs per weight. Just a matter of time until they creep into SpaceX's market, while SpaceX tries to build a starship inspired by the Futurama rocket.
It is insanely valuable, both commercially and strategically.
If it weren’t that hard to replicate, several countries (and Bezos/Blue Origin) would have replicated it by now.
I think you vastly underestimate how difficult rocketry is. There’s a reason “rocket science” is colloquially a metaphor for an extremely difficult and technical task.
>If it weren’t that hard to replicate, several countries (and Bezos/Blue Origin) would have replicated it by now.
There is a 100% chance multiple countries/companies will have replicated it in the next decade. If SpaceX never existed, they likely would have achieved it at the same pace regardless.
This is the same with EVs. If Tesla never rose, the world EV market outside of Tesla would have seen precisely the same rise.
There is a tendency to attribute the early movers with innovation in the inevitable, where we all stand on the shoulders of others and just reach a little higher.
As to the rocket science misnomer, that's a space race hangover where an engineering role was extremely public and celebrated, but in actual reality "rocket science" is a mediocre field with miserable pay and high unemployment.
As to how valuable it is, "insanely"? The world has a fairly finite launch need, such that SpaceX made a whole new business -- Starlink -- to make work for their capacity. Economically the space launch business is relatively minuscule.
I question your EV take. Tesla proved a business model, the technology and path from niche sports car to the best selling car on earth, and now on to the lowest cost per mile robotaxi. Simply knowing that a solution exists and is financially viable, is enough to motivate the competition.
Through almost all of Tesla's existence, its business model was ironically the sale of gasoline vehicles. Because, of course, Tesla's entire business model relied upon selling green credits to incumbent ICE vehicle makers.
So it didn't really prove much of a business model, effectively being parasitic.
>the technology and path from niche sports car to the best selling car on earth,
The overwhelming bulk of the technology advancements that enable modern EVs -- from advances in batteries to cameras to sensors to embedded controllers and CPUs -- is thanks to the smartphone industry. Modern EVs owe infinitely more to those than they do to anything Tesla did.
>knowing that a solution exists and is financially viable, is enough to motivate the competition.
I think of this much like compact fluorescents. Remember those? We all rushed to transition, and then they were absolutely demolished in every metric -- efficiency, colour, and most importantly the amount of environmental contamination when disposed -- by LED lights. I feel like we're going to feel the same about early EVs.
>now on to the lowest cost per mile robotaxi
Is this a serious comment?
Absolutely a serious comment and why Waymo will likely be unable to compete. Tesla is designing their fleet for the lowest cost per mile, and likely the real reason why they have pushed back on Lidar versus camera - humans can drive with binocular (some with just monocular) vision, why use Lidar as it’s more expensive.
If a Waymo car costs $100k and is good for 100k miles, then that’s $1/mile. Tesla’s robotaxi is likely designed for 500k miles at under $0.10/mile. Tesla’s robotaxi will be able to drive me for cheaper than I’ll be able to drive myself. Think of a cab company with scale enough to build their own cabs, optimizing for the lowest cost per mile, and with $0 labor costs. That’s going to be their huge win. And no, I’m not an Elon fan, but I do understand his business plan, and it’s brilliant. And no, I don’t think it’s going to displace personal cars. There’s too much burst demand during rush hour for a peak scaled fleet to be cost effective (I haven’t run the numbers, so even that slight may be incorrect). However I do expect to never have to drive myself once I hit retirement age, and I’ll probably not even own the vehicle. My hope is that there will be a fast follower, able to compete against Tesla, otherwise I’ll be paying Tesla $0.99/mile instead of $0.12/mile.
> The overwhelming bulk of the technology advancements that enable modern EVs -- from advances in batteries to cameras to sensors to embedded controllers and CPUs -- is thanks to the smartphone industry.
Would you point me to a source where I can read about this?
If Tesla never rose, the world EV market outside of Tesla would have seen precisely the same rise.
would have seen the same rise _eventually_. I know from a friend that worked R&D at a major car company that Tesla really lit a fire under then and 'forced' them to push their own EV experiments from proof of concepts to commercial product much faster than they where originally thinking about doing it.
The EV ventures of most automakers are massive money losers (just as it always has been for Tesla outside of selling green credits and subsidies). But for sure they all rushed to get there not because the EVs themselves were valuable, but because of the insanity of the capital markets where Tesla is valued at a trillion dollars at a 200x P/E, while the rest of the market is at like a 7-14x ratio. Everyone wanted some of that irrational hype.
The problem lies in forming and managing such a huge organisation that deals with the problem in an efficient and lean way, not the technical aspects.
The materials science aspect is a challenge, not to produce, but to produce with a sane cost.
The rocket science aspect of things (namely the linearisation of the booster model in order to be able to be solved in constant time by an MPC) is more or less a solved problem.
Coordinating such complex interconnected systems will always remain one.
SpaceX itself replicated the DC-X from the 90's. The reason the DC-X was cancelled was because of the economics. Reusable rockets are a solved problem, with only the economics of it a barrier (see Space Shuttle). SpaceX has to rely on their own investor funding for Starlink to remain a viable entity.
The DC-X was stupid, dumb, insane, bunkers, and a waste of money and not replicable.
"sure, let's put four RL-10 hydrogen engines on it! They are expensive and the worst possible for the low-altitude flights we are going to do so some sucker will believe we can do multi-stage reusable flights to orbit".
"sure, we don't actually need to go to orbit anyway. That was always a dumb idea. Who said we ever wanted to do that? No, suborbital is really useful, we promise. And we are going to do it with a single stage (using hydrogen) cuz we are so smart and the future and everything!"
"sure, let's build a specialized hydrogen tank in a stupid shape."
"sure, let's give the whole single-stage low-altitude rocket a k00l shape that makes it more expensive."
"People knowing it can't be done shouldn't interrupt people doing it wrong."
Funny how many words it takes to say Elon’s dad in fact own a stake in an emerald mine, invest in Elon’s business, and give Elon an incredibly privileged upbringing.
There's a lot of unanswered questions from the article which is seems anyone with even mediocre skills as a journalist should be willing and able to answer.
It mentions the mine "collapsed" in 1989. Does that mean literally? Just financially? Was there an insurance payout? Did everyone lose their investment? Did Errol Musk own 1% or 90%?
* In quoting another article, it does say: "Errol Musk, an engineer, owned a small percentage of an emerald mine and had a couple of good years before the mine went bust and wiped out his investment."
Elon graduated from college less than 10 years later but says he was $100k in debt. Was he actually in debt? Did he spend years paying that off or did he (or someone else) pay it off shortly after in a lump sum or very quickly?
His dad provided $20k of a $200k seed round for Zip2, was the rest also from friends and family or more institutional investors? Did his dad receive equity for that or was it a gift?
Reading through the article again (and a little more closely than before) it doesn't seem Elon had "an incredibly privileged upbringing" but maybe that's a mix of good PR and this now being 40-some years ago? They're referred to as upper middle class but if that's all it takes to be "incredibly privileged" then 90% of the kids born to people reading HN are also incredibly privileged.
Arnold's piece seems to contradict basically everything else in the article so I'm not sure what to make of it, but it also sounds like he made about $350k profit over the lifetime of his ownership in the mine? Certainly not nothing but it's not opulence and it doesn't sound like most (any?) of that made its way to Elon.
> Elon mostly lived with his father, who says he owned thoroughbred horses, a yacht, several houses and a Cessna. One of their homes was in Waterkloof, a leafy suburb of Pretoria that was popular with foreign diplomats.
> Wanderlust ran on both sides of the family. On holidays, Errol and his kids would travel, he said: to Europe, Hong Kong, throughout the United States. Or they'd take the plane to Lake Tanganyika [in Zambia], where Errol had a stake in an emerald mine.
That sounds more privileged than upper middle class to me.
am i crazy for not caring if a company in a foreign country obtains trade secrets and manufactures the same thing? like we're all humans and we all want access to whatever it is they're building, it seems like more people building the thing is a good thing. if that impacts Samsung's profits, why do I care? its not like corporations give a shit about me
Well who are you? Maybe there is no reason for you to care.
TSMC alone is 12% of Taiwanese exports. The entire semiconductor industry is 25% of Taiwan's GDP. It's obvious why the Taiwanese government and society, to say nothing of TSMC's shareholders, would care.
