> Swift has turned into a gigantic super complicated bag of special cases, special syntax, special stuff...
That's true, but only partly true. It already was a gigantic super complicated bag of special cases right from the start.
Rob Rix noted the following 10 years ago:
Swift is a crescendo of special cases stopping just short of the general; the result is complexity in the semantics, complexity in the behaviour (i.e. bugs), and complexity in use (i.e. workarounds).
Apple's new Swift language has taken a page from the C++ and Java playbooks and made initialization a special case. Well, lots of special cases actually. The Swift book has 30 pages on initialization, and they aren't just illustration and explanation, they are dense with rules and special cases
Of course, that doesn't mean that it didn't get worse. It got lot worse. For example (me again, 2020):
I was really surprised to learn that Swift recently adopted Smalltalk keyword syntax ... Of course, Swift wouldn't be Swift if this weren't a special case of a special case, specifically the case of multiple trailing closures, which is a special case of trailing closures, which are weird and special-casey enough by themselves.
A prediction I made was that these rules, despite or more likely because of their complexity, would not be sufficient. And that turned out to be correct, as predicted, people turned to workarounds, just like they did with C++ and Java constructors.
So it is true that it is now bad and that it has gotten worse. It's just not the case that it was ever simple to start with. And the further explosion of complexity was not some accidental thing that happened to what was otherwise a good beginning. That very explosion was already pretty much predetermined in the language as it existed from inception and in the values that were visible.
From my exchange with Chris regarding initializers:
"Chris Lattner said...
Marcel, I totally agree with your simplicity goal, but this isn't practical unless you are willing to sacrifice non-default initializable types (e.g. non-nullable pointers) or memory safety."
Part of my response:
"Let me turn it around: Chris, I totally agree with your goal of initializable types, but it is just not practical unless you are willing to sacrifice simplicity, parsimony and power (and ignore the fact that it doesn't actually work)."
Simplicity is not the easy option. Simplicity is hard. Swift took the easy route.
[...] when you first attack a problem it seems really simple because you don't understand it. Then when you start to really understand it, you come up with these very complicated solutions because it's really hairy. Most people stop there. But a few people keep burning the midnight oil and finally understand the underlying principles of the problem and come up with an elegantly simple solution for it. But very few people go the distance to get there.
-- Steve Jobs (borrowed and adapted from Heinelein)
But they can silently drop their mistakes without ever mentioning them again, for example Garbage Collection, Modern Objective-C Syntax, Cocoa-Java.
While they will do this and just start treating the old thing as if it were brand new and shiny, it helps if they actually do have some new shiny thing.
Happy to rebrand Objective-Smalltalk into AppleTalk. The network protocol was dropped 15 years ago, so that could work.
Lasers have very limited applications, they have an inherent line of sight limits, and even the most powerful ship mounted lasers that can do like 50kW, take a minute to boil a kettle of water away, more if you wrap it in tinfoil.
And a shot might cost $10, the laser itself cost $$$, fits only in a cargo container, and requires crazy amounts of juice.
Meanwhile a simple AA gun needs none of those things and can kill things just fine.
At this point no one is talking about using lasers to defend against hyper-sonic missiles (at least not anywhere near the target). All of the current laser systems require being focused on the targets for some amount of time to "burn though", which means they are only suitable for lower-speed targets (drones, cruise missiles, and some low-level ballistics).
You would need to have significantly stronger lasers to try and "burn through" on something moving that fast.
For completeness I should mention that there was quite some work on trying to get laser defenses against ballistic missiles on their "boost" phase (when they were launching, so slow enough to track a point in the missile), for example George Bush's "Star Wars" defense system. These would have been space based (some of the testing involved mounting on 747s, but I don't think that was ever an end-goal), but never made it near production.
Lasers will probably only be used for point defenses against drones which isn't useless but they aren't the cheap future panacea everyone seems to think.
None of that refutes anything that was said. macOS is a third-class citizen measured by market share, and the total sum of annual Mac profits is lower than what the iPad ecosystem makes in a year.
Consumers do not want the Mac. Datacenters don't want Apple Silicon. People want the iPhone, they want Airpods, but the M-series Macs have spent 5 years changing absolutely nothing.
> and the total sum of annual Mac profits is lower than what the iPad ecosystem makes in a year.
