It's not so much teenage credulity, or coddling parents. Teen suicide is the easily quantifiable tip of the iceberg when it comes to mental health outcomes. Conspicuously it started trended up after 2008, around the nascence of Facebook and smartphones:
> Following a downward trend until 2007, suicide rates significantly increased 8.2% annually from 2008 to 2022, corresponding to a significant increase in the overall rates between 2001 to 2007 and 2008 to 2022 (3.34 to 5.71 per 1 million; IRR, 1.71)
That's also when the Great Recession happened, giving young people bleak outlooks for their future, outlooks which never really recovered. Nothing was fixed, and things have only gotten worse since then.
Dead end jobs with little to no benefits, no pensions, time off, low pay and few hours count as "employment".
Their parents and grandparents had pensions and could work at one employer for the entirety of their careers with growth opportunities, and could afford homes and healthcare while doing so.
Your hypothesis might be right, but I've provided data, and you're providing opinions. I'm fine with being wrong in my claim, but I didn't earn the downvote when no-one seems to have a clearer hypothesis with better evidence. First, show me that this shift is peculiar to 2008. And then show me that this is what teenagers are killing themselves over.
Just to test out the OP articles theory, I was about to write some unit tests. I decided to let Opus 4.5 have a go. It did a pretty good job, but I spent probably as much time parsing what it had done as I would have writing the code from scratch. I still needed to clean it up, and of course, unsurprisingly, it had made a few tests that only really exercised the mocking it had made. A kind of mistake I wouldn't be caught dead sending in for peer review.
I'm glad the OP feels fine just letting Opus do whatever it wants without a pause to look under the covers, and perhaps we all have to learn to stop worrying and love the LLM? But I think really, here and now, we're witness to just another hype article written by a professional blogger and speaker, who's highly motivated to write engagement bait like this.
That is the thing ... How long ago did we get Agent mode. Like in CoPilot that thing is only 7 months old.
Things evolve faster then people realize... Agent mode, then came mcp servers, sub agents, now its rag databases allowing the LLMs to get data directly.
The development of LLMS looks slow but with each iteration, things get improved. As yourself, what will have been the result of those same tests you ran, 21 months ago, with Claude 3.0? How about Claude 4.0, that is only 8 months ago.
Right now Opus 4.5 is darn functional. The issue is more often not the code that it write, but more often it get stuck on "its too complex, let me simplify it", with the biggest issue often being context capacity.
LLMs are still bad at deeper tasks, but compared to the last LLMs, the jumps have been enormous. What about a year from now? Two years? I have a hard time believing that Claude 3 was not even 2 years but just 21 month ago. And we considered that a massive jump up, useful for working on a single file... Now we are throwing it entire codebases and is darn good at debugging, editing etc.
Do i like the results? No, there are lots of times that the results are not what "i wanted", but that is often a result of my own prompting being too generic.
LLMs are never going to really replace experience programmers, but boy is the progress scary.
I can't say my opinion has changed. It didn't give me results that more exciting or useful than Sonnet. Is it worth 3x price per token? I'm not so sure.
(It wasn't clear in my comment, but I already use agents for my code. I just think the OPs claims are overblown.)
This is only true if the code it wrote is something you can just sit down and write without any reference.
Now do something like I did: An application that can get your IMDB/Letterboxd/Goodreads/Steam libraries and store them locally (own your data). Also use OMDB/TMDB to enrich the movie and TV show data.
If you can write all that code faster than read what Claude did, I salute you and will subscribe to your Substack and Youtube channels :)
Oh btw, neither Goodreads, IMDB nor Letterboxd have proper export APIs so you need to have a playwright-style browser automation do it. Just debugging that mess by writing all the code yourself is going to be hours and hours.
The Steam API access Claude one-shotted (with Sonnet 3.7, this was a long time ago) as well as enriching the input data from different sources.
> If you can write all that code faster than read what Claude did
I think you need to parse my comment a little more keenly ;)
> The Steam API access Claude one-shotted (with Sonnet 3.7, this was a long time ago) as well as enriching the input data from different sources.
This story isn't different to the usual "I made a throw-away thing with an LLM and it was super useful and it took me no time at all". It's very different to the OP stating you can throw this at mature or legacy codebases and reduce labour inputs by 90%. If the correctness of the code matters (which it will as the codebase grows), you still need to read the code, and that still takes human eyes.
Na-ion cells have roughly half the volumetric and gravimetric energy density of NMC, so it's double the weight and double the space. Apart from still being at least as—if not more—expensive as LFP, they also have a sloping voltage curve, vs lithium with is relatively flat, which poses problems for voltage conversion, and these engines are going to be taking kilovolts of power. So I think those problems would need to be solved first.
I'd imagine it makes providing constant AC voltage to the engines pretty tough. It's even a problem for home energy storage because you need an inverter that can handle a very wide range of input voltages. Most inverters will cut off well before the Na-ion battery is full drained, vs a lithium which can go pretty much all the way to 0%.