Yes, because if you tolerate that you disincentivize actually developing the IP in question in favour of stealing everyone else's, which leads to nothing being developed.
Some countries have all the talent and manufacturing and sourcing advantages. Once they take the lead, you might never be able to keep up.
Your engineers lose their jobs, your businesses go bankrupt, you exit that entire field entirely for your entire population. Slowly your ability to do work begins to evaporate.
It's happened before and it'll continue to happen.
> Slowly your ability to do work begins to evaporate.
To be clear about this, you can still physically do the work same as before, it's just become uneconomic/not competitive. It's not all bad though. Having another country take the bulk of the market leaves you free to differentiate and specialize in one or more profitable niches that are not being served well by what is now the main supplier. You might end up dominating some fraction of what's now a vastly bigger market, with improved economies of scale that you're free to exploit as well.
That's a bit like an exercise left to the reader. Or wishful thinking.
We hope that it works out that way, but there's simply no guarantee. It's not an economic law.
If the country supplying said thing can have domestic companies deliver at lower margin, you're still kind of screwed. Their internal competition fills out the niches, which they can then export.
there's something that rubs me wrong about the argument of "we need IP laws because otherwise someone else might do the same thing as us, but better, and we won't be able to compete"
I really can't answer that, it's your call if you care or not.
I think that the more of a free for all folks stealing tech as they wish will push companies who do the development work towards more proprietary / DRM and similar solutions ... I don't like that.
Allowing people to profit from their inventions / investments encourages more such development, and without that discourages it or encourages less good options.
We are doing much better off with the status quo than TSMC firing everyone who isn't of the same religion, and even then, it's not hard for a spy to pretend to be <insert religion here>.
Because religious vocabulary, etiquette, and clothing are learned from decades of involvement usually as a small child.
It’s kind of like saying you’re going to pretend to be a different class, but more particular.
> I have pretended to be multiple kinds of Christian with success
I don’t know what context you mean. But I guarantee others can tell you aren’t a member of the group even if Christianity aims to be open in inviting new people.
Another common situation is they recognize you have some connection but can tell you aren’t an active participant.
Harder to pretend when speaking with colleagues that are the real deal. Also very likely to "step out of line" accidentally over time as lot of time is spent with colleagues.
It's not hard to pretend to be anything for 5 minutes to someone who doesn't know anything about what you're doing.
Naively, I was assuming TSMC would still operate like TSMC, not the Taliban, and that they'd focus on doing their work, not religious performance, but I guess anything is possible!
Suggesting TSMC start operating like the Taliban to protect their trade secrets seems like an obviously bad idea, though.
But we don't live in a vacuum, so I guess it's all just opinion pieces.
Okay, substitute Taliban for the Church of Scientology or the Church of Jesus Christ of Latter Day Saints.
It's a marginally less absurd proposal.
My point stands.
Proposing extreme religious nepotism as a good counter-measure to deal with espionage 1) would be minimally effective (see the IDF infiltrating Hamas and Iran regularly), and 2) is ridiculous.
Is there a Druze world government out there that I’m not aware of? :) It’s absolutely possible to convert to Judaism, for reference, though the faith is non-proselytizing (not actively seeking converts) and the status of converts is disputed among the more extreme Orthodox groups (but what isn’t).
Money and coercion is what corporations have now and it doesn’t work very well because the incentive is to violate it as soon as a better offer comes along.
Compartmentalize your company. Make sure people from one compartment are unaware about things people from other compartments are doing. Don't transfer/hire across compartment boundaries.
You don't need to deliberately create silos in order to secure a project. We've had people working on top secret projects and while people may not have known the specifics the company was kept up to date on the progress and outcomes of the project. We brought people in from other teams after an extra vetting process and additional NDA specific to that project.
Grandpa worked heat treat at the alegany national forge, where they made stuff like the beams for the empire state building, periscope tubes, and the 16" guns for the biggest battle ships, each thing had to be tempered in a very specific way, and to deal with spying and espionage, the charts and instuments used on the shop floor were all deliberatly wrong,and the written instructions were also wrong leaving the person in charge to know how to convert the given instructions into
what was actualy done through a secret method, not complicated, but essentialy impossible to reverse engineer.
There is a story of soviet engineers who somehow were invited to tour the RR jet turbine factory, and were given shoes that had extra sticky soles they wore only for the tour, which ewere then used to anyalise the metal chips that get picked up from the shop floor.....
never ends, expected.....even honored
IBM has kept researching in Albany. They license manufacturing technology to other parties even if they gave up trying to build their own fabrication facilities a decade ago.
"Out of nowhere" would be a stretch. IBM may not have mass-market volume anymore but they kept up the production chain to build mainframes (and I would guess other "specialty" products) for institutional customers.
IBM has been publicly working toward 2nm since at least 2021, though, so it doesn't seem like it's coming from nowhere. Unless this is a completely different 2nm development, here's a presser about it from 2021, titled "Introducing the world's first 2 nm node chip" (https://research.ibm.com/blog/2-nm-chip)
Unless it has great yields with 0 issues, there's always things to learn from. It's also possible the IBM process isn't what it seems and there's more to it.
I love that there is the headline, an intro, a heading for a section about the iPhone 18, and then another heading titled "TSMC says employees tried to steal trade secrets", which is literally a word-for-word substring of the headline.
There is literally no indication in the article that this has anything to do with apple other than them being a potential user of the TSMC 2nm process. Strange they tried to connect this story to apple.
Yep. It doesn’t detail what was stolen, how they were found, if they’re arrested right now, or who it is suspected they are working for. Useless article beyond the headline.
> [...] A materials scientist working for Rubycon in Japan left the company, taking the secret water-based electrolyte formula for Rubycon's ZA and ZL series capacitors, and began working for a Chinese company. The scientist then developed a copy of this electrolyte. Then, some staff members who defected from the Chinese company copied an incomplete version of the formula and began to market it to many of the aluminium electrolytic manufacturers in Taiwan [...]
TIL! The only time (knock on wood) I ever had a motherboard go bad was an MSI from around 2002, due to a blown capacitor. To their credit, MSI had great support at the time, and sent me a new one after I sent a photo of the blown capacitor.
Serious props to TSMC for having the processes in place to catch this or figure it out, sacking the alleged bastards, announcing the insider breach publicly (and accepting the consequences of a moment of corporate vulnerability, but at the same time showing their transparency and commitment to protecting client IP), and further not allowing mere job loss to be the end of the story. I had no particular opinion about TSMC, but my respect for them has moved up a notch now.
Could someone help contextualize what parts of this manufacturing process are considered the most important and closely guarded trade secrets? I'd love to hear some slightly more concrete examples.
How easily could another company replicate this process if they knew all the key details? It was my understanding that access to photolithography machines was one of the major obstacles in replicating chip manufacturing processes.
> Could someone help contextualize what parts of this manufacturing process are considered the most important and closely guarded trade secrets?
There's way too many fine details to even begin to list. Modern chip fabbing is the closest thing on the planet to actual dark magic, and the difficulty only rises exponentially with every new fabrication node. Literally any part of this could be considered a "trade secret" if it's not already described publicly as part of patents, academic research or both.
It's nice to see TSMC's internal security teams are detecting these things, but it would be more surprising news if this kind of IP theft wasn't happening to be honest...
My probably-racist sounding comment or nationalistic is my concern for software being owned by a certain entity and then this group is going to be used for the foundation of AI with our military, it will be interesting to see how that turns out.
edit: in this case someone pointed out it's a different nation so I'm surprised
I meant that plugging in a thumb drive and downloading the company jewels is sure to be noticed. I was hoping for something a tad more clandestine. USB storage devices on my corporate laptop are auto mounted as read only and need policy exceptions to be able to write.
This was not just from a thumb drive as that is very tightly controlled at TSMC (they did catch someone right away before it got leaked). The employees were caught printing info and removing it from the company (caught due to magnetic ink setting off the metal detectors) as well as using phones to take pictures of their laptop info while connecting from offsite. Taking pictures of remote laptops is a more covert way, but both employees were caught through suspicious pattern analysis and review of access logs of people right before they quit the company.
I always wondered - why people use a nation-state notion? Why not just a state or a country? USA is not a nation-state, or UK, or India, or Russia, etc. Does your question imply that none of these countries are capable of orchestrating this and only nation-states can do it?
