Greg's recent activity

  1. Comment on Rick Astley - Pink Pony Club (Chappell Roan cover, 2025) in ~music

    Greg
    Link
    I did not expect to see this in my feed, but it put a much needed smile on my face and it’s a genuinely good cover!

    I did not expect to see this in my feed, but it put a much needed smile on my face and it’s a genuinely good cover!

    7 votes

  3. Comment on US Department of Government Efficiency plans to rebuild Social Security administration codebase in months, risking benefits and system collapse in ~society

    Greg
    Link Parent
    Yeah, that end bit about testing seems like a very easy route towards punitively reducing some or all benefits and destroying the ability to generate the authoritative calculations that anyone can...

    Yeah, that end bit about testing seems like a very easy route towards punitively reducing some or all benefits and destroying the ability to generate the authoritative calculations that anyone can point back to.

    What do you do when the US government owes you money and their shiny new computer system says they don't? Sue, I guess, but that was out of reach for most people at the best of times - now it's out of reach, and won't have accurate records to compare to, and will just be ignored even if a judge does rule in your favour.

    I couldn't guess between malice and incompetence here; if there's one thing I've learned from those Signal chats it's that there's so much incompetence that even the things that look like an evil masterplan might just be stupidity. But there's also so much malice that I wouldn't put any deliberate cruelty beyond them either. The outcome is just as bad either way though, so it probably doesn't matter.

    6 votes

  4. Comment on Amid calls for sovereign EU tech stack, Swedish startup Evroc raises $55M in Series A funding to build a hyperscale cloud in Europe in ~tech

    Greg
    Link Parent
    That wouldn't be my take on it, but I do see where you're coming from. We do at least seem to have reasonable adults on both sides of the table for this one, so hopefully it works out fairly for...

    That wouldn't be my take on it, but I do see where you're coming from. We do at least seem to have reasonable adults on both sides of the table for this one, so hopefully it works out fairly for everyone involved.

    2 votes

  5. Comment on Things progressives get wrong in ~society

    Greg
    Link Parent
    It sounds like we're very much in agreement on why there's no meaningful voting option for economic reform, in that case! I guess I interpreted your original post a bit more along the lines of...

    It sounds like we're very much in agreement on why there's no meaningful voting option for economic reform, in that case!

    I guess I interpreted your original post a bit more along the lines of "the progressive focus is misplaced and needs to be redirected" and less "there's a top-down effort denying progressives the opportunity to focus on one of the most important issues" - and to be clear I think both are true to an extent, but without acknowledging the latter it's easy to forget that progressive voters are the ones trying, and sometimes even succeeding, when it comes to giving those struggling people healthcare, and education, and housing, and a broad economic safety net.

    6 votes

  6. Comment on Things progressives get wrong in ~society

    Greg
    Link
    If I were to really boil it down it seems like the old rallying cry of "no war but the class war" is overly reductive, but placing the culture war above the class war is counterproductive. The...

    If I were to really boil it down it seems like the old rallying cry of "no war but the class war" is overly reductive, but placing the culture war above the class war is counterproductive. The question that leaves me with, and one I genuinely don't know the answer to, is whether putting the former over the latter is actually happening and if so why?

    Is it that progressive voters are genuinely more concerned by social issues than broad economic reforms? Is it that the media just makes it seem that way by what they choose to highlight? Or is it that voters have no meaningful option to vote for broad economic reform anyway, so the best they can do is support social progress? If I were to guess I'd say more of the latter two options, with a sprinkling of the first.

    In a little under two years (in the US) we're going to have the opportunity to come together and swing the pendulum back the other way.

    You've got a lot more faith than I do in the ongoing integrity of the US democracy under Trump. I'm not suggesting he'll suspend voting per se, but I'm absolutely saying it'll be hard to trust the results.