I don’t know much about the kind of giant motor that would be used in a ship, but as a general principle: the load that a motor presents to the drive electronics does not resemble your house in the slightest.
To power your house (or, more generally, supply vaguely sine-wave like output at a constant voltage), you need a converter that will convert DC at the battery voltage to AC at the desired voltage. If a buck converter is used, for example, the AC voltage can only ever be lower than the battery voltage. If you use a cheap square wave inverter, it’s possible that the output and input voltages must actually be equal.
A motor, though, is a highly inductive load, and large motors will and do operate from truly gnarly supply waveforms as long as the current waveform is approximately correct. Industrial VFDs (variable frequency drives) do unspeakable things involving switching a DC bus voltage across the motor via H bridges at tens of MHz, which is a horrible thing to do the the wiring between the drive and the motor if it’s not extremely short. (There are, recently, some guidelines that specific types of wire with twisted conductors, better than average insulation, and high quality shields should be used to improve tolerance of the fact that rather impressive standing waves can appear in the wiring if the wiring is a quarter wavelength or longer.). I can easily imagine designing a VFD that works just fine over a respectable range of DC input voltages by adjusting its duty cycle accordingly.
One way to think of this is that a VFD looks kind of like a buck converter where the inductor is free in the sense that it’s already right there in the motor. If it’s designed right, it will handle the battery’s full voltage range, and the inductor will still be free :)
> I can easily imagine designing a VFD that works just fine over a respectable range of DC input voltages by adjusting its duty cycle accordingly. [...] I can easily imagine designing a VFD that works just fine over a respectable range of DC input voltages by adjusting its duty cycle accordingly.
I imagine it's not the waveform or current that matters so much, as the voltage. These motors would be powering massive blades encountering incredible resistance, so you need megavolts to move them, with an input voltage all the way down to near zero.
> H bridges at tens of MHz
Imagine the MOSFETs on this thing! Do they have something that scales up to MV? That sounds like an engineering challenge in itself.
Full disclaimer: electronics is not my wheelhouse, though I have played around with motor controllers.
Right. One instance of metal theft in any country is enough to discredit the argument. As someone who lives in Australia, I've seen it show up in the news here just once. And I've spent time in other first world countries including the US, so my opinion doesn't come from a place of ignorance.
I live in New Zealand and I see or hear of thefts in my old neighbourhood (Woolston). Yet I can't recall anything about the issue on the news.
Every bit of reliable information I've had points to Meth users (although many years ago I knew of opiate users trying to get copper).
I was recently in New Orleans and had two theft surprises (one positive, one negative:
1: walking in Gretna I noticed an aluminium ladder under a house. You don't leave them visible at home because they get stolen (I presume for metal)
2: an Uber driver pointed out the theft of Aluminium guard rails. Obviously missing at road side. Needed grinders since they were welded infrastructure. I haven't seen much of that level of theft in Christchurch yet.
As a further point of contrast between the US and Oceania, the kind of copper theft happening in the OP are disabling communications systems:
> From January to June of this year, 9,770 incidents of intentional theft or sabotage on communications networks were reported, according to the Internet & Television Association, a trade group known as NCTA. That is nearly double the number reported in the prior six-month period. The attacks disrupted service for more than eight million customers.
> The cut lines have disrupted 911 emergency calls and internet and landline services, shut down at least one school and left whole city blocks in the dark.
... and to contrast, here in Australia, the mobile networks hadn't properly QAed their production deployment and access to emergency services causing an outage for 13-hours, likely causing people to die. The response was immediately establishing an inquiry and enacting new laws within 31 days of the incident.
In the past few years I've seen so many instances of day-light smash and grabs happening in the US in broad daylight filling my feeds. I've been waiting for the other shoe to drop – will that wave of criminality reach here? Certainly there's been an significant uptick of knife crime in Australia, people with mental health issues shooting cops and stabbing civilians, and the housing crisis is really causing problems around homelessness, but even still, Australia has nothing on the brazenness and scale of what is happening in the US. So when I say that third world problems are affecting the US, I refer to this broader situation.
So three instances in the EU for the year. I suppose that's something. All I know is I don't see people with angle grinders taking down local infrastructure in Australia. Maybe if we matched in population it would be the same? I don't know.
But I'm checked out of this argument, because I'm certain anyone who's lived in the US and, say, comes to Australia, would recognise the stark difference in social outcomes and crime levels. But why reflect on that at all? It's far easier to quibble over minutiae.
Nobody even implied coverage. You have three more examples than before.
> I'm certain anyone who's lived in the US and, say, comes to Australia
Live in America. Have come to Australia. Missing your point. You guys have your own conniptions you dial the social anger scale to 11 over. But! You’re typically better than this at presenting and refining views in response to evidence.
