It's also worth noting that's US stocks. The story in Europe or Argentina would be completely different, and there was no way to figure that out in 1920.
Terms of service often contain illegal provisions. I teach kids to flout them too. One of the biggest sins in school is kids learn to follow rules uncritically.
There are specific protections allowed when the goal is to maintain / break compatibility. If Microsoft locks competitors out, competitors are quite often permitted to pick the lock.
I can't comment on this situation since I don't know the details, but it's very likely this is fully legal.
See Oracle / Java API lawsuit, garage door opener suit, etc. To see where the lines sit.
I am quite confident that the bandwidth cost is absolutely not a concern for Microsoft, and that the obvious goal is for them to capture the market.
The "C/C++" extension github repository is 4MB. Probably the download size for the extension itself is a fraction of that, but I won't bother measuring. It was downloaded 400 times over the last minute (there is a live counter on the extension page [0]).
That's a 25MB/s or 200Mb/s bandwidth, for one of the most popular extensions. Multiply by the top 10 extensions and you get the bandwidth of an average home optic fiber connection...
I am glad you have insider knowledge to be so confident. I would rather those costs go towards furthering VS Code than helping out Cursor. This comes from someone who uses Cursor and not the biggest fan of MSFT.
Pure speculation but I would see the more logical argument being Cursor is a for pay product, why should they have access to the marketplace?
Because MS didn’t write most of the extensions yet engineered things conveniently such that you have to use their service to get them. Other text editors somehow manage to not lock people into similar dilemmas. They’re not profiting from running the marketplace or providing VS Code for free, it’s about locking people into a product. Cursor should be allowed access because interoperability is a societal net-benefit.
> those costs
…are likely minescule. I run similar services at my day job, just at a much larger scale than a text editor app marketplace, and know the precise cost to run everything. I am often disturbed that people might actually think cost:revenue is tight enough that they should defend a behemoth about callously gating access to it.
> I am often disturbed that people might actually think cost:revenue is tight enough that they should defend a behemoth about callously gating access to it.
I think it's more likely that they imagine themselves in Microsoft's shoes. After all, it's a very popular editor and the mechanism of vendor lock-in is clever - give away the editor under the noble banner of open source, while jealously gate access to the plugin ecosystem that makes the editor as useful as it is.
So no, I don't think they earnestly believe that the egress costs are anything more than pocket change. But it's almost certainly what they would argue if they were in Microsoft's position.
Couldn’t it wind up being easy for Cursor and other variants of VS Code to be long run beneficial for VS Code itself? Seems like having a different third party team extending your stuff and testing it, could be hugely valuable, they take risks and move fast, the upstream project gradually learns from what works for the forks, people contribute various other new extensions.
In the age of LLMs, community is worth its weight in platinum, cutting off Cursor just incentivizes them to develop some new better thing with better technology (cough Zed, Ghostty) to compete with VS Code which won’t benefit Microsoft because it’ll be separate. What’s the use in not just open sourcing the C extension? With more people moving off C anyway, might as well get the free community contributions
MSFT want to build their own Cursor aka Copilot Agent.
They can build a better product with their resources effectively extinguishing Cursor, who will then need to find a way to differentiate.
If Cursor was smart, they would have decoupled from the beginning as they had first mover advantage. They will now have to adapt while fending off competition from MSFT and the other players.
MSFT meanwhile, have discovered that this market is too profitable to be left untouched. They have probably been building their agent for a while and have now decided to launch while simultaneously blocking direct competition. They already have an ecosystem with users who have switching inertia. It's a brilliant yet ruthless move.
Cursor is a small team, MS is a titanic enterprise. I highly doubt that Cursor could exceed MS when their entire product is built on VSCode in the first place and they can't even seem to describe their usage policies to their paying users.
Aren't you curious how 4MB of typescript can parse and understand C++ code? It doesn't. It downloads an additional 200MB binary language server that does all the work.
It is a public website and a public service - it's like saying "hey I got free lemonade here, but you can't have it unless I decide I like you first."
If you're giving something away online for free, then you are giving it away for free. I'll never understand the cognitive dissonance of "conditionally free".
