Please don't be that guy with 1 karma who never took interest in Hacker News until he wrote a blog post and then, for some reason, has to post a summary of his blog post as a comment. Before the coming of AI slop this just wasn't a thing: if we want to read your post we can read your post.
China leads the world in battery manufacturing, it's not really so hard in an absolute sense, but gee do you have to get the cost down and get the quality up so you don't have batteries blow up because of manufacturing defects.
To make gas turbines though you need to get the last bit of performance available out of metals, manufacture parts with punishing tolerances, and do it with high quality so turbines don't blow up because of manufacturing defects. At this point in time China cannot make competitive gas turbines. China's COMAC C919 airliner is 100% dependent on western engines for instance.
The market for aircraft engine gas turbines is at least 5x the market for industrial gas turbines. We could eliminate natural gas for electricity overnight and GE would keep spinning. Even if it becomes hard to justify development of land-based gas turbines it will still be possible to repurpose jet engines for this purpose.
(Though I really want to see a fast reactor coupled to a supercritical CO2 or helium brayton cycle turbine!)
Horses for courses, right? They build these entire peaker power plants to sit idle most of the time, just to handle the couple dozen or hundred hours of maximum load in the year.
4 hours at a time hopefully buys you your way through such a peak, even if it doesn’t get you through weeks at a time… and presumably you can top it up from base load after demand peaks for the day, right?
In the equatorial and desert regions they’re talking about, I think both seasonal supply changes and demand profile might be more consistent than, say, dreary cold parts of Europe..
That's a big part of what drives me crazy about the papers I've seen on this work. People like to quote some levelized price of energy and say "this costs less than an AP1000" and you never see modeling of "here's the price in California, here's the price in New York"
(3) annual variation (you might get 3x the sunlight in summer than winter? do you build 180 days of storage or do you overbuild solar systems 3x and get Casey Handmer to dream up something useful to do with that excess energy that doesn't have a stupendous capital cost?)
The "have a natural gas backup plan" is worse than people think for a lot of reasons, in terms of the laws of political science and economics you're going to find that that tail winds up wagging the dog, the whole market will get designed around the needs of those turbine owners, it's going to cost the same if they are running turbines for 5 minutes a year or 20 days a year, and ultimately you're never going to get rid of it. If you really plan to run the system 5 minutes a year what are the odds it will really work when you need it? You have to not just support the turbines you hardly ever use, pipelines you hardly ever use, drilling and storage operations for gas, etc.
Yeah I think it really only works well in the central latitudes. The American South, the tropics and equatorial regions, Southern Europe, etc. that get a lot of high-angle sunlight year-round.
Otherwise you need to deal with long distance power transmission across many political jurisdictions, or local generation for night or other low-solar times. To my mind, nuclear base load with solar and batteries used for things like charging cars or demand-shifting probably makes more sense. It's really a shame that we mostly withdrew from building nuclear plants for the past three or four decades; instead we have reached the point where we have to re-learn how or recreate the tooling needed to build them and they are still too damn expensive.
nuclear base load makes sense from a pure grid engineering PoV, which i think is why it appeals to so many HN folk. Sadly, it makes no sense economically anymore, so it will probably never be built at scale.
They quit building coal burning power plants in the US at the same time and for the same reason they quit building nukes…. The cost of the steam turbine and heat exchanges. I mean, the steam generators for a PWR are bigger than the reactor vessel, it is just insane to use hot water at high pressure to boil water at low pressure.
If nuclear is ever going to be competitive it is going to have to be some kind of reactor that runs at higher temperatures with a gas turbine power set. Could be a fast breeder or HTGR or molten salt but not water cooled. Anyway it is not a bird in the hand.
> (3) annual variation (you might get 3x the sunlight in summer than winter? do you build 180 days of storage or do you overbuild solar systems 3x and get Casey Handmer to dream up something useful to do with that excess energy that doesn't have a stupendous capital cost?)
How about overbuild and interconnect? You don't have to overbuild for your personal worst case if you can buy power from somebody else if it comes to it.
Everyone I know who used the 360 has a lot of nostalgia for it.
It was well thought through. I mean, the 360 architecture is still with us because it avoided the traps that killed the PDP-11, VAX and the 68k -- as much as I loved the PDP-11. It's true when it came out that nobody had any idea what a general-purpose operating system looked like and it took a decade for them to productize VM so that you could run as many operating systems as you needed simultaneously.
Not the person you're replying it to. I had that happen with Microsoft's OneNote. I really liked it and I had no problems parsing the XML files it made so I could pipe data into other systems but then it went 100% cloud storage.
It was a classic example of "Microsoft made something surprisingly good, made people think it was crap because they stuffed five OneNote icons on your taskbar and three up your nose, and then they wrecked it", must have come from the same mind that thought buying Activision was a good idea.
Like an application to book a haircut at a salon or log into the WiFi at a hotel that does something "on rails" and should work without any learning or training is one thing. A complex application for professional creative work is something else entirely.
Then there are the things that are just weird, like life got me looking at mobile games lately and I was really shocked to see how many of them have a horrific onboarding experience. The best games in the "clothes collecting" genre, for instance, like Love Nikki have a gentle onboarding process that teach you the game mechanics and get you into the story and maybe get you hooked. When I was developing consumer-facing social and entertainment apps (before Facebook!) I had my own belief that "easy onboarding is good for users and good for profits" and worked with founders who got it and pushed me to outdo myself. Most games in that genre have horrific UI and don't try to explain the mechanics because they could care less if users understand the currencies used in the system because they only care for the worst "whales" who click on any red dot and fold green at the slightest difficulty.
