3,679 karma · joined July 20, 2015
DeepSeek v4 Flash/Pro also exist, they are open weight and on par with Sonnet, just a bit below Opus. Again: practically useful and sufficient for 80% of things most people actually do. And most of the remaining 20% are benchmarks designed to push the limits, not productive work.
Using these already is way cheaper than your typical Claude API prices. What's still missing is a) mindshare - everyone still thinks "claude = coding" and everyone thinks he/she really needs the very best models because he/she is doing such incredibly complex stuff - and b) someone pushing such a stack as a convenient solution for corporations to easily dump their token money into, complete with user management, enrollment, monitoring, all that enterprisey stuff you need if you want to sell to, well, enterprise customers.
China knows that doing what Anthropic/OpenAI/Google/... are doing is impossible for them. No one outside of China in any sane condition will send their data to compute farms IN CHINA like people currently do with US-based frontier models. Even if they could muster the inference power.
Hence they do the second-best thing possible to attack the dominance of the US-based corporations: reduce their moat by open-sourcing models that are not fully equal, but practically useful and good enough for easily 90% of typical tasks people use agents for in their daily lives. But way cheaper to run.
As long as this arms race in AI continues, China as "number two" will have some incentive to continue open-sourcing models. But of course the US government might force a change if they continue to enforce limited public access to new frontier models - there is no market to minimize if a model is not allowed to be publicly available.
They captured like 90% of the German ticket market by closely watching Ticketmaster and basically repeating their playbook. Eventim also owns some venues, has exclusive contracts with many of them it doesn't outright own, hosts an official fan resale site, offers promotional services and generally integrates a lot of the business around large events vertically.
What's actually interesting is that it seems to be possible to compete with Ticketmaster only by copying their playbook on a large-enough scale, and Germany appears to be large enough. The Netherlands right next to Germany don't seem to have been large enough, as Ticketmaster basically controls them.
People call this exact process "vibe coding".
The typical job of a Product Manager is also not to directly perform this mapping, although the PM is much closer to that activity. PMs mostly need to enforce coherence across an entire product with regard to the ways of mapping business needs to software features that are being developed by individual developers. They still usually involve developers to do the actual mapping, and don't really do it themselves. But the Product Manager must "manage" this process, hence the name, because without anyone coordinating the work of multiple developers, those will quickly construct mappings that may work and make sense individually, but won't fit together into a coherent product.
Developers are indeed the people responsible to find out what business actually wants (which is usually not equal to what they say they want) and map that onto a technical model that can be implemented into a piece of software - or multiple pieces, if we talk about distributed systems. Sometimes they get some help by business analysts, a role very similar to a developer that puts more weight on the business side of things and less on the coding side - but in a lot of team constellations they're also single-handedly responsible for the entire process. Good developers excel at this task and find solutions that really solve the problem at hand (even if they don't exactly follow the requirements or may have to fill up gaps), fit well into an existing solution (even if that means bending some requirements again, or changing parts of the solution), are maintainable in the long run and maximize the chance for them to be extendable in the future when the requirements change. Bad developers just churn out some code that might satisfy some tests, may even roughly do what someone else specified, but fails to be maintainable, impacts other parts of the system negatively, and often fails to actually solve the problem because what business described they needed turned out to once again not be what they actually needed. The problem is that most of these negatives don't show their effects immediately, but only weeks, months or even years later.
LLMs currently are on the level of a bad developer. They can churn out code, but not much more. They fail at the more complex parts of the job, basically all the parts that make "software engineering" an engineering discipline and not just a code generation endeavour, because those parts require adversarial thinking, which is what separates experts from anyone else. The following article was quite an eye-opener for me on this particular topic: https://www.latent.space/p/adversarial-reasoning - I highly suggest anyone working with LLMs to read it.
And most people I know who love spending time on this kind of content would not care either - because they don't care whether they waste time on real or AI animal videos. They just want something to waste time with.
His investors are not investing because of his success rate in delivering on his promises. His investors are investing exclusively because they believe that stock they buy now will be worth more tomorrow. They all know that's most likely not because Elon delivers anything concrete (because he only does that in what, 20% of cases?), but because Elon rides the hype train harder tomorrow. But they don't care if it's hype or substance, as long as numbers go up.
