Did you read beyond the first paragraph? This is about 0.1234567... where the n-th digit is written in base n. I found the visual proof more interesting since it is quite intuitive. And also unlike 0.(9), for which you have to accept some unconventional arithmetic facts as part of the reasoning.
After GRH, I guess all the millennium problems are fair game now. It's only a question of who's willing to burn enough compute on a fundamental math question that will most likely not meaningfully alter your quarterly revenue report.
I think GRH will be a distinct inflection point. There's a genuine sense in which math is effectively solved. We now know that it's almost certainly possible to push as far as we want, within reason, and the limiting factor is just compute.
Interesting idea, regarding math being solved. Mathematics is notoriously practical as a tool for imparting tools to science and engineering. Often in surprising ways! With this in mind, do you think, now that mathematics is solved, we will be stunted in learning new tools for physics, chemistry etc. Perhaps those are solved also? Or perhaps we will never be able to make further progress on them, now that we have reached mathematics' edge.
Now that we have math solved and IPO cashed out etc., can we start instead allocating the capital into advancing robotics and human health or something like that that's nice and exists in the real world and not only in the tubes of the world wide web?
If stuff like cure for cancer, alzheimer...etc. is too hard.
There's a long list of other "mundane" stuff that for a long time we've been promised AI/robots wound take on, so that humans could spend more time on arts and other more pleasurable/interesting things.
Let me know when AI proposes new axioms independent of our current theories of arithmetic that are true in the standard model. Until then, I wouldn’t call math “solved”.
Apartments should be free, as long as you’re willing to move within 50km of your place of work, or anywhere if unemployed.
If you want a big house, boo hoo pay for it, but somewhere along the line we were convinced the government owns all the land and we have to pay for it. In reality, nobody owns the Earth, it belongs to every human, so at a basic level you should not have to pay for a plot to live. That would be paying for your human rights.
The problem is you will also find people like this in homeopathy. But cargo cult science is still not real science, no matter how much work these people put in their job. In the absence of widely agreed upon education and examination, all this effort is worthless once you leave these bubbles of people who believe you without evidence.
It is sufficient to find 1 wine expert who can identify wines with a reasonable correctness rate in a blind test to claim that it is possible to do so. The fact that many claim to be experts but are not doesn’t change that.
That's not how this works. I'm pretty sure you can find a magic fortune teller who got a series of predictions right in a row. But if it's just one guy and there is no method or replicable science behind it, it is almost guaranteed to be just chance.
What would you say is the difference to working with, say, Github and a docker sandbox for the agent? Or simply Cursor, which lets you run the agent in a fully customised cloud vm sandbox with env credentials being masked from the llm and PRs by default?
It works from the command line. You can spin up a sandbox around your workdir, using any backend (docker, VM, kata, firecracker, seatbelt, apple container, etc) and it just works.
You could roll your own, just like you could build your own docker implementation. But the devil's in the details. The technologies it depends on are sharp and pointy ( https://github.com/kstenerud/yoloai/blob/main/docs/contribut... ), and it's difficult to make them do the right thing.
I have rolled my own setups for a while now. The devil is indeed in the details, which is why I'm very fond of Cursor's approach. Probably the best overall if you care a lot about security and not a lot about privacy.
Well, digital centralisation and interdependency without fault hardening is a big problem. Just because it presents itself in slightly different ways across industries doesn't mean it is not a fundamental issue in our society.
I don't see much of a problem with depending on AWS or Google Cloud. It's Microsoft who are particularly flaky. And always will be, because IT people choose them regardless.
You are still "centralizing" on "someone else's computer". Just because those two have not, yet, had a MS365 event does not mean they won't have one at some point. And everyone "centralized" there will have a bad day when that happens.
AWS has had them, and is passed the growing pains. At this point when/if AWS goes down, it will still have a far higher uptime than others could manage, let alone those going it alone.
Microsoft on the other hand, has proven that it will never be reliable.