    Leading up to that the focus should be on things that unite us, not things that divide us. And the biggest thing that unites us is that we're tired of our capital controlled political systems. We're tired of politicians that are in it for their donors rather than their constituents. We're tired of the top .01% siphoning off more and more of the resources. That's straightforward and easy to understand and it will absolutely resonate.

    Republican voters are cheering for a billionaire who's stripmining the country at their own expense. So, so many of those struggling people voted for the party who'll make their struggles infinitely worse rather than the one who, admittedly, probably would've kept them roughly the same. I hope you're right that the message can get through, I truly do, but I don't believe it.

    13 votes

  7. Comment on Amid calls for sovereign EU tech stack, Swedish startup Evroc raises $55M in Series A funding to build a hyperscale cloud in Europe in ~tech

    Greg
    Link Parent
    It does involve EU funds going to the UK, and I’m absolutely not going to naively pretend there are no economic incentives for the UK to get it signed. But I do believe it’s not just economic...

    It does involve EU funds going to the UK, and I’m absolutely not going to naively pretend there are no economic incentives for the UK to get it signed. But I do believe it’s not just economic incentives either - the war in Ukraine, and defence more broadly, is one of the few areas the UK actually has been a reliable ally and shown genuine leadership already; logistics and the free flow of funds is a necessary part of making preparations for whatever might happen next, and excluding one of the biggest players from a significant portion of that would add friction and close off options, regardless of NATO. I genuinely believe it matters, completely independent of money or nationalistic “winning”.

    On an emotional level, I was absolutely furious with Johnson for his insistence on making fishing a major issue during the whole Brexit farce (and for almost everything else he and his administration did, to be honest). It was an absolute non-issue then, it’s a rounding error in UK GDP (0.03%, according to a quick search), and it was pretty clear he only pushed it as something big because he felt it might actually be possible to paint as a win for his party when all the significant stuff was already understandably and correctly lost because the UK held no cards and the whole idea was idiotic in the first place.

    So, the point I find us at now is wringing my hands and saying “this was a petty distraction that I was decrying Johnson for focusing on over actual important issues, the new UK government is finally starting to move forward and act like adults with real challenges to address, and now France is focusing on it over more important issues?!”. I genuinely do not care about “UK interests” vs “EU interests” as separate things - I’d be happier to see the UK back in the EU under fair, equal terms than I was with the UK having all those one sided exemptions - and I see a focus on fishing in particular here when the EU is pushing it through exactly the same lens I did when the UK was pushing it.

    If it comes to it and it’s truly a blocker, sure, just give the EU the damn fish. It’s too small to matter and there are much bigger issues at stake, so that’s the pragmatic choice. But forcing that choice in the first place by holding a bigger and much more important issue hostage behind something petty leaves a bad taste when the EU does it just the same as when the UK (and specifically the previous UK government) was doing it.

    2 votes

  8. Comment on Amid calls for sovereign EU tech stack, Swedish startup Evroc raises $55M in Series A funding to build a hyperscale cloud in Europe in ~tech

    Greg
    Link Parent
    I'm not suggesting the EU should concede, I'm just saying that tying a high impact, urgent issue to a relatively niche economic special interest doesn't seem wise from a practical perspective, a...

    I'm not suggesting the EU should concede, I'm just saying that tying a high impact, urgent issue to a relatively niche economic special interest doesn't seem wise from a practical perspective, a diplomatic one, or an optics one. Sign the defence treaty now, worry about fishing afterwards, that's all.

    2 votes

  9. Comment on Amid calls for sovereign EU tech stack, Swedish startup Evroc raises $55M in Series A funding to build a hyperscale cloud in Europe in ~tech

    Greg
    Link Parent
    As far as I’m aware CoreWeave’s business is basically just being a GPU-focused cloud provider - I haven’t heard of them doing anything wildly unusual with the actual tech stack (although I may...
    • Exemplary

    As far as I’m aware CoreWeave’s business is basically just being a GPU-focused cloud provider - I haven’t heard of them doing anything wildly unusual with the actual tech stack (although I may have missed it, I haven’t used their platform beyond a brief trial a year or so ago).