These days when it comes to "slow software", I think five years is an eternity from a platform perspective so if you start something you may need to rebuild it. Like I find it hard to justify developing on anything other than the web platform (like "write once and it runs on platforms you never thought of like VR headsets", "for christ's sake did you realize every Windows software vendor had at least one InstallShield engineer back in the 1990s and IT was pulling their hair out updating all their desktops?", "it took three weeks for the App Store to reject your update? what did you think would happen?") but the app I am working on right now depends on React components that haven't been updated since 2021 and the bill is coming due.
Oh man, the state of dress up games is a personal pet peeve of mine, poor onboarding tutorials is just one problem. The art direction is often simply atrocious, proper game loops barely exist (let alone creative ones), the microtransactions are the default business model. I honestly don't know what causes it, perhaps the games are just not that profitable so they don't attract highly skilled or at least conscientious game developers? I came to accept that they are just made by people who either dislike/don't care about games in general or dislike dress up games specifically. The best dress up games are normally made by Nintendo who have a strong game design and gave development culture in general.
P.s. As a side note, I started making small dress up games that I draw myself to scratch my own itch. I am not a programmer so they are absolutely horrendous and I only show them to friends. Making videogames is not as easy as making paper clothes for paper dolls like I did as a kid, but it is so fun and I feel 30 years younger somehow. :D
Yeah, I think those games have more potential than most people think. I love voting mechanics (spent a lot of time on ranking systems) and some have good stories ranging from "the heartwarming story of Nikki exploring a continent, making friends and proving her mad skills" to "a vicious fight for survival in an ancient court"
What's a good voting system? I remember checking out Dress to Impress (not a mobile but an incredibly popular Roblox game), and it had a somewhat unpleasant, imo, voting system -- players vote for each other's outfits and the score is a sum of the scores your co-players' gave you which incentivizies players to give everyone the lowest score. The game tries to "punish" this behavior by detecting "unfair" voters and lowering the weight of their votes, but I feel like you could tweak the original incentives instead by changing the mechanic. Having to rank outfits instead of giving each outfit a score could work I think? But there are probably other problems I'm missing. I dnn't play too many mutliplayer games.
(On the other hand, while I believe thinking about "gaming" the ranking takes away from the fun of the game, Dress to Impress is massively popular, so maybe this frustration is what makes it popular. IDK if kids play it because there are no competitors in the genre or because some competitive toxicity is addictive.)
Well I'm interested in voting theory from the viewpoint of political science and also from the viewpoint of developing ranking systems. It's actually a pretty depressing topic because you can easily come to the conclusion that the public doesn't have an interest at all.
In short, pairwise rankings are real and comparable between people whereas different people might use a scale differently. It's a difficult problem to deal with "i like the bagel shop a little bit more than the rice bowl restaurant but the party member who has celiac's opinion matters more than mine" See
now if you have N options you don't have to have everyone evaluate N(N-1) pairs because preferences are correlated and should be transitive. There are many algorithms, like Elo and the methods in that paper that can be used to run tournaments and compute scores. The game Covet uses pairwise ranking and I think gets good scores.
To get back to social decision theory, it's probably not good for players to be in a team for purposes of winning and losing. Like I think sharing clothes with your friends is a really fun mechanic (might make you want to really buy or work for something!) but if people can help or hurt their friends you get into the whole can of worms of coalitions in N-person games.
I just quickly tried Covet and I can imagine how the act of comparing and picking one of the two options can feel fun. I can see myself just voting on outfits in the game without creating my own when I need some really simple "mindless" fun. Not a productive effect but also not addictive enough AFAIK to be a big problem.
It's not just "can these organizations produce brilliant output?" but "can they produce quality output?" I think you can take a lot of pride in doing ordinary things properly and giving people good experiences.
I stayed at a major chain hotel this weekend and it was a fine stay, the staff was great, but... there were just numerous little tech problems. Like the USB chargers had bad connections so I plugged in my phone overnight and it didn't charge and the next morning I found I had to twist the connector to get it to work. And then when I tried to connect to the WiFi with a Steam Deck [1] it wanted my personal info to join their rewards program, but the complicated Javascript-based dropdown menu to pick a state barely worked and after a minute of screwing with it I was able to pick the wrong state but not pick my real state so I looked up a zip code in that state and entered it.
When I got home I got an email from "no-reply@thathotelchainfoundedbyparishiltonsgreatgrandfather.com" gushing about how they want me to have the perfect experience and how badly they want to hear from me. Like yeah, they want me to call their 800 number and give me a piece of my mind.
It bugs me because I spend my day fixing problems with Javascript apps and I know all about those overcomplicated dropdowns. I get the feeling that if a staff member screwed up they'd be held accountable but people act like technology problems are like the weather and there is nothing we can do about it.
[1] granted an unusual device BUT... I don't travel with a laptop anymore, don't need to connect my phone with WiFi in 5G country and may or may not connect my iPhone. They could have just used the <select> that comes with HTML and it would have "just worked" and today you have so many options to style those with CSS
I'm totally the kind of person who loves sales, scans clearance racks regularly, and is always looking for deals but... Prime Day is the definition of "nothingburger". I mean, Kindle is on sale, Fire tablets are on sale. The things that are always on sale are on sale. It's an embarassment that the media covers Prime Day at all.
It is an absolute requirement if you want to build something that (1) uses a classifier as part of a larger system, (2) can be let off the leash with no human supervision and (3) is not an AI slop demo that will be [dead] and [flagged] in seconds after being posted to HN.
For instance if you have a predictive model for market prices that is not calibrated that's... nice. If you have a calibrated model you can add a Kelly better and you have a trading strategy that makes money. Similarly if you are classifying articles or images or other contents to make a feed you might believe that 70% or 95% or some other level of precision is "good enough" and you can set the knob and turn on the cruise control.
reply