Greg's recent activity
-
Comment on Nvim 0.12 released in ~comp
-
Comment on Nvim 0.12 released in ~comp
Greg Link ParentIt’s a meme, but there’s also a pretty large kernel of truth in most semver projects actually incrementing based on how proud the developers are of a given checkpoint. In my experience it also...It’s a meme, but there’s also a pretty large kernel of truth in most semver projects actually incrementing based on how proud the developers are of a given checkpoint. In my experience it also maps to the fact that deep fundamental unix packages like this never hit a 1.0 release - the kind of mind that’ll work on a fork of a reimplementation of a terminal text editor from 1976 and do a damn good job of it is never satisfied with their own work!
-
Comment on Google’s TurboQuant AI-compression algorithm can reduce LLM memory usage by 6x in ~tech
Greg Link ParentIt’s a published paper, that means others can and will reimplement it - and likely pretty quickly, given the speed of OSS development in the field. You might be interested in the links that @tauon...It’s a published paper, that means others can and will reimplement it - and likely pretty quickly, given the speed of OSS development in the field. You might be interested in the links that @tauon mentioned further up, too? Those are available here and now, and the results look seriously impressive!
-
Comment on Everyone but US President Donald Trump understands what he’s done - allied leaders know that any positive gesture they make will count for nothing in ~society
Greg Link ParentAgreed on all counts, and picking over the bones of the empire on its way down is a very apt description of the situation. I think it’s totally reasonable to assume Trump himself is fundamentally...Agreed on all counts, and picking over the bones of the empire on its way down is a very apt description of the situation.
I think it’s totally reasonable to assume Trump himself is fundamentally compromised by Russia, I just also think that there are enough other influences in play - on him as a person, and on what his government does outside of his personal control - that it’s tougher to untangle or predict the actions and outcomes as a whole.
-
Comment on Everyone but US President Donald Trump understands what he’s done - allied leaders know that any positive gesture they make will count for nothing in ~society
Greg Link ParentI’ve been meaning to come back to this, because I think it deserves a proper reply! The short version of my thinking is that it’s probably not as clear cut as there being any single agenda. I’m...I’ve been meaning to come back to this, because I think it deserves a proper reply! The short version of my thinking is that it’s probably not as clear cut as there being any single agenda. I’m seeing it as a gestalt of different factions and interests, with the one necessary linking factor between them being an abject lack of empathy. If that’s the case, it means faction A won’t oppose faction B doing something destructive unless that thing also happens to interfere with faction A’s goals - and even then it’d have to interfere enough that it’s worth the political capital to fight it rather than just let it slide.
I agree with you that it’s not the case that things are playing out as they are by accident, basically, but I also think it’s not the case they’re all reading from the same sheet - each group is just seeding ideas where they can, capitalising as far as possible on random whims when they get blurted out, and ignoring most of the rest of what happens because they simply don’t care.
Add in factions C, D, E, and so on, plus the fact that most of the people involved will absolutely throw their own faction under the bus if it’s sufficiently in their individual interests to do so, and you end up with a situation that I think is risky to try and predict using any single lens.
This is why I was talking about partial reasoning: I think there’s absolutely an “advancing Russia’s interests by weakening America” faction, and I think there are also plenty of useful idiots doing so unknowingly across the whole morass, and I think there are a lot of people who’ll look at something like the Iran war, realise it’s going to weaken the US, and see that as nothing more than a chance for some quid pro quo to advance their own agenda so not have any reason to stand in the way.
Russia gets their splintering of global power, Hegseth gets his holy crusade, Trump gets his narcissistic feeling of power and strength, whoever’s making those massive insider trades before the tweets go out gets even richer, Thiel gets his biblical end times, the Epstein society get their ongoing protection, etc etc
None of them really care who’s driving or how many thousands of people get run over along the way - and I think none of them are driving, they’re all just shouting different directions over the top of each other and making the occasional grab for the wheel when they can - as long as they’re safe inside the vehicle and it’s not actively going away from their particular destination.