A more important question is where do we draw the line of abuse? If someone links to my website and that's okay with me, but someone else does and I don't like it, do I have the right to conditionally block access to them? And do they have the right to circumvent that to regain access that I freely give to others?
Unless said beer contains numerous nanoprobes that phone home every measurement detail about your insides while they traverse your intestinal tract, the beer is a lot freer than vscode.
> I'll never understand the cognitive dissonance of "conditionally free".
it's not a cognitive dissonance. Lots of places have conditionally free stuff - it's a form of price discrimination (coupons, special deals etc).
Microsoft is within their rights to make their servers conditionally free. What the community can respond with is to move to a different server, if such conditions are not within the bounds of the community's lines.
> I'll never understand the cognitive dissonance of "conditionally free"
I don't think I understand. You don't understand how something can only sometimes be free? Like, free parking only on weekends? Free entry for young children? And free software depending on who you are and what you are going to do with it?
> Because they run a marketplace for Vs code they should also support paywalled forks?
Since, because of the marketplace, MSFT (somewhat) "monopolized" the access to extensions, they should not block other applications (forks) that also attempt to access the marketplace.
Those quotes around monopolized are really doing some heavy lifting considering that it is utterly trivial to use alternative marketplaces on (edit: flavors of) VS Code.
Seems to me this is plainly the community wanting its cake and to eat it too.
Why does Microsoft have the right to cause users unrestricted bandwidth use through updates and ads and spying? That's a real cost Microsoft is forcing onto users.
If bandwidth is so precious, why isn't Microsoft paying users for the bandwidth they use pushing ads to their PC? Why isn't it considered onerous for them to foist tens of gigabytes in updates every week? This is a direct cost to consumers that Microsoft is pushing on them. Do you think that's fair? Or do you want to admit that your entire premise and argument is nonsense corporate apologism?
My experience is that for anyone sufficiently famous and polarizing, there are widespread false allegations. It's hard work to work from primary sources and sort fact from fiction.
It's impractical to check everything, do I tend to do deep dives spot checking a small number of things.
For readers, I'd suggest the same thing here. Disregard claims on the Internet, or even court rulings, and just look at primary evidence. Pick a small number of issues.
I make this statement generically, without prejudice to the outcome here.
I'm not sure what you mean. I generally agree with you — but I think in the case of Trump you have to disregard at least 26 [1] public allegations of rape if you want to give him a pass, blame his fame, or partisanship, or whatever.
Right.. the credulousness of these people is insane. “I can’t believe the guy who said he liked to sneak backstage at the Miss Teen USA pageant and bragged about grabbing women by the pussy would assault someone!”
It's worth noting that Stormy Daniels' description of her encounter with Trump also amounts to rape. I don't think she ever used the word, but it's clearly what she describes.
Ms Daniels said she "blacked out" despite consuming no drugs or alcohol after Mr Trump prevented her from leaving the room by blocking the door. She said she woke up on the bed with her clothes off.
"I was staring at the ceiling and didn't know how I got there, I was trying to think about anything other than what was happening there," Ms Daniels testified.
Ms Daniels, whose real name is Stephanie Clifford, said she did not tell Mr Trump to stop. "I didn't say anything at all," she said and that she left the hotel room quickly afterwards.
>you have to disregard at least 26 [1] public allegations of rape if you want to give him a pass,
Allegations mean little, and for celebrities they tend to pile up proportionate to their fame. We live in a society that has absolutely no disincentives for false allegations of rape, and that has only grown more true the last few decades.
Instead of disregarding 26 allegations, one has to wonder why anyone would regard them in the first place. Furthermore, for many people, their regard/disregard is highly selective and comes down to the politics of the accused.
There is no incentive to make up allegations against most of those people. But if you make up a false allegation against a presidential candidate, it could cost him the election and move national politics in the direction you favor. How many allegations did Trump have against him before vs. after running for president?
There is no incentive to make up allegations, period. Lying about sexual assault in court is perjury and jeopardizes victims as much as the defendant.
The simpler correlation is that most of the people on that list respect the law and do not consider themselves beyond reproach. Mind you, Tate was fleeing Interpol on human trafficking charges when he was arrested. These men know what they did wrong which is why they lash out when accused instead of respecting due process.
>There is no incentive to make up allegations, period.