> I always wondered - why people use a nation-state notion?
It's a shibboleth, similar to HN favorites: "orders of magnitude" and "orthogonal". Shibboleths are often social signifiers, sociology has lots of research on why people use them
New Jersey is a state. I do not think that they have the wherewithal to attempt such. Scotland is a country, but I don't think their occupiers would allow them to try. Granted, it is an awkward term, and I too wish there was a better one.
>Does your question imply that none of these countries are capable of orchestrating this and only nation-states can do it?
It's not clear to me that Russia can summon the ambition. But point conceded.
Well, is China a nation-state or a multi-national state, or essentially just a country(state)? English is my third language, so I just wonder do I miss some nuance here.
They said [1] in article but put iPhone 18 chip process just for clickbait… Disappointed.
[1] No details have been shared on the nature of the information obtained. It is likely that it relates to the 2nm process in general rather than anything specific to Apple’s A20 chip.
Hard to feel bad for Apple when a lot of this is about locking customers out of their own devices and then cheaping out with contract manufacturing. I almost want to cheer on the employees doing this.
Right before I left Samsung Austin Semiconductor (Samsung’s fab in the U.S.), in 2019, they were phasing out local share drives in favor of a self-hosted cloud that Samsung created. The supposed reason was better security, though it’s unclear to me why they couldn’t globally apply whatever rules they wanted to enforce to all office locations, instead of forcing everyone to use a remote endpoint. The throughput was absolutely terrible, like < 1 MBps. My department had some large files, so that was fun.
One such file was an Excel file that was more script than anything else. We had to have labels in a specific format on every machine we owned in the fab, which was something like 250 of them. The normal stuff like its id number, and also which points of contact for a technician and engineer, as well as their photos and phone numbers. Manually balancing and re-balancing every time a shift gained or lost an employee would’ve been obnoxious, so naturally instead countless hours were spent coercing VBA and ODBC to query a DB containing employee info, extracting and resizing their headshot, applying all of this to a template label, queuing a print job, and repeating. It was pretty fun to watch, honestly. I think I also had created a floor plan map somehow, and it would do its best to group a given technician’s assigned machines such that they minimized distance traveled during inspections. Anyway, the large file size was due to it caching the headshots (might have made a hidden tab for each? I don’t remember) for better performance, as that had proven to be a bottleneck.
This abuse of Excel might be in the running for a new form of esports.
It's always worth remembering there's exactly one reason Excel is abused -- IT isn't giving "non-developers" access to tools they need to automate their work.
I can only imagine the thousands of lines of code I saved by using Excel '97 for a scientific tabular calculation effort alone.
Because I had already done about the same thing in HP basic before PC's existed, when there wasn't a spreadsheet app anyway.
And it was on my mind the whole time.
It didn't take many more years before clients wanted their results in XLS on one tab and the invoice on another tab. Rather than the DOC files they had been wanting, which were reasonable facsimiles of the top linen typewritten paperwork I used to fax then mail by USPS or Fedex at the beginning. With the original invoice attached using a "gold" paperclip, nothing ever stapled or folded.
Oh how low I have sunk since then :\
Anyway it was pretty easy to switch from Word to Excel using the same fonts and similar formatting, even though direct integration was provided for, I just switched. When clients printed it they looked almost indistinguishable from my old typewritten multi-font documents.
And that just happened to put the paperwork onto the same spreadsheet as the data. Left one blank tab for future AI use, and the two tabs to send to the clients are in place to the right of that.
Eventually there were quite a few tabs in between the blank placeholder and the final two tabs that the clients would receive. Each with its own function to further leverage built-in Excel capabilities and make further progress toward a "paperless" office.
Saved gobs of code there too, but it gets "worse" from here ;)
It all came together by writing more VBA modules until there was an automation "shell" around the working system. Once complete it would stop so you could look at the data, decide what to type into the blank tab ("naturally", in the absence of AI) then hit the button.
And you were done.
while I don't doubt that such situations also exist, that wasn't the reason for any of the many "Excel abusers" I've encountered in different positions. Quite to the contrary, these people all had access to the appropriate tools, but their whole thinking was totally formatted and fixated on Excel as their go-to tool for everything:
be it things better done with a database, a word processor, a diagramming program, a label generator, a Form editor, a markup language, a web page, anything: they had all the tools at their disposal but no, no, they felt the odd compulsive need to do it with only Excel instead…
…often leading to problems down the line when the limitations of Excel for the use case (for which it wasn't made) would show more and more but they wasted already so much time and (needless) effort doing it in Excel that they would be even more reluctant to the possibility of switching to any more appropriate tool for the task.
Another reason I did so (at a different job, actually, the one immediately prior to Samsung) was that the only other language I knew to any degree was Python, and while I did compile one Python program for Windows to more easily distribute it, Excel Just Worked.
That one estimated nominal and peak loading on electrical distribution circuits by querying a DB for consumption on a given branch, and comparing it to a table containing current handling capabilities of various kinds of ACSR (overhead power lines). It was easier to do that in Excel because people were already used to it for various calculations.
Also if you’re wondering, yes, we had some bespoke software whose name escapes me that was specifically created to model electrical circuits; unfortunately no one knew how to use it, but one guy thought he did, and convinced management that we didn’t need to pay for training because he would train everyone else. He did not.
In hotel industry they need to calculate and set a price for each room/night. Typically that's done by solving convex optimization (e.g. simplex method) and using "shadow values" from it as the price per room.
Lo and behold, not every math package provides "shadow values". So you either buy a specialized math tool, or... use Excel, that has it built-in.
The VBA hammer is real... it can force any screw, nail, staple into any hole you desire.
Until the web version takes over, and you can no longer connect to anything real.
I doubt that the people who use Excel care enough about computers to be deterred by cloud.
The cloud version of Office365 Excel doesn't use/support VBA iirc. It also won't support, for example, connecting directly to a database.
Last night enjoyed reading the HN topic
"Lessons from writing a Kubernetes Security book" at
https://news.ycombinator.com/item?id=44784495
all about the joys and problems developers encounter using kubernetes/k8.
And reading now, today here about Excel/VBA, it struck me that something was missing: couldn't I eliminate 90% of computer software and ease 80% of software devops simply by building/selling a system that ran purely Excel atop kubernetes?
I mean, isn't that all there is to computing nowadays? Plus a little social media? And some AI for fun and maybe a little Excel/k8 work as frosting on the cake.
EDIT: Upon some further thought and prompting by helpful anonymous others, I have found a slightly different path: a FORTH/kubernetes system that will be used to bootstrap versions of VBA & Excel to the cloud. This will be made available to the great unwashed masses of humanity (and their AI friends) for implementing social media for all.
Largely, yes. Another department built a Python script that mostly was a driver for Selenium, because while they weren’t allowed direct access to some DB, nothing stopped them from automating a script to click various buttons in the shitty IE6 web app and scrape the results. I also recall they found that while passwords for said web app needed to be changed every N days, there was no lookback, so they had the script rotate between two passwords.
In my case, another not-insignificant factor was that multiple people on the team were quite good at abusing Excel and VBA, but only two others knew any other programming languages. One of them was an engineer (not SWE, engineer as in fab tool engineer - I think he was an EE) who occasionally wrote stuff in C for fun, and the other was also an engineer, but one who had convinced management that we should create an in-house SWE team. Great idea, and we weren’t the first to do so; unfortunately our manager was abruptly replaced, and his successor saw no value in the proposition.
Bingo. Working in an automotive OEM the entire company works on excel specifically excel vb scripts. No one knows why and no one has figured out how we got here. But it’s slow and steady abuse excel.
How do you do fellow Germans
Also how the simplest, fastest and most secure way to send someone from another corporation a file for collaboration is an encrypted zip and phone in the password.
Roko's Basilisk or whatever it is should more properly be called Excel's Visual Basicilisk - once Excel becomes self-aware it is going to punish everyone who tortured it for decades.
Maybe we need an IOCCC or Demoscene for Excel
https://excel-esports.com/
https://youtu.be/QwNoFOUiSiE
It just sounds like Mail Merge to me. Which if you were using a computer for work in the 90s, you might know something about. (It was also convoluted) :)
> it’s unclear to me why they couldn’t globally apply whatever rules they wanted to enforce to all office locations, instead of forcing everyone to use a remote endpoint
My guess is that they're worried that you'll download data and then copy it out of the device while the device is offline. An employee could even "lose" the device, giving an attacker unbound time to extract data from it.