So the company that makes between 50-60% of all profits in personal computers has created a market where it makes 100% of the profits, but albeit smaller than the whole PC market. That's terrrrible, what was Apple thinking!
Market share is far from everything when people live in poverty and do not have money to spend on good hardware and software. Apple makes stuff for affluent people, and then makes a ton of money from those rich folks. Making Apple the most valuable company in the history of humanity. Boy, that's a terrible place to be in!
I shouldn't have to repeat myself; this still doesn't refute the claim that Apple has ceded the consumer compute market. Cheap Macs have flooded the used market for years, and people still gravitate towards plastic Wintel boxes and Chromebooks.
> Apple makes stuff for affluent people
is just repeating the original claim upthread:
>> they've priced the consumer/pro-sumer out of the market prettymuch and so B2B is the more sustainable paying population.
The fact Apple maximizes for profits, and does not care about market share, does not mean it has ceded the market at all. It’s the exact contrary. Apple’s making money akin to the #2 position while being #4 and that’s an issue for you?
Once again you retreat to anecdata; how can you prove that used Mac laptops are not popular?
No, treating people with hostility and escalating the situation only makes it more likely that someone will snap and attack a cop.
People generally do not shoot at cops, because whether or not they hit the target doing so is pretty much signing their own death sentence. All cops have to do to protect themselves is to not provoke people to fight-or-flight reactions.
> What is cheap is nuclear, what is expensive (and requiring massive subsidies) are intermittent renewables.
This has not been the case for a while now.
Specifically German source to avoid questions about regional variation, from 2024; only "small rooftop" PV (both with and without batteries) even has an overlap with nuclear, and even then at the low end of the cost range for nuclear:
I suspect we're still going to get new nuclear reactors, but for the radioisotopes and not because of any questions about cost or supply diversity or dunkelflaute.
That "source" is complete BS. It is a theoretical model that models LCOE (Levelized Cost of Electricity), pure generating cost. It does says nothing about the real world, nor does it model the cost of electricity, which has to include system costs. See
Those system costs rise dramatically with the share of intermittent renewables in a grid, making LCOE almost completely meaningless and downright misleading for such a grid. Full Levelized System Cost of Electricity is a better measure.
The discrepancy can also be seen in the real world, where despite falling generating costs for intermittent renewables, electricty prices correlated extremely strongly with the share of intermittent renewables in a grid.
The check against the Real World™ is also another item where this Fraunhofer piece of disinformation fails. They claim costs of 13-49 cents for nuclear.
However, in the Real World™, nuclear power plants profitably produce for 3-5 cents in Switzerland (world-renowned for its low, lower, lowest prices!).
So something very obviously does not add up.
When you look at the model more closely, you see exactly what doesn't add up: all their assumptions going into the model are lopsided, to the point of being absurd:
1. For construction costs, they take as the baseline not the average of current construction costs, nor the median either. They take the highest construction costs seen up to that point. As the baseline. And then increase it. This statistical malpractice even without knowing any details. And the details make it even more ridiculous: their baseline is a FOAK build (prototype), the reactor design (EPR) has been discontinued by the manufacturer, because it was too hard to build, and the country (Finland) had to rebuild its nuclear construction know how.
2. For the lifetime of the plant, they assume 40 years. Nuclear reactors in the US, for example, are all being extended to 80 years, which is also the design lifetime of modern reactors. This obviously doubles the Capex per kWh produced.
3. They assume comically low utilization down to only 20% capacity factor. For references, the US nuclear fleet has a capacity factor of over 90%, as did the German fleet. The reasoning they give is that they want nuclear to run only as a backup to the intermittent renewables. Which is ridiculous, because it allows the intermittent renewables to shift the cost for their intermittency onto nuclear.
4. They also assume much higher capital costs than they do for intermittent renewables. Not sure how they justify this, possibly with the fact that the guaranteed subsidies for intermittent renewables make them a low risk.
5. To balance things out, they assume higher utilization rates for wind and solar than have ever been achieved in the Real World™. And those utilization rates are decreasing, not increasing.
Put those components together, and you get around a factor 10. So this "study" makes assumptions that result in nuclear looking around 10x more expensive than it should be. Which, surprise surprise, pretty much matches the discrepancy between what this study shows and the evidence from the real world.