That's obviously not true. For example, this woman confessed to making up an sexual assault allegation for political purposes:
>One of Justice Brett M. Kavanaugh’s accusers admitted this week that she made up her lurid tale of a backseat car rape, saying it “was a tactic” to try to derail the judge’s confirmation to the Supreme Court.
>Lying about sexual assault in court is perjury and jeopardizes victims as much as the defendant.
So what? If I make up an allegation against you, there is little risk to me unless you can PROVE I lied. But if the "evidence" against you is just my word, what can you do with that to establish that I am lying?
I think your argument is spot on, but there is important context which can be revealed by doing the same list for assassination attempts. Trump is qualitatively different from these other people - it just isn't because he is famous and polarising.
And Vladamir Putin (0), seriously? Good luck to anyone who attempts to make a public accusation against him. There will be a fatal fall through a window in their future. He could have raped 200 women and nobody would say a thing.
There is widespread fraud in the government. It needs to be addressed. There is widespread inefficiency too.
I think the people in DOGE have the skills and access to address it.
I have no evidence that they are doing so, and some evidence of widespread loyalty tests which, while not identical, remind me of how Stalin came to power.
However, absence if evidence is not evidence of absence, and some evidence is not the same as proof.
I have dozens of explanations which fit the facts, and I don't have any way to determine which, if any, is correct.
> There is widespread fraud in the government.... There is widespread inefficiency too... I think the people in DOGE have the skills and access to address it.
Given that just getting the names of the people involved in this process incurred Musk's wrath and accusations of criminal behaviour... how can you have any justified belief in people having 'skills' to address 'fraud' and 'inefficiency'?
We'd need some common definition of 'fraud' in the first place. Many of the things that have been labelled 'corruption' seem to just be 'things Musk doesn't like'; I suspect 'fraud' would be similar.
"Inefficiencies" - we have the Chesterton's Fence idea to illustrate that what might be 'inefficient' is intentional with an overall positive purpose. Again, define 'inefficiency'. The rate at which firings have been happening may certainly be 'efficient' from an operational standpoint, but having to scramble to rehire key people who shouldn't have been fired in the first place is 'inefficient' at best.
> I have dozens of explanations which fit the facts, and I don't have any way to determine which, if any, is correct.
I'm not sure we have enough verifiable 'facts' that can support many conclusions at all, and I think that 'fact' itself is evidence of intentionality in keeping the public in the dark about what's going on and why.
LLMs tend to be very naive in their outputs when you start asking for anything below surface level. If you ask it how to audit something, it'll probably give you a solid high level answer - look at a, b, c and try to build a narrative about how they relate and then look for deviance (I'm not an auditor and I didn't use an LLM for this). Once you start trying to look at the mechanics of how to actually do that, that's when it will start "hallucinating" or just generally swirl. It's the side effect of having a ton of training data on what something is but not much data on how to do it in practice.
This may change at some point in the future, but I would hardly say that using an LLM is "close to having someone with that experience and knowledge," or maybe it is "close" but it isn't a substitute for "having" when dealing with serious topics.
I've found that when cross checked against my own expertise, LLMs have dubious "knowledge" at best. Trusting the output with anything you already don't know would just be Gell-Mann amnesia.
There's a lot of just plain simple fraud too. I've seen embassies issue visas only with bribes, or employees simply collect salaries without doing their jobs. As in you're hired to review documents by some legally mandated criteria, and they simply toss them into piles without even glancing at them and go home early.
This is sort of the group that interests me the most, there is some notion of the government as sort of an employer of last resort in some areas which is a progressive/liberal idea, though id imagine with the areas most impacted by offshoring these jobs are disproportionately in red / Trump supporting states.
And even if you’re ok with getting rid of these jobs, the biggest impact might not even be the loss of these jobs but the loss of the consumers who had these jobs spending money in their local communities.
In addition to the money that's spent, these people had great health insurance. This helped to subsidize rural hospitals. Between this and cuts to medicaid, more rural hospitals will close.
A basic problem is they're trained on the Internet, and take on all the biases. Ask any of them so purposed edX to MIT or wrote the platform. You'll get back official PR. Look at a primary source (e.g. public git history or private email records) and you'll get a factual story.