    As I understand it they’re notable for two reasons: firstly, they’re using their GPUs as collateral to finance loans for ongoing expansion, which honestly makes a lot of sense to me and is pretty common with capital assets in other industries already. Secondly, they’re relatively new and managed to catch the hyperscalers a little flat footed a couple of years ago when GPUs were in short supply and the initial wave of AI/ML hype was taking off - they had plenty of supply already negotiated with NVIDIA and in the pipeline for customers while Google and Amazon weren’t offering H100s for love nor money and Microsoft was throwing whatever they did have at OpenAI.

    5 votes

  10. Comment on Amid calls for sovereign EU tech stack, Swedish startup Evroc raises $55M in Series A funding to build a hyperscale cloud in Europe in ~tech

    Greg
    Link Parent
    You’re right, it is, and I’m vehemently in favour of closer European integration and I don’t blame the EU as a whole for remaining cautious after the UK fuck ups of the last decade. Buuuut looking...

    You’re right, it is, and I’m vehemently in favour of closer European integration and I don’t blame the EU as a whole for remaining cautious after the UK fuck ups of the last decade.

    Buuuut looking at the UK’s most serious attempt to move forward by joining a shared defence pact, and blocking it over fish of all things, really isn’t a great look right now either: https://www.politico.eu/article/uk-eu-defense-pact-really-does-depend-on-fish-european-minister-warns/

    4 votes

  11. Comment on New breakthrough in AI cancer detection is pushing accuracy levels to an unprecedented 99% in ~health

    Greg
    Link Parent
    Yeah exactly, that's specifically the drop in incidence attributed to the HPV vaccine, I added it as a recent-ish (last 20 years) example where a single breakthrough really did make a massive...

    Yeah exactly, that's specifically the drop in incidence attributed to the HPV vaccine, I added it as a recent-ish (last 20 years) example where a single breakthrough really did make a massive change. I didn't write out the whole tangent that was in my head, but it was pretty much "most of the progress comes from cumulative advancements and those really add up over time, but big wins in a single vertical do sometimes happen too".

    7 votes

  12. Comment on The Donald Trump US administration accidentally texted me its war plans (gifted link) in ~society

    Greg
    Link Parent
    I get you, and if it helps at all I am in no way interested in giving this guy any kind of absolution. I know your question was probably rhetorical, but honestly my primary drive for following the...

    I get you, and if it helps at all I am in no way interested in giving this guy any kind of absolution. I know your question was probably rhetorical, but honestly my primary drive for following the thread of possibilities here is curiosity. It’s a big, important, fucking insane situation involving some utterly huge coincidences - trying to piece it back together helps scratch that itch it’s creating at the back of my brain.

    6 votes

  13. Comment on The Donald Trump US administration accidentally texted me its war plans (gifted link) in ~society

    Greg
    Link Parent
    Balance of probabilities I still agree with you, it was probably just a genuine mistake, but with @psi's point in mind I do think it's harder to be sure. If he had a brief flash of conscience but...

    Balance of probabilities I still agree with you, it was probably just a genuine mistake, but with @psi's point in mind I do think it's harder to be sure. If he had a brief flash of conscience but still wanted to protect himself to the extent he goes down as "fuck up" rather than "traitor" in Trump's eyes, playing it as a mistake would be the way to go and we'd have no way of knowing.

    Exposing the chats seems a lot more powerful than just refusing to participate, and more likely to trigger the kind of outcry that could at least possibly lead to some external oversight, too. I don't have a lot of hope there, but still, better than us not knowing, because it's not like those lives were going to be better protected otherwise.

    Guy's still chosen to be part of the fascist administration, I've got no sympathy for him either way and no reason to think it was a change of heart, I can just see a more plausible path for it to have played out the same way if it were.