-
Comment on Google’s TurboQuant AI-compression algorithm can reduce LLM memory usage by 6x in ~tech
Greg Link ParentYeah, that absolutely seems plausible - will be interesting to see how (and if) models from the big players change as this newly published stuff cross-pollinates between them and mixes with...Yeah, that absolutely seems plausible - will be interesting to see how (and if) models from the big players change as this newly published stuff cross-pollinates between them and mixes with whatever tricks and techniques they each haven’t published!
-
Comment on Google’s TurboQuant AI-compression algorithm can reduce LLM memory usage by 6x in ~tech
Greg Link ParentIn which case used server farm RAM prices really go down, it’s a win-win! Win-win not guaranteed, management takes no responsibility for job losses, riots, rising fascism, economic collapse,...In which case used server farm RAM prices really go down, it’s a win-win!
Win-win not guaranteed, management takes no responsibility for job losses, riots, rising fascism, economic collapse, plagues of locusts, or indigestion.
-
Comment on Google’s TurboQuant AI-compression algorithm can reduce LLM memory usage by 6x in ~tech
Greg (edited )LinkBetween this from Google (it’s a very significant optimisation to KV caching, for anyone following along), and DeepSeek’s “Engram” conditional static memory lookups, it’s looking like there’s...Between this from Google (it’s a very significant optimisation to KV caching, for anyone following along), and DeepSeek’s “Engram” conditional static memory lookups, it’s looking like there’s another decent step up in model performance coming. I’m actually a little surprised to be saying that, I was starting to think we were plateauing a bit and potentially resorting to slightly more brute force approaches to making improvements, so it’ll be interesting to see how much more steam there is in this broad class of LLM architectures!
[Edit] DeepMind’s recent paper about doing actual, publishable mathematical research with an LLM is very significant too IMO. This isn’t some beam search brute force approach to constructing a technically novel but ultimately unreadable thousand page proof, it seems to be a research team genuinely creating new knowledge with the aid of an LLM, which strikes me as an important tipping point in capability.
-
Comment on Britain mandates heat pumps and solar panels in new homes from 2028 in ~enviro
Greg Link ParentLarger surface area and thermal mass, to compensate for smaller temperature differential (bearing in mind a lot of places use hot water running through radiators, not forced air like much of the...Larger surface area and thermal mass, to compensate for smaller temperature differential (bearing in mind a lot of places use hot water running through radiators, not forced air like much of the US).
Bigger radiators are the more standard option, and not quite such a big job to install as underfloor, but it’s still a pretty significant replacement of the heating system as a whole rather than just the heat source.
-
Comment on OpenAI shuts down Sora AI video, Disney drops planned $1B investment in ~tech
Greg Link ParentI find it sad that the companies in control of tech I find fascinating are such unmitigated arseholes that everyone's celebrating their downfall. And to be very clear here: I don't think anyone's...I find it sad that the companies in control of tech I find fascinating are such unmitigated arseholes that everyone's celebrating their downfall. And to be very clear here: I don't think anyone's wrong to be pleased, I just see it as yet another "this is why we can't have nice things" when the incentives favour organisations that justify that reaction.
I watched us go from text interfaces to blocky 16 colour sprites to Mario to Sonic to Jurassic Park to Toy Story to Tomb Raider to Skyrim to Avatar. Transformative jumps in quality happened every few years, and then went from huge studio projects to rendering in real time at home a few years later, and my jaw dropped every time. I had the same reaction to seeing a computer create photorealistic video from nothing but a rough description barely two years back, even if it does have the occasional imperfection.
To anyone who's followed or worked on computer generated imagery in the 90s and 2000s, it's a genuinely stunning technical achievement - doubly so considering it came only a decade after even getting a computer to recognise a bird in a still image was an unreasonable ask, let alone asking it to create video of one in flight from scratch.
But yeah. I'm starting to believe none of that really matters. We as individuals are seeing fuck all benefit from it, the astronomically wealthy are using it as yet another dice game with the economy, and their minions are using it as yet another avenue to spread misinformation and set us against each other.