Just for LCOE.
You then have to add the system costs.
Pretty much the entire rest of the industrialized world are getting into nuclear, expanding their nuclear programs or even ditching their nuclear exits (the most recent: Belgium and Taiwan, with Switzerland on the cusp). There is a pledge to triple nuclear power endorsed by essentially all the economic powers in the world, and they are also following through.
Germany will get there in the end, but we have always been more stubborn in insisting on our mistakes. The longer it takes, the more expensive it will be.
But they're not all in. The politician who is being attacked in this very article for being insufficiently pro-nuclear is also against wind power.
And if the author believes the non-mainstream claim that nuclear is cheaper than renewables then they could have stated that rather than mysteriously avoid the topic.
Germany is still all-in on the failed Energiewende. Last I checked, this is due to the coalition agreement with the SPD, who have not yet managed to come to terms with their mistake.
The problems of the intermittent renewables wind power and solar are inherent, they have nothing to do with being "for" or "against" it. Stating simple facts is not being "against" something.
Well, unless you are such a blind-eyed partisan that any criticism of your religion, no matter how justified, is viewed as heresy. We don't have such blind partisanship here, do we? I mean apart from the crucial topics of vi vs emacs and static vs dynamic typing?
And nuclear being cheaper is not a "non-mainstream view". It is a simple, easily checkable fact.
Let's say that I'm in the same info-bubble as you and the OP and we all agree that renewables are more expensive than nuclear and furthermore that this point is so well accepted that it's not even worth mentioning.
Still in that case, it's kind of weird to attack a pro-nuclear politician for not replacing deadly coal fast enough without even mentioning that he is against wind as well.
Or to avoid any reference the exact same thing happening in your own country.
That "info-bubble" is also known as The Real World™.
Adding more intermittent renewables at this point is useless. Last year we installed significant new wind capacity and output from wind for that year actually dropped.
Although the weather did play a role (which is bad enough by itself!), this is not an anomaly: we have reached a point where we have way more intermittent renewable capacity than demand, so when the sun shines and the wind blows it already needs to be curtailed. Added capacity does not add output here, it just adds to the curtailment.
And when the existing intermittent renewables don't produce, the newly installed capacity also does not produce.
Capacity factors for both wind and PV have dropped, and will continue to drop even faster. And they still need full backup with fossil fuels, because the guaranteed production is either exactly zero or close to zero.
> They have shown that the information density in our existing languages is extremely low: small prompts can generate very large programs.
"Write a book about a small person who happens upon a magical ring which turns out to be the repository of an evil entities power. The small person needs to destroy the ring somehow, probably using the same means it was created"
Small prompts leading to large programs has absolutely nothing to do with programming languages and everything to do with the design of the word generators used to produce the programs — which ingest millions of programs in training and can spit out almost entire examples based on them.
> This tells me you’ve never looked at a demand curve. In for example California the demand swings from 18 GW to 50 GW over the day
Have you been looking at "net demand" curves? Total demand variation is not too large over the day. The wind/solar production enormously increases the magnitude of remaining demand difference over the day.
Nobody is talking about batteries to deal with demand swings between seasons though. Capacity has to accommodate whether it's nuclear or fossil or battery or renewable. The issue is day to day variation. And it does not matter how much wind/solar capacity you have, you can't supply demand without storage. That is untrue of other generation types.
Other generation might use batteries to take the edge of peaks, but that would only be done if it made total cost cheaper. That's not the case for renewables. If there were no other generation then they would have to use storage, so it's always going to make them more expensive.
The net demand curve varies 30 GW over the period you posted?
It goes 7 GW negative.
The problem with nuclear power is that about all costs are fixed. It costs 18-24 cents/kWh when running at 100% for 40 years excluding backup, transmission, final waste disposal and taxes.
Now remove any earning potential from large portions of the day coming from renewables and storage and the economics simply does not pan a out.
> The net demand curve varies 30 GW over the period you posted?
Right. Due to solar/wind.
Gross demand is much flatter. Not completely flat, but it's obvious that it does not require anywhere near the amount of storage or variation that renewables alone would require.
Are you sure with the numbers? Maybe for failed projects like Vogtle it may be true but otherwise, the cost is about 4.7ct/kwh everything included looking at swiss open data. And Goesgen didn't run at 100% CF all these years.