The tendency to reaffirm popular beliefs would make current LLMs almost useless for actual historical work, which often involves sifting fact from fiction.
Couldn’t LLMs cite primary sources much the same way as a textbook or Wikipedia? Which is how you circumvent the biases in textbooks and wikipedia summaries?
A raw LLM is a bad tool for citations, because you can't guarantee that their model weights will contain accurate enough information to be citable.
Instead, you should find the primary sources through other means and then paste them into the LLMs to help translate/evaluate/etc, which is what this author is doing.
Yes, I could, but I'd have a 90% chance of hurting myself. Ditto for giving some cancer with radiation, many covert poisons, and similar. I can name a dozen ways I could kill or harm you, but:
1. I'd need to be wacko enough to want to harm you
2. I'd need to be sane enough for a major engineering project
3. I'd need to be competent enough to succeed without hurting myself
The set of individuals with all the traits required is small. It's not zero -- we had Ted Kazinsky -- but very small. Most major crimes are also never solved, and many more people can just take a gun and shoot someone. That's probably even harder to track down since a lot more people have guns than labs and PhDs. Esoteric technologies make for easy investigations.
So in the end, I think state actors have these. They also make good plot lines for a novel or video game (I have about a dozen I came up with in the context of a half-baked creative project).
But for real life individuals, I don't see how they'd become prolific. We'll probably see a crazy someday, but nowhere close to gun homicides. They just don't make much practical sense for an individual.
World War II was won because of US and (less advertised) Soviet manufacturing capacity.
Ukraine is more constrained by weapon supplies, especially drones and artillery, than by manpower.
I'm not sure both sides are playing the same game. In the game I think might be afoot, if your economy is management consultants, therapists, and hair stylists, that's "bad" GDP. Resources and manufacturing are "good" GDP. Resources are complex in that stone are renewable, and some go away when you export them.
I'm not even going to talk about education or, related, R&D capacity.
high finance/management can be quite valuable when you’re trying to plan a conflict. this is how you get the industries developed that are capable of designing a system to turn sand into missile-targeting systems.
i’m extremely skeptical of people’s ability to view the economy from on high and pick out the “good, productive gdp” and the “bad, makework gdp”.
That's not quite my point. "Productive" is for a reason. Is manufacturing perfume productive? Making video games?
It depends on your goals.
However, in a great power conflict, they're much less productive than tanks, drones, fighter jets, and guns.
Conversely, money thrown into weapons can't be used for quality-of-life improvements.
Neither of these is "good" or "bad" per se.
Some industries are hybrid. A ship yard is dual purpose.
If two hypothetical countries are playing this game, and one exports movies to buy ships and vice versa, when war breaks out, one country might be stuck without movies and the other without ships.
There's a huge step to I'm capability with 16gb and 24gb, for not to much more. The 4060 has a 16gb version, for example. On the the cheap end, the Intel Arc does too.
Next major step up is 48GB and then hundreds of GB. But a lot of ML models target 16-24gb since that's in the grad student price range.
At the 48GB level, L40S are great cards and very cost effective. If you aren’t aiming for constant uptime on several >70B models at once, they’re for sure the way to go!
Not arguing against 'great', but cost efficiency is questionable. for 10% you can get two used 3090. The good thing about LLMs is they are sequential and should be easily parallelized. Model can be split in several sub-models, by the number of GPUs. Then 2,3,4.. GPUs should improve performance proportionally on big batches, and make it possible to run bigger model on low end hardware.
Yeah, I’m specifically responding to the parent’s comment about the 48GB tier. When you’re looking in that range, it’s usually because you want to pack in as much vram as possible into your rack space, so consumer level cards are off the table. I definitely agree multiple 3090 is the way to go if you aren’t trying to host models for smaller scale enterprise use, which is where 48GB cards shine.
My guess is that most people on a jury would convict, but the odds of an entire jury convicting are less than fifty percent. I don't know how much less.
Tha kind of flexibility is the point of a jury. Laws aren't purely mechanical in most countries, for good reason. There's a lot of complexity and nuance, especially when some people have political power.
I don't have enough background here to make a judgement about what a jury should do, but I do have enough background to know that things like jury nullification are important to a well-functioning justice system.