So let me clarify. Here's something I said in a comment to you:
> I've seen it show up in the news here just once
So I'm aware that happens even here in Australia, and this has not been my argument from the start. Yes, crime still happens in first-world countries. So it's not that it also happens also outside the US, it's something else: the quality and the quantity.
Now–you should know this, and I don't mean to be condescending, but to be clear–there's a difference between relative and and absolute measures. Like for instance, there's some ~450 homicides in Australia every year. That means that every death basically makes the broadcast news segment. But then Australia only has ~27.2 million people. So to make a fair comparison to another country, that's about 1.6 homicides per 100K people (vs the 6.81 in the US)
So talking about copper theft, the 3 data points there I can't really quantify as copper thefts per X population. I also can't qualify the difference to the daily lives of people. Many of the data points provided outside the US seems to be targeted at business places and depots, whereas the OP article which you posted talks about something much closer to home:
> From January to June of this year, 9,770 incidents of intentional theft or sabotage on communications networks were reported, according to the Internet & Television Association, a trade group known as NCTA. That is nearly double the number reported in the prior six-month period. The attacks disrupted service for more than eight million customers.
> The cut lines have disrupted 911 emergency calls and internet and landline services, shut down at least one school and left whole city blocks in the dark.
Do you see the point I'm trying to make here? If something so brazen and on this scale was happening in Australia, it would be nipped in the bud. I strongly believe that's because the overwhelming majority of Australians feel they have a stake in the country, and, relatively speaking, this country (for all its foibles, and all our misgivings of it) looks after us.
Not just America. People are stealing copper in very rural areas in my country; in many cases the price they get is hardly paying for the petrol to drive there. We have a whole team now in my company dedicated to repairing damage from copper theft, it's rampant.
I play the NYT connections pretty regularly. Today's (#25332) purple group was slightly more tenuous than real NYT games. Often good purple groups are both difficult to infer, but also the grouping makes more sense. Good connections puzzles will have fake-out groupings, forcing you to try and find as many groups as possible before making a guess. But still I was impressed by this one!
But on my second puzzle (#25049) I had two groups all come up in a row, I don't think the NYT one ever does this – which now makes me think you could shuffle their puzzle a bunch of times to find out the connections. And these connections were dead simple apart from purple which didn't make sense. And "Things You Container" is not correct English.
It's not bad for a vibe-code. But I'd be inclined to stick to the lovingly human-created NYT puzzles.
I think the lesson here is to get feedback early and often. Do an open beta from the start, or at least focus groups where regular members of the public can give opinions on the product as it evolves.
Indeed - this should've been easily possible by having a beta subdomain pointing to the new implementation and let people comment on it to get early feedback.
First I've heard of it. Might have helped to not just listen, but to actively advertise the beta via a link on the main website. Pretty standard practice.
They could also just be people with bills to pay who are maybe faced with—by some accounts—a very challenging employment market. Or maybe due to disabilities they find the process of finding new work difficult or impossible.
That is fine, but they adopt or delegate corporate opinions onto others. I feel that if you need to lie to people because of money, your job is not honest. (I don't mean you; I mean people who need to do this because otherwise they may lose their job etc...)
Except for the "disabilities" part, which is problematic to classify, wouldn't your description broadly fit the word "losers"?
EDIT: I don't understand the downvotes. It's not a value judgement on Github employees, it's about the meaning of the word "loser". Go back to your teenage years. What's a loser? Someone, often through no fault of their own, keep being in a bad situation, having the "short end of the stick". What characteristically makes them losers is that they lack the audacity to snap out of it.
Isn't that an accurate definition of what "loser" generally means?
> Isn't that an accurate definition of what "loser" generally means?
"Loser" is a catch-all taunt that bypasses empathy. But certainly they might be 'in a losing situation', which is an important distinction.
> Except for the "disabilities" part, which is problematic to classify
Disability in this context is something intrinsic to the person (e.g., physically, mentally) that makes the employment process substantially difficult to engage with.
In addition to disability, difficulty can also arise do to any prejudice that might be levelled against them (e.g., ageism, sexism, junior vs senior, skin color, language skills, country of origin), as well as visa consideration, financial situation, etc. There's so many things that affect the risk calculus of changing jobs.
I'm inclined to agree. But occasionally, someone is so dead set on talking past every point made and or is just posting flame-bait. Engaging with them doesn't help.
I think at least one chance for them to clarify their position, or genuinely answer to critique should be given. If it's clear that it's just a drive-by low-effort comment, or if they just have an axe to grind, only then should they be downvoted. If they are serial offenders, they typically get shadow-banned.
https://jamanetwork.com/journals/jamanetworkopen/fullarticle...
> Following a downward trend until 2007, suicide rates significantly increased 8.2% annually from 2008 to 2022, corresponding to a significant increase in the overall rates between 2001 to 2007 and 2008 to 2022 (3.34 to 5.71 per 1 million; IRR, 1.71)
reply