You mean middle eastern labor and design that doesn’t fly with western regulations?
Sounds applicable!
Let’s first acknowledge that KHNP pulled out of all western projects except the Czech one after their settlement with Westinghouse. They don’t exist as an option.
Then let’s look at the Czech subsidies. They aren’t materially different compared to any other modern western nuclear construction.
They’ve shaved a few billion from the headline number but the project is still pure cost plus putting all construction and financial risk on the governments tab.
Barakah is Korean design partly based on Westinghouse patents. That's why for Barakah they had a deal with Westinghouse just like with Czechia. The design is in line with western regulations. Labor is not that relevant for such projects. By far the biggest problems are depleted supply chain and immature design (Both EPRs and Vogtle started when their design wasn't finalized. On top, EPR suffered major design changes for each build due to specific regulations, especially in UK)
Czechian govt subsidies were approved by EC and are pretty ok cost-wise. Even 11bn/reactor is fine considering FLA3 is 23bn. On the other hand, Germany spends on EEG alone each year almost a full equivalent of a failed FLA3. And with new transmission subsidies it's even higher. Both EEG and transmission subsidies are not subject to EC approval, unlike subsidies for nuclear
But agree with the other comment - your remarks sound rather racist
Hmm...any evidence for your weird and to be frank a bit racist claims about "middle eastern labor and design" as well as regulations?
You may not be aware of this, but the UAE is one of the richest countries in the world, on par with with the United States and ahead of Denmark and most of the European nations.
That is incorrect. The profitability floor of even the catastrophic FOAK FV 3 build is 9 cents/kWh for a 2% ROI, with around €130 for 4%. The EPR2 units are lower, around €90/MWh for 4% return and €70-80 for 2% return.
Contrast this with French intermittent renewables projects, which are not profitable at all, EDF receives massive subsidies for them.
Thanks for confirming how insanely expensive new built nuclear power is by using absolutely insanely low discount rates.
Hinkley Point C just got a bridging loan to finalize the plant, after 8 years of building so the remaining risk should be minimal. They got an 7% interest rate. Then EDF needs to make profit on top of that.
When using real world discount rates for FV3 you end up towards 20 cents/kWh.
I love how you quote the EPR2 units cheaper than the proposed subsidies. This looks like blind conviction rather than a factual statement.
The proposed subsidies for the EPR2 fleet is a 10 euro cents CFD and interest free loans. Sum freely, but you end up towards 20 cents per kWh.
And that is excluding for example the backup needed when suddenly half your fleet is offline at the same time. Like happened in France during the energy crisis and multiple times in Sweden last year.
I don't call out all the disinformation you put forward en-detail, that would be far too much work.
The interest-free loans are also not a subsidy, at least not the loans, as the principal has to be repaid in full. The lack of interest payment is a subsidy, but that's a lot less than the loan amount, and being a state company EDF can get pretty good interest rates on the free market. Oh, and those loans will only be for half of the investment.
My estimates put the value of this subsidy at around €20 billion. So less than 1 year of Germany renewable subsidies just from the EEG. For plants that will run for 80 years or more and produce electricity worth >€600 billion (at 10 cents).
20/600 = 0,0333 or around 3,3% of those 10 cents, or less than about 1/3 of a cent per kWh.
And of course the French state owns EDF, so when EDF makes a profit, the state gets those profits. As they have been doing consistently for the last half century or so.
So as with CFD payment, which you put as a subsidy of 10 cent/kWh, there is the tiniest grain of truth in your claims, but then inflated beyond the pale.
Speaking of the CFD: the expected wholesale price of electricity in Germany is expected to be around €90-95/MWh in the next couple of years, and prices have tended to be higher than predictions. And if the wholesale price goes above the CFD price, then EDF loses money on the CFD, because they get exactly the CFD, no less, but also no more.
Assuming the predictions are correct and apply to France as well, the subsidy would be 0,5-1 cents/kWh, so a factor 10-20 less than your claim of 10 cents, and only if wholesale prices actually stay low.
After all, the CFD is primarily necessary, because the subsidized and preferentially treated intermittent renewables have wrecked havoc with wholesale electricity prices, with prices in Germany in 2025 fluctuating wildly between + €583/MWh and - €130/MWh.