    1 vote

  14. Comment on New breakthrough in AI cancer detection is pushing accuracy levels to an unprecedented 99% in ~health

    Greg
    Link Parent
    No worries! There's so much marketing crap to wade through that I tend to approach these claims with a bit of an eye roll too, this just happened to be one I'm already familiar with.

    No worries! There's so much marketing crap to wade through that I tend to approach these claims with a bit of an eye roll too, this just happened to be one I'm already familiar with.

    8 votes

  15. Comment on New breakthrough in AI cancer detection is pushing accuracy levels to an unprecedented 99% in ~health

    Greg
    Link Parent
    I do know what you mean, the breathless headlines and sensationalist reporting help nobody, but progress really is happening. We’ve had a 33% drop in overall cancer deaths since the 90s, and a 65%...

    I do know what you mean, the breathless headlines and sensationalist reporting help nobody, but progress really is happening. We’ve had a 33% drop in overall cancer deaths since the 90s, and a 65% drop in cervical cancer specifically in the 2010s, for example.

    This paper looks like a genuine improvement on what we had before, and I’d make a solid bet on it saving lives. But the bit the article doesn’t mention is that we already had pretty damn good automated detection, and we already had trained radiologists who were and are excellent at doing the job manually. This isn’t zero to cure, but it’s a few percentage points better than we had, which in turn gets a few percent of people treated earlier, which in turn improves their prognosis by a few percent.

    These aren’t the kind of changes you can really see other than in long term statistics across whole populations. But every little step along the way means a few more people alive who otherwise wouldn’t be.

    19 votes

  16. Comment on The Donald Trump US administration accidentally texted me its war plans (gifted link) in ~society

    Greg
    Link Parent
    Very good point actually, particularly that knowing in advance it was genuine could've altered Goldberg's legal position too - maybe even put him in a situation where he would be compelled to...

    Very good point actually, particularly that knowing in advance it was genuine could've altered Goldberg's legal position too - maybe even put him in a situation where he would be compelled to report it to the authorities as soon as he was informed.

    I guess there's probably a decent amount here that only Waltz will ever know for certain, which bothers my innate curiosity, but at least we're better off than if we didn't even know there was a classified group chat to be curious about in the first place.

    9 votes

  17. Comment on New breakthrough in AI cancer detection is pushing accuracy levels to an unprecedented 99% in ~health

    Greg
    Link Parent
    Healthy skepticism is always sensible, but this looks like actual scientific research done on existing real world data to me. The paper is here (link in the article just points to the journal for...
    • Exemplary

    Healthy skepticism is always sensible, but this looks like actual scientific research done on existing real world data to me. The paper is here (link in the article just points to the journal for some reason), and references the datasets they used. It's also worth bearing in mind that these numbers are good but not implausible - they mention existing papers in related fields with accuracy in the 90%+ range, and general purpose (non-medical) image classification has been close to 100% on some datasets for a good few years now.

    The 99% number isn't from pretrained data in the way you might be thinking. The standard approach here is called k-fold validation, where you split the dataset into chunks and then measure performance on the chunk the model hasn't seen - often you'll split it into five groups, train on four of those, and measure performance on the fifth unseen one. Repeat that four more times and you've got an average for how well it'd perform if the full dataset hadn't previously been seen by the model. Those additional results you mentioned are then from testing the same architecture in the same way but on entirely different datasets, to see how it generalises to cancers that it wasn't specifically designed to handle.

    Obviously this is still susceptible to the data as a whole being biased or unrepresentative, but they're using multiple pre-existing peer reviewed datasets, so that's fairly unlikely to be a confounding factor. It seems like a genuine improvement in the state of the art, from research done in good faith by the academic community.

    33 votes

  18. Comment on The Donald Trump US administration accidentally texted me its war plans (gifted link) in ~society

    Greg
    Link Parent
    I think you're probably right, but I will say that having a journalist watch it unfold in real time is a lot more impactful - for the journalist themselves, for the story that motivates them to...

