Thanks for building what I'd hoped to find the time to build (and much better than what I would have made)! One question: do you think there is room for parallelization here, eg in the retry loop? Local models generally can handle a limited number (~ 2 digits) of concurrent requests pretty well, even on consumer hardware, which can give >10x boosts in the effective number of token/s. I've been thinking for a while about workflows that could take advantage of this, and 'fix this error' could be one (if not ideal) application. Would be curious what you think.
Linux itself, major Linux distros, npm - none of these were designed with a security-first approach. Even the things that do help with security, like package maintenance or containerization, were more incidental to other primary goals like stability, reproducibility and so on rather than being born from a comprehensive security-first strategy. They could have been, but then things would have moved slower. They even exist, like Alpine, OpenBSD, RedoxOS, but the major ones, the ones we're talking about today, were the ones who moved faster and managed to take over. That's the fundamental issue I'm talking about, the mindset shift that would be required before we could even start the Herculean effort of rebuilding much of the existing stack with different architectures, in different languages and using different development models, always knowing that, in the past, the ones who moved fast and broke things instead tended to be the ones who succeeded.
After the npm supply chain attacks people suggested automating delays before installing updates, now we're talking about automating update delivery... I'm afraid there won't be any easy or quick fix after decades of treating security as an afterthought.
I felt a bit similar about electric cars with a trunk in the front where the engine would sit in an ICE car. But that's more about esthetic expectations, like the first cars looked similar to horse carriages.
The vast majority of the endowment isn't money (dollars in bank accounts). University endowments work like private equity funds, most of the funds will be invested in assets, most of which hardly liquid enough to reasonably convert them into cash on short notice. They could try to borrow money against the valuations of those assets, but it's not sane to take on debt in order to sustain a level of expenditures that was adjusted to a much higher level of income (true more generally). Especially when the alternative of temporarily scaling back expenses is relatively easy.
Yeah, the DGX Spark could qualify as a mini PC too. The AMD chip is sold as a laptop chip I believe, but I've mostly seen it in mini PCs. And the Framework Desktop. A brand that probably carries a lot of trust among the kind of tinkerers who would consider buying a barebone motherboard in the first place.
You cannot "use the government". I'm not aware of any countries where it would work like that. You can use the judiciary, which is independent of the government, and you can use it to go against the government the same way you can go against corporations.
It's a condensed statement. There was a time when I would start a new programming project thinking about class hierarchies, maybe drawing some UML diagrams. I don't do that anymore, and I don't believe it's very common for greenfield projects anymore. But educate me if that's wrong. We've kept some of the good ideas from OOP like namespaces and interfaces and we use them in slightly different contexts now, where OOP may even still be technically possible, but it's not the primary way of doing things anymore. I believe, or at least hope, that we will see a similar kind of evolution for package managers. Where it's still possible to use other people's code, but having packages like left-pad or is-even is no longer how it's commonly being used, even if it may still technically be possible.
Visa/Mastercard are also 100% surveilled. The nightmare of your transactions being monitored and arbitrarily blocked that's peddled as an argument against CBDCs or public financial infrastructure like Pix is already a reality for all of traditional finance (with the exception of physical cash, which happens to also be a central bank product). Sometimes even balances get confiscated, read the horror stories by Stripe customers. Americans seem to believe that the same thing is necessarily worse if it's done in the public sector. Other countries tend to see it differently. Just let them be.
I don't have an answer what the alternative is going to look like. But smarter people than me may find something. C/C++ are doing fine without package managers. Go at least has a more capable standard library than Rust. But I'm not sure if Go's import github approach is the answer.
One idea I've been entertaining is to not allow transitive imports in packages. It would probably lead to far fewer and more capable packages, and a bigger standard library. Much harder to imagine a left-pad incident in such an ecosystem.
My pet theory is that package managers will one day be seen like we see object-oriented programming today. As something that was once popular but that we've since grown out of. It's also a design flaw that I see in cargo/Rust. Having to import 3rd party packages with who-knows-what dependencies to do pretty much anything, from using async to parsing JSON, it's supply chain vulnerability baked into the language philosophy. npm is no better, but I'm mentioning Rust specifically because it's an otherwise security-conscious language.
Well, the two chips I mentioned (DGX Spark uses the GB-10) are both a SoC, so no motherboard needed there. I don't know if that's the full explanation, but it could be a factor.
The SoC design with unified memory is generally well suited for residential use because it's quite energy-efficient, quiet and small (compared to traditional GPU-powered gaming rigs). Great performance-per-annoyance, so to say.