It's not really the price, it's the artificially and unnecessarily introduced fluctuations that cause problems for investors.
And of course those renewables that are causing al this havoc get vastly more subsidies per kWh than this. And preferential loans, preferential feed-in, preferential regulations etc.
Anyway, your claim was 20 cents of subsidies per kWh. The real value ranges from less than a cent (could even go negative) to maybe up to 2 cents. So very generously you inflated by a factor of only 10-40.
So you can see why I don't debunk all of the disinformation you put forward, just some of it. It's too much work.
Do you think that demand is willing to pay 18-24 cents per kWh excluding backup, insurance, transmission, final waste disposal etc.?
I can tell you they won’t. Which is why there’s currently 0 commercial nuclear reactors under construction in the US.
You also have to look at it from incentives. Why should a person or company with solar and storage buy horrifyingly expensive nuclear power from the grid when their own installation delivers?
Well, they don’t.
Why should their neighbors prefer horrifyingly expensive grid based nuclear electricity to their neighbors excess renewables?
Well, they don’t.
Do you know what they do instead? These steady consumers. They buy financial instruments like futures and PPAs to ensure steady price and supply.
The problem for new built nuclear power is that these financial instruments costs a fraction of the price new built nuclear power requires.
New built nuclear power is the fax of the internet age. It is time to let go.
The exact opposite of what you write is true. Please inform yourself.
They are currently buying up all the nuclear they can get their hands on, reactivating shut down plants, investing both in startups and in existing players to build as much nuclear as they can:
Storage would mean just to reroute the energy to storage, otherwise you need to lower the power plant‘s output what doesn’t happen fast in nuclear power plants
That's true, but only partly true. It already was a gigantic super complicated bag of special cases right from the start.
Rob Rix noted the following 10 years ago:
Swift is a crescendo of special cases stopping just short of the general; the result is complexity in the semantics, complexity in the behaviour (i.e. bugs), and complexity in use (i.e. workarounds).
https://www.quora.com/Which-features-overcomplicate-Swift-Wh...
Me, 2014:
Apple's new Swift language has taken a page from the C++ and Java playbooks and made initialization a special case. Well, lots of special cases actually. The Swift book has 30 pages on initialization, and they aren't just illustration and explanation, they are dense with rules and special cases
https://blog.metaobject.com/2014/06/remove-features-for-grea...
Of course, that doesn't mean that it didn't get worse. It got lot worse. For example (me again, 2020):
I was really surprised to learn that Swift recently adopted Smalltalk keyword syntax ... Of course, Swift wouldn't be Swift if this weren't a special case of a special case, specifically the case of multiple trailing closures, which is a special case of trailing closures, which are weird and special-casey enough by themselves.
https://blog.metaobject.com/2020/06/the-curious-case-of-swif...
Oh, and Function Builders (2020, also me):
A prediction I made was that these rules, despite or more likely because of their complexity, would not be sufficient. And that turned out to be correct, as predicted, people turned to workarounds, just like they did with C++ and Java constructors.
https://blog.metaobject.com/2020/04/swift-initialization-swi...
So it is true that it is now bad and that it has gotten worse. It's just not the case that it was ever simple to start with. And the further explosion of complexity was not some accidental thing that happened to what was otherwise a good beginning. That very explosion was already pretty much predetermined in the language as it existed from inception and in the values that were visible.
From my exchange with Chris regarding initializers:
"Chris Lattner said...
Marcel, I totally agree with your simplicity goal, but this isn't practical unless you are willing to sacrifice non-default initializable types (e.g. non-nullable pointers) or memory safety."
Part of my response:
"Let me turn it around: Chris, I totally agree with your goal of initializable types, but it is just not practical unless you are willing to sacrifice simplicity, parsimony and power (and ignore the fact that it doesn't actually work)."
Simplicity is not the easy option. Simplicity is hard. Swift took the easy route.
[...] when you first attack a problem it seems really simple because you don't understand it. Then when you start to really understand it, you come up with these very complicated solutions because it's really hairy. Most people stop there. But a few people keep burning the midnight oil and finally understand the underlying principles of the problem and come up with an elegantly simple solution for it. But very few people go the distance to get there.
-- Steve Jobs (borrowed and adapted from Heinelein)
https://blog.metaobject.com/2014/04/sophisticated-simplicity...
reply