Another equally likely explanation is that the exec in charge of their cloud services gains more prestige due to his solution being universally adopted internally, or some other crap along this line.
It is great that particular tools enable employees automate stuff and make their work more effective.
From developer point of view I see that the effort would most certainly be diverted in another kind of solution.
But yeah, "citizen developer" stuff is a thing that microsoft pushes especially in Power Platform / Canvas Apps - one programs with WYSIWIG and Excel-like formulas (PowerFX)
But then again I wonder who are the people that can program in VBA and chooses excel. Is it the constraint around software they can use? An excel being a GUI which you don't have to implement? Anyways, a net positive for business.
I'm sure the free GUI is the gateway drug, but at this point practically everybody in finance uses Excel. You can pass files and scripts around and be pretty confident that the external auditor will be able to use them.
where else can you run vba these days?
This is a pathology of megacorps who don't continually reinvest in process and support infrastructure and instead abuse backoffice desktop software inefficiently. In 20/20 retrospect, it should've been converted from Excel to a proper app as soon as it became a bottleneck.
Nice ha. Your story reminds me of why I flinch whenever I hear "just give them the spreadsheet, engineer, don't argue..."
A common and innovative solution to this problem now, that I have seen in other fabs (not semis, but other industries), is to put QR codes on each machine. That way the info behind it can be dynamic and maintained
At least they didn’t make you use Sharepoint and OneDrive. How fun using VBA and ODBC in 2019, proving the mighty Excel will go on as the new MS Access.
The original link from Nikkei Asia that the 9to5mac article is a repost of has some more information and less generic filler:
Link: https://asia.nikkei.com/business/technology/tsmc-fires-worke...
Archive: https://archive.ph/ta1kq
Industrial espionage has to be totally insane to defend against these days.
Just knowing how the thing is built doesn't seem to be enough. Comac still sources its jet engines from Pratt and Whitney for instance, despite many years of trying no local manufacturer has been able to build them to the same spec.
The product I oversee at my job is something that can only be built by people who are intimately aware of the process and have a strong understanding of the underlying engineering.
We could hand the full project file to a competitor and they almost certainly would not be able to build functional units. The failure points are fractal, so you need a strong intuition about what part you are installing, what qualities an ideal part has, what qualities the one in your hand has, how you might install it differently because of those qualities, and/or how you might change a later process to accommodate it. Or if the part should just be junked. The process is fraught with seemingly good intuitions that will ultimately lead you to failure as well.
These units also cannot be reworked, reused, or repaired, so any mistake before finalizing the build junks the entire thing.
For extremely low-entropy products, mother nature is incredibly unforgiving.
This is a bigger issue than most people appreciate, and a huge problem for the USA.
There is a specialised trade known as tool and die maker, or just die maker (https://en.m.wikipedia.org/wiki/Tool_and_die_maker) that is fundamental to a country’s ability to create industrial capacity. So far, no automation tool has been able to replace their expertise.
Without die makers, you can’t build tools that make things, you can’t build factories, and mass production in general is dependent on their skillset.
Right now, the USA only has 50k die makers, and the average age of a die maker, including apprentices, is 54. The average age of a master die maker is 73, and the average age of a journeyman is 62. A master die maker can teach about 5 apprentices to the journey man level in a work environment, after 2+ years of basic engineering school, in about 5-7 years. A journeyman may generally considered a master after 10-20 years of experience, depending on the nature of their experience.
We don’t have enough die makers to rebuild the industrial capacity of the USA, and we can’t teach the amount we need in less than 5-7decades without some kind of major change in the process of doing so. And since more than half of the master die makers are months from retirement or death, we are in an extremely precarious position as a viable industrial power.
This is why it is extremely difficult to build anything physical in the USA using only USA sourced parts and materials. It’s almost impossible to even get a decent variety of screws and fasteners made here in the USA, and we can’t easy build the machines to make screws because of the critical shortage of master die makers.
If we are to maintain the ability to build and maintain our machines, weapons, and critical infrastructure without being completely dependent on imported tools, supplies, and knowledge, we will have to reinvent the industrial process using automation or something similar to compensate for our foolish exportation and devaluation of strategic skills and capabilities.
Thank you for taking the time to post this. You've outlined problems that are absolutely solvable.
If the average age is 73 that's the issue right there. People are refusing to retire at 62 and make room for juniors. Then when they finally do retire (73 average seems like "dies in office" levels) there are 2 decades of missing juniors who haven't been trained in the pipeline.
This is the shortcoming with most technical fields. No one is incentivized to see the big picture of the training pipeline that exists well outside the scope of their own company. No one likes juniors but that is their future.
The actual killer is the pay and working conditions, and the problem is fairly intractable.
If you are smart enough to be a good tool maker, you are likely smart enough to be a good 6-figure keyboard-all-day worker. Losing a finger (or three) and breathing VOCs all day for half the pay is not very enticing.
These industries aren't glamourous for investors either. The business proposition sucks, the cost and liabilities are intense, and the margins would need to be negative to be truly competitive.
And worse than anything, the stuff that comes from China is not only 1/10th the cost, it's also now better quality.
At this point, even if die makers started making 300k a year, there isn’t enough master die makers left in the USA to bootstrap the capacity in less than 5 decades. The only practical solution at this point is to max out apprenticeship and somehow incentivise the industry so that we could flood other countries with apprentices as well. Then, we might be able to start getting things balanced in as little as 2 decades.
Otherwise, completely homegrown manufacturing is essentially dead in the water.
If we want to regrow US domestic manufacturing, we need to throw about 100B at scholarships and incentives, or figure out how to capture remaining diemaker empirical knowledge into ML / robotics.
Or maybe we could spend 10B on immigration incentives for qualified Diemakers? Idk about the global state of affairs and how that would work. At any rate we need at least 100-300k new diemakers if we really want to rebuild 100 percent domestic capabilities overnight. And then we need a career path for them 20 years farther down the road.
It was really, really foolish to allow strategic industrial capability to wither on the vine.
The problem is that these jobs are hard and the pay is crap.
SmarterEveryDay attempted to make a grill scrubber in the US. https://www.youtube.com/watch?v=3ZTGwcHQfLY
tl;dr He did it, but it went very poorly.
Yes. Realistically speaking the value that a diemaker brings should easily be worth 2x or more their current scale.
The SmarterEveryDay piece was disheartening for sure, echoes what I have been singing from the rooftops ever since the government dropped the USA origin of manufacturing requirements for all purchases. Sure, it was a kind of corporate welfare, and it made for a tricky incentive misalignment, but it kept a lot of strategic industries on life support. We would be in a way, way better position now if it hadn’t ended.
As far as I understand you still have to buy US, and need approval if you cannot source domestically.
All our military work still uses US electronics shops and machine shops.
The window has shifted as to what constitutes “us made” though. For most things “US assembled” is the norm, and when you get out into things like office supplies, vehicles, and construction materials it’s often far from true domestic manufacturing. For strategic military stuff it’s better, but often at the cost of using 1990s tech because we don’t make the newer stuff here. (It also has the effect of making things more repairable, but that argument gets weaker and weaker as time goes on)
At this point I think in practice it’s a preference for US manufacturers but a long way from robust enough to establish domestic supply chains.
Also -a lot- of military purchases are local, discretionary on a credit card, and there is no oversight on the origin there- it’s whatever they need from office max or the hardware store.
That may be the case in many industries, but I don’t think it applies in this case. The industry isn’t flooded with people looking to move up, it’s starving for interest. The problem is that you can farm out the work overseas for much less than you can do it locally, because our automation is decades behind China.
Chinese skilled labor wages aren’t that low anymore. The reason they can do most work for 1/10 of the local cost is advanced automation, government VC investment, and somewhat less c-suite greed. The incentive alignment of the government VC is also distinct from typical US VC.
We can’t even build the machines that would enable that kind of automation in the USA. We’d need to buy them from China, IF China was willing to sell them to us. Unlike the USA, China is probably not going to be stupid about diluting their advantages. I’m sure the CCP hasn’t forgotten the decades where we were unwilling to sell them high precision machine tools lol.
Unfortunately, we are more likely to see a slow deterioration of the US economy and infrastructure, until we hit about a 5-6x wage advantage over China which we could potentially use to copy the Chinese playbook and rebuild our presence.