    I think you're probably right, but I will say that having a journalist watch it unfold in real time is a lot more impactful - for the journalist themselves, for the story that motivates them to write, and for the headlines it generates. It'd be a lot easier to cast doubt on screenshots, too, compared to having an independent third party witness it all firsthand. The fact that the chat exists at all is already a crime, so that wouldn't strike me as a dealbreaker either.

    Honestly, if this were somehow a whistleblower or deliberate leaker it's exactly what I've been asking for when I say people on the side of freedom and the rule of law need to understand the media better and start playing hardball!

    Including telling the reporter why he was invited in a DM, before inviting him.

    This is the key, for me. Give the reporter a heads up to protect yourself a little. Hell, send a separate message from a burner phone saying "The messages you're seeing are real, please treat the info on who added you as an anonymous source" if you want to maintain plausible deniability for it being nothing more than a mistake when the group members figure out you were the one who added the journalist.

    Wishful thinking is nice and all but the man signed up for the job, it's worth assuming he wants to have it. We keep looking for some principled savior and I really think we have to stop painting that image onto people and squinting in hopes it lines up.

    Yeah. It's almost a mirror universe Hanlon's razor - no need to assume noble reasoning when we live in a Four Seasons Total Landscaping world of incompetence. Even if this were somehow the point they chose to draw the line (and I don't think that's likely), it doesn't absolve them for all the blindingly obvious lines they happily crossed before this.

    Could maybe have been a staffer with access to the device acting as a genuine whistleblower, I guess? That'd more explain the lack of a defensive request for anonymity or anything like that. It'd also be a beautiful example of why you maintain proper fucking security measures around your classified communication systems and devices in the first place.

    14 votes

  19. Comment on EU paves the way for iPhones and Android devices to ditch USB-C entirely in ~tech

    Greg
    (edited )
    Link Parent
    Scaling up to larger distances (between cities) it’s worthwhile to go wireless if you’re doing something very latency sensitive, so I can imagine people hearing that snippet and remembering it as...

    Scaling up to larger distances (between cities) it’s worthwhile to go wireless if you’re doing something very latency sensitive, so I can imagine people hearing that snippet and remembering it as a broader “wireless = lower latency for the most demanding applications”.

    High frequency trading gets a measurable boost from the extra propagation speed and the straight line distance (compared to the cables being installed to follow the actual geography) at the tens to hundreds of km range, and that demonstrably outweighs any encoding/decoding overheads, but for a 3m mouse cable you’re talking nanosecond differences in travel time for a theoretical best case. That maps to tens of GHz in data processing terms - within the capabilities of specialised tech, because we live in the future, but below the noise floor for general computing and way outside human perception. [Edit: come to think of it I wonder if sticking with wires but switching back to good ol' PS/2 to avoid the USB stack would make a difference? I doubt it'd be perceptible, but it might at least be measurable... Anyway, I should be working right now!]

    Tom Scott did a video about an exchange that uses a 60km spool of fiber as an intentional speed bump, if you’re interested in a real world example of this kind of thing.

    3 votes

  20. Comment on EU paves the way for iPhones and Android devices to ditch USB-C entirely in ~tech

    Greg
    Link Parent
    Oh that’s cool, I hadn’t done much digging since I tried to hunt down an Android with magnetic wireless a few years ago. Good to know it’s getting on track! Interestingly, searching around Qi2...

    Oh that’s cool, I hadn’t done much digging since I tried to hunt down an Android with magnetic wireless a few years ago. Good to know it’s getting on track!

    Interestingly, searching around Qi2 does bring up a few articles about Apple’s involvement in defining that standard and some musing on why they’ve chosen not to fight it to “protect” MagSafe for themselves, so it sounds like a change of position from them might have been what got it across the line.