I've said many times what incredible technical achievements these models are, and I stand by that. Some of the researchers involved have made enormous advancements in our understanding of information theory, statistical modelling, and both physical and computer science more broadly.
So I'm sad for them, and even more so for what it says about our society as a whole, that the incentive structure we've built is to use all of that for output that's legitimately called slop 99% of the time. I'm sad that we've so thoroughly given up on any advancements being used for the greater good that we celebrate their downfall, and that we're probably right to do so.
-
Comment on Everyone but US President Donald Trump understands what he’s done - allied leaders know that any positive gesture they make will count for nothing in ~society
Greg (edited )Link ParentI think it’s actually really important to hold on to this as at least partial reasoning for anyone trying to understand what’s going on here. Some of his actions make sense in the context of...I think it’s actually really important to hold on to this as at least partial reasoning for anyone trying to understand what’s going on here. Some of his actions make sense in the context of enriching certain industries or individuals, and that’s almost certainly because they’re taking advantage of his corruption and gullibility even if they didn’t engineer the situation, but other things he does just don’t track even then.
I think part of it is straightforward stupidity and now advancing cognitive decline too, but a decent amount of his actions do also make sense if you contextualise them as “what decision would weaken the US’s global dominance in this moment?”. If you look at probably-Russia’s interests in terms of rebalancing power by damaging the reliability and credibility of the largest player, it makes more sense who’s likely to be seeding some of the crazier ideas with no obvious payoff even for vested financial interests.
[Edit] Typos
-
Comment on OpenAI shuts down Sora AI video, Disney drops planned $1B investment in ~tech
Greg Link ParentI thought Sora was actually pretty impressive, and definitely a long way ahead of anyone else when they first demoed it (although that was a relative lifetime ago and plenty of others have been...I thought Sora was actually pretty impressive, and definitely a long way ahead of anyone else when they first demoed it (although that was a relative lifetime ago and plenty of others have been catching up). Then again, I’m sure the sales team would find a way to over promise in an absurd way even if the product literally cured cancer, so you could well still be right there!
My bet’s more on risk/reward though - video’s computationally expensive and likely to get them sued by companies with deep pockets, I imagine the economics just weren’t workable for the relatively limited amount end users were willing to pay.
-
Comment on US regulator bans imports of new foreign-made routers, citing security concerns in ~tech
Greg Link ParentIf this were a normal administration it'd be a horrible sign of incompetence and/or malice. As it stands, we already know loud and clear that those are both the case, and this particular idea is...If this were a normal administration it'd be a horrible sign of incompetence and/or malice. As it stands, we already know loud and clear that those are both the case, and this particular idea is so goddamn stupid and shortsighted that it'll probably just fall apart under a little bit of scrutiny.
I know that's not a lot of reassurance, but I don't think this in particular is that much to really worry about given the circumstances. At least not in comparison to a lot of other things they're doing.
-
Comment on US regulator bans imports of new foreign-made routers, citing security concerns in ~tech
Greg Link ParentLooks like Comcast contracts out their own-brand stuff to CommScope, which is a subsidiary of Amphenol and does have at least some manufacturing in the US. No idea whether that’s where this...Looks like Comcast contracts out their own-brand stuff to CommScope, which is a subsidiary of Amphenol and does have at least some manufacturing in the US. No idea whether that’s where this specific hardware is made, but it seems like a possibility for a US-based company that’s going to have to deal with this one way or another?
The problem (well, one of the many problems) with the Trump administration is we can’t even assume it’s straightforward corruption and work backwards from there. It could be that they’re deliberately doing this to benefit a few specific businesses - and I’d bet on Comcast being one of them if so, it’d be entirely their style to cut off the supply of routers to the entire US just to force their own customers into some absurd equipment rental fee. It could be that they’re doing it to extort fees from those businesses to get their hardware on the approved list and the businesses are just as pissed off about it as everyone else. It could be that they’re planning on mandating some kind of absolutely insane surveillance and censorship infrastructure built right into the network hardware and don’t want anyone working around it. Or it could be that someone in the office thought this made for a solid “China bad, America strong” headline and literally didn’t even bother to ask whether there are any US made alternatives before implementing it.