I really hope I’m completely wrong lol.
That's not a high bar... Boeing can't seem to reliably manufacture their own designs.
What product is it?
A relative of mine worked at a medical devices company (brain sensors). She told me how small intricacies of the manufacturing process were critical to reach good enough yield or functioning devices, at all. The critical process steps were closely guarded and only a handful employees knew how to do them. The devil is often in the details - and the moat, too.
> Just knowing how the thing is built doesn't seem to be enough.
See perhaps:
> Tacit knowledge or implicit knowledge is knowledge that is difficult to extract or articulate—as opposed to conceptualized, formalized, codified, or explicit knowledge—and is therefore more difficult to convey to others through verbalization or writing. Examples of this include individual wisdom, experience, insight, motor skill, and intuition.[1] An example of "explicit" information that can be recorded, conveyed, and understood by the recipient is the knowledge that London is in the United Kingdom. Speaking a language, riding a bicycle, kneading dough, playing an instrument, or designing and operating sophisticated machinery, on the other hand, all require a variety of knowledge that is difficult or impossible to transfer to other people and is not always known "explicitly," even by skilled practitioners.
* https://en.wikipedia.org/wiki/Tacit_knowledge
Building even basic things is difficult. Friend of mine tried like dozen companies from around the world to manufacture very basic milled aluminium part. None of them could make it. Only one company could make 1 or 2 right out of 100.
Sorry, but I find that hard to believe. Part certainly wasn't very basic.
Slab of aluminium milled to a flat triangular-ish shape about 10mm thick, then inside hollowed out to half depth and 8 counter-sunk holes non-tapped. Common issues: - item is not flat - nicks and dents on the edges - non uniform depth of the counter-sunk holes - holes in incorrect locations - unapproved changes to design - scratches - pores (poor material quality) - incomplete milling
Then parts have to be hard anodised and sand blasted.
- Sand blasting uneven - Sharp edges remaining - Streaks - Contamination - Exposed bare metal (it is only acceptable in the holes that could be used for hanging)
Companies also have virtually non existent QA. You can list them things that will cause rejection and they send it anyway.
One company told him after sending a batch that was complete carnage "We can give you $100 coupon for next order!" he insisted on rework. So they sent again, even worse.
I can't remember but I think Lance Air or Epic got split in a sale and a company bought the type design/blueprints but ran into issues actually manufacturing from them.
Or for analogy, following a recipe doesn't make you a good cook.
> Or for analogy, following a recipe doesn't make you a good cook.
For an analogy, no one cares if you're a good cook if you're able to make a passable burger. Most of the demand is not for the best burger money can buy, the just want a burger.
Or for analogy, following a recipe doesn't make you a good cook.
Following a recipe can you close enough for thousands of Door Dash customers to put the original restaurant out of business.
I feel like you're reading too deeply into my shallow analogy.
I'm not sure - what if China can flood the world with 80%-as-good-as-the-real-thing knockoffs at half the price and put everyone else out of business?
But it also depends on how precise the recipe is - if it's described down to the exact movements the cook needs to do, which may be replicated via a machine...
No recipe accounts for ambient temperature/humidity, very few for altitude differences, etc. etc. It still takes knowledgable tweaks to get just right.
Such recipe could exist.
At the end of the day, the map won't ever be the territory. Typically the properties that makes something successful are those which cannot be specified. If it were otherwise and those qualities could be specified in a reproducible manner, that thing would not be observably successful.
They exist in the industry, where a small variation of one parameter could affect tons of products. Same for individual restaurants, cooks know how to set up their equipment. For amateur cooks at home it does not matter. They do not need that level of consistency, and it would get quite expensive.
Are P&W and Rolls Royce the only companies in the world capable of manufacturing high-end jet engines end-to-end?
No, but they provide the best efficiency/performance for the buck. China can produce its own jet turbines, but they have to trade off performance or longevity to do it.
Interesting. Thanks.
They don't make them end to end either. Their jet engines are made up of parts supplied by thousands of suppliers.
Obviously, but I would imagine that most of the “magic” isn’t in the supplied components, but in the finished product. Otherwise reproduction would be easier.
Don't forget GE Aerospace. It gets a bit weirder too since you have joint ventures like CFM and Engine Alliance.
Good point. Isn’t there also Safran?
Yes. Their engines are mostly for military applications, tough, so they are less well known of the general public. Other than that, they are half of CFM International, but again they are not very visible.
Which Comac? I thought they all used GE (CFM for Comac 919) or Russian/Chinese sourced engines.
That's the OP's point: COMAC is using CFM LEAP 1-C engines on the C919.
To be fair, they have taken the effort to build the CJ 1000A engine - which is on wing testing should the tangerine fellow cut them off. But its Plan B at best.
We don’t have sensors that can grok the full building process that deep human experts have.
I frequently wonder what steps SpaceX security has to take, given the insane geopolitical significance of reusable rockets and cheap access to orbit.
There is a very nice chapter in the somewhat dated but classic book Business Adventures [1] on trade-secrets and what happens when employees of one company move to another. In chapter 11, "A man, his knowledge, and his job", there is a story of a "space-suit" manufacturer Goodrich suing an employee for moving to its rival Latex for stealing trade-secrets. The story is timely in context of Meta hiring researchers from open-ai, deepmind etc for 100s of millions for the knowledge in their heads of the recipes which work for making superior LLMs - the knowledge of which is empirical and may take years to discover.
[1] https://www.amazon.com/Business-Adventures-Twelve-Classic-St...
You could also argue that they're hiring people for the general expertise they have in developing frontier LLMs. Distinguishing trade secrets from general knowledge is difficult and employees building the latter to make them more valuable to new employers is an explicit policy goal of US employment law.
What is the story?
Lots of space technology is classified as weapons subject to export control. ITAR has plenty of rules about who can see information. US immigration status generally has to be green card or citizen, and country of origin and any second citizenships are considered.
("export control" in this sense really doesn't have to do with moving a physical object out of the country but sharing information, to the extent that a conversation in an elevator could be an export violation. most export violations amount to emails being sent to the wrong person)
When I worked briefly in defense, for example, there would be regular random searches of my stuff as I exited the building and security would wander the building and look at what you left out on your desk while you went to lunch. Entirely seriously they told us not to wear our badge in public if we left the building and not to leave our laptops in our cars because someone might follow us and steal it. Had colleagues who were visiting a foreign country for work have their hotel rooms obviously thoroughly and messily searched while they were out.
They also do national security missions so there are folks there with high clearances.
Thing is that even if you did steal a bunch of information, that doesn't mean you could just copy and be successful. Any one of a million things can go wrong with a self-landing rocket that will cause it to explode, you can't just steal the whole system of operation that keeps these things from happening.
You couldn't steal all of the secrets of a circus performer and suddenly be able to juggle chainsaws while riding a unicycle.
I would say the main element of their success is not really in any specific close kept secrets - its in actually committing to reusable rockets and keep working until they had a working system.
While there are some really nice components and clever ideas (Merlin/Raptor engines & very good guidance tech) this all really has been doable for decades in less efficient form.
But so far no one other than Space X has been able to win against all the naysayers who were so sure only single use rockets are ever going to work, get enough funding to build a partial RLV & then operate it successfully as a business.
I don't think it depends on any single technology or a set of them only they have access to - rather that they have been able to persist and see it through, unlike all the other RLV projects that never got funding to go past the paper stage or very simply not viable (Space shuttle).
So, this is admittedly a little tinfoil, but I wouldn't be surprised if Musk is happy for some degree of espionage to happen. If it looked like there was a possibility of China getting this capability first, it would light a fire under the US government to give financial and regulatory assistance to the Starship program.
https://www.patsnap.com/resources/blog/elon-musk-claims-pate...
Musk does seem to think in terms of how much money he can get from the government for his companies. But to be fair, government subsidies are a successful strategy for entrepreneurs who want to make a lot of money.
Maybe they shouldn't be? And I think honest people can have that debate.
But you can't really argue against the effectiveness of government subsidy as a path to prosperity for the guy getting the money.
Despite his faults he does seem to care about the actual lofty mission statements of his companies more than just straight profit. Otherwise, you wouldn't see bizarre things like him directing Tesla to open source all their patents.