Elon's investors are happy with his success rate only in terms of continuously generating hype. Which, I have to admit, he's been able to keep up longer now than I ever thought possible.
Tesla's sales have suffered, yes, and Elon's image is a significant contributor to that, besides all the reasons directly related to the cars themselves.
But Tesla's stock price is still stuck in irrational heights, not even remotely justifiable by the company's performance.
It just seems that people reconsider purchasing a physical object way quicker than they reconsider a stock investment. Maybe because the stock investment, especially in TSLA, is considered more like a gamble - "as long as others also think that this stock will skyrocket, even just because they think that others like me think it will skyrocket - as long as that's the case, I'm good with buying shares".
For non-technical supply chain managers to pick up enough technical chops to understand the stuff they are supposed to manage the supply chain of is comparatively more difficult.
Especially when fierce negotiations to push the price down are not the highest priority, but robustness of supply chains, having alternative options that technically work, and ensuring quality according to tight specs are paramount. Which is how I assume ASML supply chain management to work.
In addition to that, you'll have the problem of inventory differences, which is often cited as being an even bigger problem with store theft than the loss of valued product. If the inventory numbers on your books differ too much from the inventory actually on the shelves, all your replenishment processes will suffer, eventually causing out of stock situations and thus loss of revenue. You may be able to eventually counter that by estimating losses to billing inaccuracies, but that's another complexity that's not going to be free to tackle, so the 1% inaccuracy is going to cost you money on the inventory difference front, no matter what.
You don't really need an army of sales managers to sell such a product. Going lean on management and more heavy on engineering is therefore a good idea if you want to keep the lead you have.
They don't have a real website advertising their services, but they seem to do well, probably their customers know them. They've run their business continuously for at least those 10 years I've lived at that spot. I could smell the soldering oven running constantly.
Overly confident, but poorly informed articles aren't commonly written by teenagers anymore, but by LLMs.
And inspecting each part is not enough to understand how, together, they achieve what they achieve. We would need to understand the entire system in a much more abstract way, and currently we have nothing more than ideas of how it _might_ work.
Normally, with software, we do not have this problem, as we start on the abstract level with a fully understood design and construct the concrete parts thereafter. Obviously we have a much better understanding of how the entire system of concrete parts works together to perform some complex task.
With AI, we took the other way: concrete parts were assembled with vague ideas on the abstract level of how they might do some cool stuff when put together. From there it was basically trial-and-error, iteration to the current state, but always with nothing more than vague ideas of how all of the parts work together on the abstract level. And even if we just stopped the development now and tried to gain a full, thorough understanding of the abstract level of a current LLM, we would fail, as they already reached a complexity that no human can understand anymore, even when devoting their entire lifetime to it.
However, while this is a clear difference to most other software (though one has to get careful when it comes to the biggest projects like Chromium, Windows, Linux, ... since even though these were constructed abstract-first, they have been in development for such a long time and have gained so many moving parts in the meantime that someone trying to understand them fully on the abstract level will probably start to face the difficulty of limited lifetime as well), it is not an uncommon thing per se: we also do not "really" understand how economy works, how money works, how capitalism works. Very much like with LLMs, humanity has somehow developed these systems through interaction of billions of humans over a long time, there was never an architect designing them on an abstract level from scratch, and they have shown emergent capabilities and behaviors that we don't fully understand. Still, we obviously try to use them to our advantage every day, and nobody would say that modern economies are useless or should be abandoned because they're not fully understood.
„which are even worse“ is an assumption you make, you do not have any data to back that up. From what I regularly see, they are not worse at all. People without sun on their balconies do not buy balcony solar kits in the first place. Also there’s the fact that people can spend time to meticulously optimize every single panels’ location (which they usually don’t when someone places 30 panels on a roof in a single day, it’s just about getting them up there quickly). If you are interested in numbers, browse https://www.reddit.com/r/Balkonkraftwerk/ - it’s German language but Reddit does quite well with auto-translation as far as I know. Every month, people post their yield numbers for comparison there.
> No, because when you get any significant amount of solar installed you start to get negative prices on sunny hours and need to shut them down. If home solar setups won't shutdown then some other panels in the grid would.