-
Comment on Kill chain - on the automated bureaucratic machinery that killed 175 children in ~society
Greg (edited )Link ParentIt’s not even that things need to be expensive per se - genuine efficiency should be cheaper, you just need to have a willingness to spend up front for savings later (putting capital into newer,...It’s not even that things need to be expensive per se - genuine efficiency should be cheaper, you just need to have a willingness to spend up front for savings later (putting capital into newer, better systems and tools when appropriate, for example) and an understanding that the savings might not be obvious or easy to quantify. Sometimes the payoff is clear - sometimes you really do just get more done in less time with a better tool - but a lot of the time you need to consider things like aggregate total hours wasted for staff and clients before you see that the overall time cost of having more staff is actually lower.
Obviously businesses don’t care about wasting their customers’ time as long as it doesn’t impact profits, but governments should be accounting for that, and accounting for things like “how many hours of people’s lives does the tax money we’re spending represent, and how many hours can we save them by spending it?” when determining what the net cost of a project actually is. In reality we’re lucky to get an analysis that considers the concept of externalities at all in any meaningful way, let alone one that takes the next step and maps it back to intangible benefits rather than assuming monetary value is a benefit in and of itself and not just an approach to keeping score.
[Edit] For what it’s worth, I’m wary of accepting friction in processes as a good thing - inaction can be just as harmful as action, but people tend to have a bias towards doing harm by inaction because it feels like they’re not actually doing harm by being passive - so I’m more inclined towards the “do more, faster, just set your goals and incentives properly” approach in an ideal world. But compared to the status quo of “fast with actively bad incentives” when businesses get involved, I can understand the preference.
-
Comment on Our commitment to Windows quality in ~tech
Greg Link ParentThis is such an important factor! I’m at the “NixOS, terraform, hundreds of cloud instances, tens of physical machines” end of the spectrum and I still stop and consider when to do major OS...you cant drive to work on Monday, a ruined pair of pants, or for installing an OS a computer that doesn't work at all
This is such an important factor! I’m at the “NixOS, terraform, hundreds of cloud instances, tens of physical machines” end of the spectrum and I still stop and consider when to do major OS changes based on how long it’d take out of my day if I need to deal with a weird bootloader issue or new driver version incompatibility or whatever, or how big a problem it’d be to have that machine out of action for a few days if I won’t have time. For people who don’t have the experience of what might go wrong, and definitely don’t have the experience to be confident in fixing it if it does, that’s a much bigger deal!
There are a lot of people who are frustratingly uninformed too, and I’d argue that’s probably driving more general inertia than people who consciously decide to hold off in case of potential problems, but if we can get people to the point they’re considering the choices I think it’s good to acknowledge the potential time and stress costs if things don’t work as expected.
-
Comment on The women leaving the new right in ~society
Greg LinkYup, there it is. Zero morality, zero care for all the other people being hurt thanks to the racism, xenophobia, science denial, corruption, and straight up incompetent handling of major life or...“If I was a guy, maybe I’d still be on the right,” she tells me. “It’s much easier if you’re a white guy not to feel any of this pain. There’s no one planning to disenfranchise you. It’s all upside. It’s all this big program for how you should live a better life and be surrounded by nubile women.” She believed it was human to enjoy being on top. “I don’t want to punish anyone,” she says. “But the reality is: It is stupid, it’s impractical, it has to end.”
Yup, there it is. Zero morality, zero care for all the other people being hurt thanks to the racism, xenophobia, science denial, corruption, and straight up incompetent handling of major life or death decisions on behalf of millions of people. No introspection on the idea that "being on top" might, perhaps, be a hollow and purposeless victory if it comes solely as the result of pushing others down. Not even the slightest hint of consideration that being somewhere in the middle and raising standards across the board could leave you in a better place than a life as the most important person in the toxic crater of your own making. Just a "practical" decision to leave because you personally are being harmed.
I am utterly disgusted with these people and everything they stand for, and I'm terrified of their mentality because I don't have the faintest idea how to even begin engaging with someone who's capable of making decisions with no regard to who'll be harmed as long as it isn't themself.