It’s primarily in the government interest in subsidising, except when it’s a corrupted subsidy.
This worked for at least one Burisma employee
I imagine SpaceX having pretty fishbowl conference rooms for customers in the center of everything, with guest network ports just segmented off from the main but using the same hardware. Oh man that would stress me out if I was IT. And of course the '<Customer name> needs to print something off and needs access to the MFCs'. No, you print it out for them, like has been discussed and agreed to and keeps with ITAR. 'No, they need access, and now. Because I'm a sales guy and I won't tell them no' but if they get into/past the MFC, it's all on IT and IT being bad at their job/security, not the sales guy that demanded they get physical access.
Spacex rocketry tech is subject to ITAR regulations. That restricts who they’re allowed to contract with, data encryption and handling, but altogether those regulations are quite bare. It likely wouldn’t be enough to stop a state actor or rogue employees.
I think ITAR is mostly just to stop the outright sale of controlled items to foreign entities, not necessarily to prevent IP theft or corporate espionage.
I worked on an ITAR-controlled camera once and it was drilled into me than even allowing a non-US person to view the output images of the camera constituted an ITAR violation.
Right, which is all well and good if you're a benevolent actor, but ITAR really doesn't do anything at all to stop you from intentionally committing (corporate or actual) espionage if you want to. There should still be controls to prevent you from violating ITAR either accidentally or on purpose.
The controls are you go to jail if you get caught. At least this is what one is told in the mandatory ITAR compliance trainings. That seems to be how the law works for most crimes.
It's an incentive but it's certainly not a control. Controls stop something from happening, they don't respond to it after the fact. Controls would be things like security checkpoints, network monitoring, random searches, that sort of thing.
"You will go to jail if you kill someone" isn't a control against murder.
[dead]
I don't think what SpaceX is doing is that hard to replicate. There's already competitors launching smaller payloads for smaller costs per weight. Just a matter of time until they creep into SpaceX's market, while SpaceX tries to build a starship inspired by the Futurama rocket.
It is insanely valuable, both commercially and strategically.
If it weren’t that hard to replicate, several countries (and Bezos/Blue Origin) would have replicated it by now.
I think you vastly underestimate how difficult rocketry is. There’s a reason “rocket science” is colloquially a metaphor for an extremely difficult and technical task.
>If it weren’t that hard to replicate, several countries (and Bezos/Blue Origin) would have replicated it by now.
There is a 100% chance multiple countries/companies will have replicated it in the next decade. If SpaceX never existed, they likely would have achieved it at the same pace regardless.
This is the same with EVs. If Tesla never rose, the world EV market outside of Tesla would have seen precisely the same rise.
There is a tendency to attribute the early movers with innovation in the inevitable, where we all stand on the shoulders of others and just reach a little higher.
As to the rocket science misnomer, that's a space race hangover where an engineering role was extremely public and celebrated, but in actual reality "rocket science" is a mediocre field with miserable pay and high unemployment.
As to how valuable it is, "insanely"? The world has a fairly finite launch need, such that SpaceX made a whole new business -- Starlink -- to make work for their capacity. Economically the space launch business is relatively minuscule.
I question your EV take. Tesla proved a business model, the technology and path from niche sports car to the best selling car on earth, and now on to the lowest cost per mile robotaxi. Simply knowing that a solution exists and is financially viable, is enough to motivate the competition.
>Tesla proved a business model
Through almost all of Tesla's existence, its business model was ironically the sale of gasoline vehicles. Because, of course, Tesla's entire business model relied upon selling green credits to incumbent ICE vehicle makers.
So it didn't really prove much of a business model, effectively being parasitic.
>the technology and path from niche sports car to the best selling car on earth,
The overwhelming bulk of the technology advancements that enable modern EVs -- from advances in batteries to cameras to sensors to embedded controllers and CPUs -- is thanks to the smartphone industry. Modern EVs owe infinitely more to those than they do to anything Tesla did.
>knowing that a solution exists and is financially viable, is enough to motivate the competition.
I think of this much like compact fluorescents. Remember those? We all rushed to transition, and then they were absolutely demolished in every metric -- efficiency, colour, and most importantly the amount of environmental contamination when disposed -- by LED lights. I feel like we're going to feel the same about early EVs.
>now on to the lowest cost per mile robotaxi
Is this a serious comment?
>now on to the lowest cost per mile robotaxi Is this a serious comment?
Absolutely a serious comment and why Waymo will likely be unable to compete. Tesla is designing their fleet for the lowest cost per mile, and likely the real reason why they have pushed back on Lidar versus camera - humans can drive with binocular (some with just monocular) vision, why use Lidar as it’s more expensive.
If a Waymo car costs $100k and is good for 100k miles, then that’s $1/mile. Tesla’s robotaxi is likely designed for 500k miles at under $0.10/mile. Tesla’s robotaxi will be able to drive me for cheaper than I’ll be able to drive myself. Think of a cab company with scale enough to build their own cabs, optimizing for the lowest cost per mile, and with $0 labor costs. That’s going to be their huge win. And no, I’m not an Elon fan, but I do understand his business plan, and it’s brilliant. And no, I don’t think it’s going to displace personal cars. There’s too much burst demand during rush hour for a peak scaled fleet to be cost effective (I haven’t run the numbers, so even that slight may be incorrect). However I do expect to never have to drive myself once I hit retirement age, and I’ll probably not even own the vehicle. My hope is that there will be a fast follower, able to compete against Tesla, otherwise I’ll be paying Tesla $0.99/mile instead of $0.12/mile.
> The overwhelming bulk of the technology advancements that enable modern EVs -- from advances in batteries to cameras to sensors to embedded controllers and CPUs -- is thanks to the smartphone industry.
Would you point me to a source where I can read about this?
If Tesla never rose, the world EV market outside of Tesla would have seen precisely the same rise.
would have seen the same rise _eventually_. I know from a friend that worked R&D at a major car company that Tesla really lit a fire under then and 'forced' them to push their own EV experiments from proof of concepts to commercial product much faster than they where originally thinking about doing it.
The EV ventures of most automakers are massive money losers (just as it always has been for Tesla outside of selling green credits and subsidies). But for sure they all rushed to get there not because the EVs themselves were valuable, but because of the insanity of the capital markets where Tesla is valued at a trillion dollars at a 200x P/E, while the rest of the market is at like a 7-14x ratio. Everyone wanted some of that irrational hype.
The problem lies in forming and managing such a huge organisation that deals with the problem in an efficient and lean way, not the technical aspects.
The materials science aspect is a challenge, not to produce, but to produce with a sane cost.
The rocket science aspect of things (namely the linearisation of the booster model in order to be able to be solved in constant time by an MPC) is more or less a solved problem.
Coordinating such complex interconnected systems will always remain one.
so it still "rocket science", you saying that the hard part is to make it lean and efficient
isn't that what they do in SpaceX????, your comment literally counterproductive
It is comparably hard to creating a lean and efficient car manufacturing company.
Yet you don't hear people throw the phrase "automotive science" around.
Rocket science has some niche parts that are very hard indeed. But this is not where the major challenge lies.
SpaceX itself replicated the DC-X from the 90's. The reason the DC-X was cancelled was because of the economics. Reusable rockets are a solved problem, with only the economics of it a barrier (see Space Shuttle). SpaceX has to rely on their own investor funding for Starlink to remain a viable entity.
The DC-X was stupid, dumb, insane, bunkers, and a waste of money and not replicable.
"sure, let's put four RL-10 hydrogen engines on it! They are expensive and the worst possible for the low-altitude flights we are going to do so some sucker will believe we can do multi-stage reusable flights to orbit".
"sure, we don't actually need to go to orbit anyway. That was always a dumb idea. Who said we ever wanted to do that? No, suborbital is really useful, we promise. And we are going to do it with a single stage (using hydrogen) cuz we are so smart and the future and everything!"
"sure, let's build a specialized hydrogen tank in a stupid shape."
"sure, let's give the whole single-stage low-altitude rocket a k00l shape that makes it more expensive."
"People knowing it can't be done shouldn't interrupt people doing it wrong."
This comment has big "I could build a better SpaceX I just don't want to" vibes.
If only my dad had an emerald mine then I'd be the billionaire rocket scientist CEO
https://www.snopes.com/news/2022/11/17/elon-musk-emerald-min...
TLDR: It's nonsense.