Contrary take: I believe we will see an expanded market for capable PCs that can be sanely put in a living space. By extension of the gaming PC niche to local AI. Both NVidia and AMD are developing product lines in that direction (DGX Spark, Ryzen AI Max). And Linux will be more prominent than ever, due to several independent reasons: MS dropping the ball hard on Windows, SteamOS making Linux attractive for gamers, 'digital sovereignty' as a trend, and Linux being the de facto standard for hosting AI (or anything really).
I don't like it. It's inevitable, but no reason to cheer it on. I find it similar to Google Mail or YouTube autotranslating content without opt-in (and sometimes opt-out). It's continuing a trend of you can't really trust the content you see is the content someone else sees or what they sent. It says it only changes accents, soon it'll filter swear words and what else? The end game for the legal use of such tech is always injecting ads. And with this particular tech, we know that the legal uses will be a negligible fraction of the real uses.
Lean isn't the most loved proof assistant by mathematicians, it's not the most suited to formal verification of software, it just sort of works for both. Yet it's got the thing that's arguably the hardest to achieve, momentum. Compounded by the fact that in the AI age, the amount of publicly available human-written code directly affects how well agents can produce new code.
They do also make a lot of money selling hardware, and as things stand today that business happens to make them look like the first tech giant to actually profit from the AI boom (because the hardware they've been developing internally for years happens to be among the best consumer-grade options for self-hosting LLMs). Making their hardware more attractive to tinkerers could be a winning move right now.
Hexagonal, with shaded colors? QR Codes are, by definition, square and binary and traditionally use black and white. They're also used for a different purpose typically. They could easily have made them look more like QR Codes if they had wanted to, but they made their own artistic choices. Which I love btw, but they could have maybe chosen better wording. Something like 'fingerprint' or 'mugshot' would have conveyed the idea of it being useful for identification, if not perfect, much better.
As one lecturer put it: modern probability theory derived from two foundations - measure theory and gambling. The latter explains why it has long lacked mainstream mathematical recognition :)
But that's all in the past. Probability is absolutely established in math academia today, Fields medals and all. And despite its applied nature it's pervasive even in pure math.
Genuinely curious, what does that mean? Is the PDF standard now open and free, or does Adobe wield any power? Does it own trademarks (for pdf itself, not Acrobat Reader or sth like that)?
Maybe someone with more time at hands could look up what Google said with respect to ads and what happened later.
This is one of the rare instances where it's very easy to predict the future: the prompt auction market will look similar to the existing online ad market, financial firms will pay for prompt streams for sentiment analysis, companies and interest groups will pay to have their products or agenda included favorably in the training data for future open weights models... any way you can think of that LLMs can be monetized, you will see it happen. And fast. The financial pressure is way too high for there to be too long of a honeymoon phase like we had with web 2.0
Is an agent-ready website so obvious poison? If I'm running a plumber shop in East London, then I'd want agents to know that just as much as I want Google (Search) to know that. The same will be true for most real-world businesses. Only sites that make money by selling their users' data and eyeballs obviously stand to suffer.
So the difference between a theoretically pure prediction market and Kalshi is ... this interview? The CEO saying that he thinks others, who do not answer to him in any way, will be doing something to enforce some notion of fairness.
If you're being that puritan about the definition, then having a "real" prediction market is completely impossible. Because actors like the DOJ do not wait for a statement by the Kalshi CEO to bring charges. And rational actors will know and anticipate that, and hence preemptively comply. So you never get the unfettered version of a prediction market.
I don't think it makes sense to be that puritan about a definition that the thing it's trying to define becomes an impossibility. Polymarket, Kalshi et al are clearly prediction markets in the messy reality that we live in, and we're figuring out as we go what the legal reality of a real-world prediction market is and should be.
Why would it be insane? We have globally accepted answers for the area of each country (modulo territorial disputes, geological changes and similar). One would expect the same thing to be possible for the circumference. So to most people it will be a surprise that it is, in fact, impossible. It is mathematically impossible because the problem is underdetermined, and it is practically impossible because agreeing internationally on how to fully determine the mathematical problem seems unrealistic.
Where do you see a difference? Like you said, there is a libertarian argument that can be made for why insider trading is desirable. If the bet is easily manipulable, like how many times someone will visit a place, then the rational response is for others not to bet on that market. The same argument still holds.
You can disagree with the libertarian argument, but I don't see how you can say that Polymarket et al. are something other than a prediction market. Can you explain where you see the difference?