-
Comment on I hope you don't use generative AI - an essay about my experience offering an open-source tool in ~tech
Greg Link ParentI'm with you on the broad strokes there - anger at copyright and anger at AI are both rooted in anger at the corporate entities using those things to do harm, and frankly people are absolutely...I'm with you on the broad strokes there - anger at copyright and anger at AI are both rooted in anger at the corporate entities using those things to do harm, and frankly people are absolutely right to be pissed off about that. But I think the idea of tying the ethical concerns you're raising back to copyright law as currently implemented is wrong enough that it would actively make things much worse if people followed that through to its logical conclusion.
Model training just doesn't fit within the current IP law framework at all - either legally or conceptually. You say they took the valuable labor without compensation, but the works they were taking are already published on the internet*, there to be distributed. The "taking" isn't the problem, that's what you're supposed to do with something that's been published. What they've done that pisses people off is devalued those works, and I'm yet to see any sensible suggestions for how IP law could have combated that without causing vastly larger problems, even with the benefit of hindsight.
I'm not saying people are wrong to be angry, upset, worried, or anything else about that devaluation. But I absolutely am saying that using the lens of copyright, or even of conceptual ownership more broadly, just doesn't fit the problem. Focusing on whether or not the tech companies were justified, ether legally or morally, in running statistical analysis on publicly available text and images is a distraction at best. It's pretty much guaranteed to descend into a fuzzy argument about where exactly the line is between "acceptable things to do with a publicly available document your computer already has in RAM" and "unacceptable things to with that same document your computer already has in RAM".
It's a question of economics, of power imbalance, of financial value, of social value, and of technical capability. If you start from the premise that it's extraction without credit - a framing that I do quite like - it's a question of what extraction with credit could or should look like.
*Yes, it does appear that some of the companies involved torrented huge numbers of books. The hubris and hypocrisy there pisses me off greatly, particularly because they could absolutely have afforded to just buy the books legally - but nothing I'm saying would actually change in a meaningful way if they'd paid for them like they were supposed to and each author got their $3.27 per book in royalties.
-
Comment on I hope you don't use generative AI - an essay about my experience offering an open-source tool in ~tech
Greg Link ParentI'm hoping we get some more robust approaches to testing for race conditions, deadlocks, and memory leaks sooner rather than later! I feel like those issues have always been sidelined by a focus...I'm hoping we get some more robust approaches to testing for race conditions, deadlocks, and memory leaks sooner rather than later! I feel like those issues have always been sidelined by a focus on unit testing frameworks that don't really have good ways of catching them, and that's going to be an even bigger problem in a "code's a black box, just trust the tests" world than it is now.
-
Comment on I hope you don't use generative AI - an essay about my experience offering an open-source tool in ~tech
Greg Link ParentEven more annoyingly, we absolutely could be using code generating LLMs in a deterministic way, and I think there are likely to be some really technically interesting possibilities down that path...Even more annoyingly, we absolutely could be using code generating LLMs in a deterministic way, and I think there are likely to be some really technically interesting possibilities down that path - it's just unlikely to be commercially viable.
At a guess, fully deterministic models are going to end up in the same conceptual niche as Haskell: interesting to the small subset of us who really care about the mathematical basis of how things work, but probably a bit too constrained to ever get serious traction.
But I'd also guess that the absolute lack of consideration for determinism or even basic predictability when generating code is going to end up being the same kind of ticking time bomb as the lack of memory safety in C (not to say the two things are similar per se, just that I expect them to create the same kind of widespread, slow-burn security and functionality problems). We're going to end up with whole swathes of code that not only has a high risk of difficult-to-spot bugs, but that has no standardised approach to finding and fixing those bugs even if we later discover a way to prevent them.
Huh, that's an interesting one, I didn't know that - perils of being one of the people figuring things out right at the start, I guess! Or perhaps a statement on Torvalds' opinion of himself, if some of his comments to other devs over the years are anything to go on... Probably a bit of both.