Funny how many words it takes to say Elon’s dad in fact own a stake in an emerald mine, invest in Elon’s business, and give Elon an incredibly privileged upbringing.
There's a lot of unanswered questions from the article which is seems anyone with even mediocre skills as a journalist should be willing and able to answer.
It mentions the mine "collapsed" in 1989. Does that mean literally? Just financially? Was there an insurance payout? Did everyone lose their investment? Did Errol Musk own 1% or 90%?
* In quoting another article, it does say: "Errol Musk, an engineer, owned a small percentage of an emerald mine and had a couple of good years before the mine went bust and wiped out his investment."
Elon graduated from college less than 10 years later but says he was $100k in debt. Was he actually in debt? Did he spend years paying that off or did he (or someone else) pay it off shortly after in a lump sum or very quickly?
His dad provided $20k of a $200k seed round for Zip2, was the rest also from friends and family or more institutional investors? Did his dad receive equity for that or was it a gift?
Reading through the article again (and a little more closely than before) it doesn't seem Elon had "an incredibly privileged upbringing" but maybe that's a mix of good PR and this now being 40-some years ago? They're referred to as upper middle class but if that's all it takes to be "incredibly privileged" then 90% of the kids born to people reading HN are also incredibly privileged.
Arnold's piece seems to contradict basically everything else in the article so I'm not sure what to make of it, but it also sounds like he made about $350k profit over the lifetime of his ownership in the mine? Certainly not nothing but it's not opulence and it doesn't sound like most (any?) of that made its way to Elon.
> Elon mostly lived with his father, who says he owned thoroughbred horses, a yacht, several houses and a Cessna. One of their homes was in Waterkloof, a leafy suburb of Pretoria that was popular with foreign diplomats.
> Wanderlust ran on both sides of the family. On holidays, Errol and his kids would travel, he said: to Europe, Hong Kong, throughout the United States. Or they'd take the plane to Lake Tanganyika [in Zambia], where Errol had a stake in an emerald mine.
That sounds more privileged than upper middle class to me.
It seems to me companies ... don't care out side some easy to do basic things.
But when push comes to shove if manufacturing is cheaper in a country where lots of folks want to steal your things. -shrug- Short term profits win.
am i crazy for not caring if a company in a foreign country obtains trade secrets and manufactures the same thing? like we're all humans and we all want access to whatever it is they're building, it seems like more people building the thing is a good thing. if that impacts Samsung's profits, why do I care? its not like corporations give a shit about me
Well who are you? Maybe there is no reason for you to care.
TSMC alone is 12% of Taiwanese exports. The entire semiconductor industry is 25% of Taiwan's GDP. It's obvious why the Taiwanese government and society, to say nothing of TSMC's shareholders, would care.
Yes, because if you tolerate that you disincentivize actually developing the IP in question in favour of stealing everyone else's, which leads to nothing being developed.
is it impossible for society to advance without IP laws? is capitalism a requirement for advancement?
No. Of course not. This is just something IP holders say because they profit.
Obviously, if this happened, Taiwan would not exist now.
Do non-absolutes exist?
Some countries have all the talent and manufacturing and sourcing advantages. Once they take the lead, you might never be able to keep up.
Your engineers lose their jobs, your businesses go bankrupt, you exit that entire field entirely for your entire population. Slowly your ability to do work begins to evaporate.
It's happened before and it'll continue to happen.
> Slowly your ability to do work begins to evaporate.
To be clear about this, you can still physically do the work same as before, it's just become uneconomic/not competitive. It's not all bad though. Having another country take the bulk of the market leaves you free to differentiate and specialize in one or more profitable niches that are not being served well by what is now the main supplier. You might end up dominating some fraction of what's now a vastly bigger market, with improved economies of scale that you're free to exploit as well.
That's a bit like an exercise left to the reader. Or wishful thinking.
We hope that it works out that way, but there's simply no guarantee. It's not an economic law.
If the country supplying said thing can have domestic companies deliver at lower margin, you're still kind of screwed. Their internal competition fills out the niches, which they can then export.
isnt that what tariffs are for?
there's something that rubs me wrong about the argument of "we need IP laws because otherwise someone else might do the same thing as us, but better, and we won't be able to compete"
I work at a tech company and I’d rather have other countries pay my company to do it, rather than do it themselves with less-than-US salaries
I really can't answer that, it's your call if you care or not.
I think that the more of a free for all folks stealing tech as they wish will push companies who do the development work towards more proprietary / DRM and similar solutions ... I don't like that.
Allowing people to profit from their inventions / investments encourages more such development, and without that discourages it or encourages less good options.
[flagged]
We are doing much better off with the status quo than TSMC firing everyone who isn't of the same religion, and even then, it's not hard for a spy to pretend to be <insert religion here>.
> We are doing much better off
I’m not making a value judgement. I’m making a prediction about material conditions.
> TSMC firing everyone who isn't of the same religion
I didn’t say that. But circles of a trust around important information.
> it's not hard for a spy to pretend to be <insert religion here>.
Yes it is.
Can you explain why? I have pretended to be multiple kinds of Christian with success despite being none of them.
Because religious vocabulary, etiquette, and clothing are learned from decades of involvement usually as a small child.
It’s kind of like saying you’re going to pretend to be a different class, but more particular.
> I have pretended to be multiple kinds of Christian with success
I don’t know what context you mean. But I guarantee others can tell you aren’t a member of the group even if Christianity aims to be open in inviting new people.
Another common situation is they recognize you have some connection but can tell you aren’t an active participant.
Harder to pretend when speaking with colleagues that are the real deal. Also very likely to "step out of line" accidentally over time as lot of time is spent with colleagues.
It's not hard to pretend to be anything for 5 minutes to someone who doesn't know anything about what you're doing.
Naively, I was assuming TSMC would still operate like TSMC, not the Taliban, and that they'd focus on doing their work, not religious performance, but I guess anything is possible!
Suggesting TSMC start operating like the Taliban to protect their trade secrets seems like an obviously bad idea, though.
But we don't live in a vacuum, so I guess it's all just opinion pieces.
What are we on about now? The taliban?
Please don’t be facetious
Okay, substitute Taliban for the Church of Scientology or the Church of Jesus Christ of Latter Day Saints.
It's a marginally less absurd proposal.
My point stands.
Proposing extreme religious nepotism as a good counter-measure to deal with espionage 1) would be minimally effective (see the IDF infiltrating Hamas and Iran regularly), and 2) is ridiculous.
> Proposing extreme religious nepotism as a good counter-measure
Please read the thread again.
This is not a prescription. This is a description of how high trust markets work.
See this classic economics paper “Jewish Diamond Merchants in New York”: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=349040
>even then, it's not hard for a spy to pretend to be <insert religion here>.
There are religious groups which you can only be born into. Some people believe that one such religious group wields the most power on our planet.
Is there a Druze world government out there that I’m not aware of? :) It’s absolutely possible to convert to Judaism, for reference, though the faith is non-proselytizing (not actively seeking converts) and the status of converts is disputed among the more extreme Orthodox groups (but what isn’t).
The religion installing wires above NY so they can cheat it?
> groups have to return to more traditional forms of trust (family, religion, creed, etc)
a relevant acronym is MICE (money, ideology, compromise/coercion, ego). Religion would fall under ideology.
See e.g. https://en.wikipedia.org/wiki/Counterintelligence
Money and coercion is what corporations have now and it doesn’t work very well because the incentive is to violate it as soon as a better offer comes along.
I’m saying it has to be deeper and more durable.
Compartmentalize your company. Make sure people from one compartment are unaware about things people from other compartments are doing. Don't transfer/hire across compartment boundaries.
This is such a bad idea.
Care to say why? Security always comes at a cost ...
You don't need to deliberately create silos in order to secure a project. We've had people working on top secret projects and while people may not have known the specifics the company was kept up to date on the progress and outcomes of the project. We brought people in from other teams after an extra vetting process and additional NDA specific to that project.
With industrial espionage, NDAs aren't going to cut it ...
For the vast vast majority of companies, their success comes down to culture and execution rather than "the tech secret". You can't really steal that.
I will grant you, the specific case of TSMC is definitely in the rarer case where there are true tech secrets.
Apple does this.
Most companies aren't Apple.
Apple is great because their hardware and software is so well integrated.