You are equating the time during which a home solar owner cannot use his own solar power with the time during which there are negative electricity prices. This is grossly wrong. In 2024, Germany had 457 hours of negative power prices (see https://www.pv-magazine.de/2025/01/03/bundesnetzagentur-457-...). That’s roughly 5% of the year. Typical home solar power usage if no battery at all is installed is about 50%. If we talk about batteries, which are increasingly getting common in balcony solar installations due to significant price drops, it’s more like 80-90% of power that the owner can use directly.
> You confuse the "maximum possible outcome" with real life. No one knows if these $300 setups will last 30 years, that was never tested because that requires well 30 years. My estimate is they won't because electronics from the lowest price range very rarely do.
The panels are the exact same panels used for large-scale solar installations. These are tested and guaranteed by the manufacturer for 30+ years. Nobody doubts that they’ll reach that lifespan in most cases. The inverters are a negligible amount of kWh invested, as I pointed out in the parallel threads’ posting. So you can easily buy one or two replacements over the 30 year timespan without impacting the EROI of the panel at all. Also, a lifespan of 30 years does not mean that the panel fails after 30 years. It just goes below a defined point of efficiency (80% of original peak power). You can very well use it for another 10 or 20 years, you just have to accept that it produces only 80% of the original output.
> Then even if they could last that long, half of them will end up in a dumpster after a few years because people move and can't always take their panels along.
That’s not what I see, because it is surprisingly hard to dispose of solar panels in practice. They do not fit into the typical „dumpsters“ people use to dispose their regular trash. You would be able to dispose of them for free at the next recycling center in Germany, as they are mandated to take them, but most cars cannot be used to transport solar panels as they are too large, so it’s not trivial to get them there. From what I observe, people therefore simply sell or donate the panels to the next renter/owner when they move, which is obviously a good idea as they are usually installed on a balcony or garden house or whatever and you usually buy the matching installation equipment for a particular situation which you wouldn’t be able to use at your future home anyway.
I find it quite interesting that you did not object to my second argument about the psychological impact/use of this technology.
Non-optimal positioning is already included in your "Measured (not projected) yield in Germany". Because that is the difference between a projected yield (under optimal conditions) and actual measurements of actual panels, which are practically never optimally placed.
> Then you won't consume all that it produces because you're at work during the day and there are no appliances running except the fridge maybe.
That is relevant for an economic calculation, but it is entirely irrelevant if you want to determine the point after which the panel breaks even regarding the energy used for its production vs. the energy produced by it. In that case, every single kWh counts, whether the owner of the panel economically profits from it or whether he or she just donates it to the grid without compensation.
And clearly, we are discussing the energy break-even here, as indicated by "The usual estimate to produce a PV panel is 600-1000 kWh per 1sqm."
> However from the physics/ecology perspective they make no sense at all and many of the panels installed today will never recoup the electricity used to produce them, making them a net-negative impact for the environment.
As I've demonstrated with concrete calculations, that you seemingly accept as valid as you perform the same calculations with roughly the same numbers, the EROI of solar panels even in Germany over their lifespan is clearly in the positive range. Maybe they "only" recoup 3x or 5x their investment, and not 20x, but they are a net positive regardless. Any number above 1x is.
In addition to this, as I've described in another posting here (https://news.ycombinator.com/context?id=45490555) there's the psychological side of things, where cheap and easily profitable balcony panels for everyone are a gateway drug for "normal people" to get actively involved in the field of renewable energy and must not be underestimated in their ability to open the minds of people for other, more efficient actions to get closer to a carbon-neutral energy economy. Since those activities tend to be heavily inhibited by broad refusal that's often not based on factual arguments, but simply on inertia in people's minds ("we've always done it the other way") and a certain lazyness to actively grapple with new technologies and developments, this effect is at least as important as the actual impact on the energy grid. Just like on the stock market, the raw numbers are only half of the story. Psychology is the other half.
Scientific publications say otherwise. It's quite hard to come by any numbers, but page 11 of https://www.wisdomlib.org/uploads/journals/mdpi-sust/2025-vo... says that typical solar inverters require about 15 MJ/kW of power for their production in total, which would amount to approx. 4 kWh per kW of inverter power. A solar panel square meter produces about 200-250 watts of peak power, so it needs inverter power that cost about 1 kWh to build. Let's triple that, because inverters have a typical lifetime of 10 years in contrast to the 30 years of solar panels. 3 kWh for the inverter is negligible when compared with 300-2000 kWh for the panel itself. So we can just ignore that.
Batteries are interesting. According to https://www.mdpi.com/2076-3298/12/1/24, it takes about 35 kWh in total to produce 1 kWh of battery capacity. Let's say we'll need 250 Wh of capacity for our 250 WPeak solar panel square meter (the rule of 1 kWh capacity for each 1 kWp is a typical estimate applied when sizing solar installations for residential homes). That makes up about 8 or 9 kWh to produce this battery capacity. Admittedly that doesn't include the raw materials, of which the cell requires quite a few expensive ones. Unfortunately I wasn't able to find a good resource on that, so I resorted to asking ChatGPT for a rough calculation, and it came up with about 140 kWh for our 250 Wh LFP cell, which doesn't sound entirely wrong, as it assumed a cell weight of 1,5 kg and splitted that up into different materials. The weight matches what I would expect from personal experience with LFP batteries.
Basically, we can just ignore the inverter and must add about 150 kWh for the battery to our 300-2000 kWh for the panel. That does not substantially impact an EROI calculated from a 1000 kWh assumption for the panel alone.
And this is a calculation based on Germany. Again: weather conditions are far from optimal for solar in Germany. It's much better in many regions in China, where solar panels are made. They can easily achieve EROIs of 20+ with solar there, which is probably the reason why China installs absolutely HUGE numbers of panels. But according to you, they must be "delusional" over there.
That is clearly wrong. Even the worst-case embodied energy assumptions for solar panels estimate the cost of producing a square meter of solar panel area at 2000 kWh (the best cases are around 300, see https://en.wikipedia.org/wiki/Energy_return_on_investment#Ph... ). A square meter of solar panel area produces an average of 200 kWh of power per year in Germany (which implies a pessimistic assumption, more sunnier countries can get a multiple of that). This means that even in the worst case, the solar panel has amortized itself from the perspective of embodied energy after 10 years. On average it will be more like below 5 years. Solar panels however have an expected lifetime of well over 30 years and require no maintenance if installed correctly.
Balcony solar power plants are sort of a gateway drug into actual, practical participation in the renewable energy sector. They are easy to install, cheap, have a clear and fast way to profitability, and provide significant gamification value (people who buy these kits tend to start with constantly monitoring their energy generation and usage in apps afterwards). That "ice breaker" effect should not be underestimated. It can pave the way to way more substantial actions (or to the acceptance of actions taken by others) that people wouldn't have considered otherwise simply due to inertia of the status quo.
> look ugly
That is YOUR taste. I consider most balconies with solar panels to look futuristic and cool. Garden houses with solar panels on the roof also look way cooler and more modern than without them.
So the actual "byzantine incomprehensible mess" (which is indeed the correct description) is the MSI format and Windows Installer, not WiX.
Is it that hard to admit that you didn't properly read the comment you responded to?
They will most certainly be able to make the US subsidiary of Hetzner turn out any kind of data it technically has access to. But if Hetzner is not entirely stupid (and they are usually pretty smart) they set up their internal networks such that the US admins cannot access data located in the EU.
The problem with EU subsidiaries of Amazon and Microsoft is that the principal corporation is located in the US and subject to US jurisdiction, and that eventually, the owning company is always able to make a subsidiary comply with its demands, so it's virtually impossible to set up an impenetrable barrier between them. It's the other way round with Hetzner. A subsidiary can't command the owning company around, not even if the US government wants it to.
However, writing that kind of code maybe makes up 5% of my work. Analysis, trial-and-error, discussions etc. make up the other 95%, and AI only seldomly helps with that. It can sometimes be useful for research and spec ingestion, but it quickly becomes dangerous in those cases because as soon as you enter any kind of niche area (and unfortunately my work has a lot of those) LLMs tend to hallucinate and present made-up "knowledge" with enviable certainty.