If they compartmentalize their hardware and software departments, we might as well split them into a hardware and a software company.
Grandpa worked heat treat at the alegany national forge, where they made stuff like the beams for the empire state building, periscope tubes, and the 16" guns for the biggest battle ships, each thing had to be tempered in a very specific way, and to deal with spying and espionage, the charts and instuments used on the shop floor were all deliberatly wrong,and the written instructions were also wrong leaving the person in charge to know how to convert the given instructions into what was actualy done through a secret method, not complicated, but essentialy impossible to reverse engineer. There is a story of soviet engineers who somehow were invited to tour the RR jet turbine factory, and were given shoes that had extra sticky soles they wore only for the tour, which ewere then used to anyalise the metal chips that get picked up from the shop floor..... never ends, expected.....even honored
Locally in Taiwan, there’s growing suspicion that the 2nm process technology was leaked to Japan’s Rapidus.
People are questioning whether the technology was leaked to Rapidus through Japanese equipment suppliers.
make no sense since Rapidus 2nm process is from IBM.
idk which shocking, Japanese company try to steal TSMC or IBM have 2nm process out of nowhere
IBM has kept researching in Albany. They license manufacturing technology to other parties even if they gave up trying to build their own fabrication facilities a decade ago.
"Out of nowhere" would be a stretch. IBM may not have mass-market volume anymore but they kept up the production chain to build mainframes (and I would guess other "specialty" products) for institutional customers.
[flagged]
IBM has been publicly working toward 2nm since at least 2021, though, so it doesn't seem like it's coming from nowhere. Unless this is a completely different 2nm development, here's a presser about it from 2021, titled "Introducing the world's first 2 nm node chip" (https://research.ibm.com/blog/2-nm-chip)
[flagged]
What’s going on big guy? Maybe take a break from commenting
I thought HN commenter would be smart people, turns out they just good at Tech and sucks at everything else
> since Rapidus 2nm process is from IBM
Unless it has great yields with 0 issues, there's always things to learn from. It's also possible the IBM process isn't what it seems and there's more to it.
We can make hypersonic missiles.
I still want to get a look at China's. Right down to the metallurgy.
This article was just the headline repeated in various forms with some generic filler
So strange
I love that there is the headline, an intro, a heading for a section about the iPhone 18, and then another heading titled "TSMC says employees tried to steal trade secrets", which is literally a word-for-word substring of the headline.
When there isn't any info but you have to turn in 400 words to your editor/get paid by the word...
There is literally no indication in the article that this has anything to do with apple other than them being a potential user of the TSMC 2nm process. Strange they tried to connect this story to apple.
Yep. It doesn’t detail what was stolen, how they were found, if they’re arrested right now, or who it is suspected they are working for. Useless article beyond the headline.
At least, it doesnt look like chatgptese.
Reminds me of the now-infamous "capacitor plague" [1] of 1999-2007 that keeps cropping up in electronics repair.
[1] https://en.wikipedia.org/wiki/Capacitor_plague
I guess you got downvoted because the connection is not obvious. The relevant paragraph of https://en.wikipedia.org/wiki/Capacitor_plague#Implications_... is
> [...] A materials scientist working for Rubycon in Japan left the company, taking the secret water-based electrolyte formula for Rubycon's ZA and ZL series capacitors, and began working for a Chinese company. The scientist then developed a copy of this electrolyte. Then, some staff members who defected from the Chinese company copied an incomplete version of the formula and began to market it to many of the aluminium electrolytic manufacturers in Taiwan [...]
Asianometry has an interesting video on the topic titled What Happened to the Capacitors in 2002? [1]
[1] https://youtu.be/rSpzAVpnXo4
TIL! The only time (knock on wood) I ever had a motherboard go bad was an MSI from around 2002, due to a blown capacitor. To their credit, MSI had great support at the time, and sent me a new one after I sent a photo of the blown capacitor.
Serious props to TSMC for having the processes in place to catch this or figure it out, sacking the alleged bastards, announcing the insider breach publicly (and accepting the consequences of a moment of corporate vulnerability, but at the same time showing their transparency and commitment to protecting client IP), and further not allowing mere job loss to be the end of the story. I had no particular opinion about TSMC, but my respect for them has moved up a notch now.
Could someone help contextualize what parts of this manufacturing process are considered the most important and closely guarded trade secrets? I'd love to hear some slightly more concrete examples.
How easily could another company replicate this process if they knew all the key details? It was my understanding that access to photolithography machines was one of the major obstacles in replicating chip manufacturing processes.
> Could someone help contextualize what parts of this manufacturing process are considered the most important and closely guarded trade secrets?
There's way too many fine details to even begin to list. Modern chip fabbing is the closest thing on the planet to actual dark magic, and the difficulty only rises exponentially with every new fabrication node. Literally any part of this could be considered a "trade secret" if it's not already described publicly as part of patents, academic research or both.
It's nice to see TSMC's internal security teams are detecting these things, but it would be more surprising news if this kind of IP theft wasn't happening to be honest...
Only the bad criminals get caught.
My probably-racist sounding comment or nationalistic is my concern for software being owned by a certain entity and then this group is going to be used for the foundation of AI with our military, it will be interesting to see how that turns out.
edit: in this case someone pointed out it's a different nation so I'm surprised
How big are all of the design files for a modern chip? 10MB? 100gig? Billions of transistors, but surely a lot of compressible redundancy.
Not that I think you can just plug in a thumb drive and download as you please, but just a sense of scale on how much data describes the design.
> Not that I think you can just plug in a thumb drive and download as you please
Why not? It's just data, and a thumb drive stores data. As long as it fits.
I meant that plugging in a thumb drive and downloading the company jewels is sure to be noticed. I was hoping for something a tad more clandestine. USB storage devices on my corporate laptop are auto mounted as read only and need policy exceptions to be able to write.
This was not just from a thumb drive as that is very tightly controlled at TSMC (they did catch someone right away before it got leaked). The employees were caught printing info and removing it from the company (caught due to magnetic ink setting off the metal detectors) as well as using phones to take pictures of their laptop info while connecting from offsite. Taking pictures of remote laptops is a more covert way, but both employees were caught through suspicious pattern analysis and review of access logs of people right before they quit the company.
TSMC practices compartmentalization. The entire recipe is never stored in one place.
Have we topped Micron?
https://www.freethink.com/the-changing-world-order/chinese-i...
I'm confused about the iPhone 18 part; I assume that it would just be one of their usual processes, possibly with a few tweaks.
The iPhone 18 part is for clicks...it really has nothing to do with the iPhone 18 and everything to do with TSMC's N2 process.
So even if someone steals the secret sauce, does that mean they can actually cook the same dish?
No, but if you want to cook the same dish it sure helps to have the secret sauce.
When you have secret sauce, the next thing to do is hiring the right cook, which is much easier to do.
Gee, I wonder which nation-state actor was orchestrating this...
They claim it was Japan.
More specifically, Rapidus.
I always wondered - why people use a nation-state notion? Why not just a state or a country? USA is not a nation-state, or UK, or India, or Russia, etc. Does your question imply that none of these countries are capable of orchestrating this and only nation-states can do it?
> I always wondered - why people use a nation-state notion?
It's a shibboleth, similar to HN favorites: "orders of magnitude" and "orthogonal". Shibboleths are often social signifiers, sociology has lots of research on why people use them
ha-ha, thanks, makes sense.
New Jersey is a state. I do not think that they have the wherewithal to attempt such. Scotland is a country, but I don't think their occupiers would allow them to try. Granted, it is an awkward term, and I too wish there was a better one.
>Does your question imply that none of these countries are capable of orchestrating this and only nation-states can do it?
It's not clear to me that Russia can summon the ambition. But point conceded.
Hard to know what OP meant, but I took it as an oblique reference to China.
Well, is China a nation-state or a multi-national state, or essentially just a country(state)? English is my third language, so I just wonder do I miss some nuance here.
Funny thing is that Apple doesn't know how to make an iPhone.
They said [1] in article but put iPhone 18 chip process just for clickbait… Disappointed.
[1] No details have been shared on the nature of the information obtained. It is likely that it relates to the 2nm process in general rather than anything specific to Apple’s A20 chip.
[dead]
Hard to feel bad for Apple when a lot of this is about locking customers out of their own devices and then cheaping out with contract manufacturing. I almost want to cheer on the employees doing this.