AWS quite literally lost an entire region in the Middle East and has said they will not be able to restore it.
You can argue that this is a result of something outside of their control, and you would not be wrong, but if the point here is about centralization rather than simply a particular company's operational procedures, then direct strikes on datacenters should be part of that conversation.
I was an Amazon red badge in the EC2 org, fwiw, and do think AWS has excellent operational culture, far beyond what I saw while working in the M365 org, and beyond what I've seen at other tech companies and hedge funds where I've worked. But that doesn't obviate issues with centralization.
Centralization is almost a good thing for extreme cases though.
If Bank of America's central datacenter (cloud or on-prem) gets hit by a cluster bomb, taking out banking for ~half the US, you can be damn sure that very smart people are going to be working 24/7 until it's resolved.
If you have all your savings stored with a local credit union, say Seattle Credit Union which has 50,000 customers, and they get hit with a ransomware attack... well, good luck. You (probably) won't lose any actual money thanks to FDIC rules, but it might be weeks or months before their services are back online.
Similar to self-hosting vs cloud, you get a bit of a paradox where it's overall better for society if there's no interdependencies so a SPoF outage doesn't impact everyone, but things are much worse for an individual in the event a localized outage does occur due to economies of scale, or lack thereof in this case.
It’s a tragedy that much of European industry and public sector are so deeply embedded with Microsoft that they barely even consider other options, even when everything Microsoft touches seems terrible
This is also true in Oceania (especially NZ) in my experience. Companies here are much more embedded with Microsoft than companies I worked with back in the US and are less willing to consider other options. My company actually does most stuff with on prem windows servers and a bit with Azure. One aspect for NZ in particular is that there was no NZ region for Azure or AWS until 2025 and parts of the country are 70+ms from Sydney, so for anything where latency is significant, on prem stuck around here a lot later.
I feel like on-prem Microsoft stack is a very well known and controllable beast.
Its pros/cons are known and gaps can be addressed with 3p software. You just keep on old outdated, but proven version of software and keep the perimeter locked tightly.
GP isn't talking about on-prem I believe. Small/medium businesses in Europe are much more dependent on MS than their US counterparts. Alternatives are mostly out-of-scope from their point of view.
Don't forget that MS has the advantage of non-English languages support that many alternatives lack.
> I feel like on-prem Microsoft stack is a very well known and controllable beast.
I feel like it is very unknown, there is no documentation and everyone relies on comments on random forums that are now 21 years old which tells to set undocumented registry keys to undocumented values.
They’re expensive and centralized. I always thought the general wisdom was to bootstrap your business with serverless infrastructure and then eventually work your way back to an on-prem-first philosophy. I guess a lot of companies skipped that last step to…save…something?
Vulkan is super appealing if you're in the industry and have the time and resources necessary to profit from its advantages. But if you're a single dev who wants to learn game engine design, you're going to have a bad time. Most people also don't get that game engine design is very far removed from actual game design. You can have a ton of fun learning math, physics and computer science when building an engine, but beware that you'll likely be mentally and physically exhausted long before you actually get to build a fun game.
>if you're in the industry and have the time and resources necessary to profit from its advantages.
I don't even know how you get into the graphics industry these days. The bar is so high and I just don't see how you get the knowledge needed for it. I graduated years ago and don't feel any closer now than back in the mid 2010's despite having a lot more experience to point to in other parts of games.
vulkan doesn't have global state, and the error handling is better
but it's not batteries-included, and that's often to be a deciding factor at small scales
i think if you're going to dabble in engine dev, you pick which one you want depending on which part of the engine you find interesting. if you want to make a game, you pick up unity or godot or something
DirectX 12 only gets interesting when you want hardware raytracing support to make use of the new Nvidia cards on windows. Tbf that is pretty cool, which is why I actually dabbled with it a little. But it's not necessary for the vast, vast majority of graphics applications.
People have claimed that competitors are catching up for a long time. But these journalists never understood why Nvidia was so dominant. And it seems they still don't. The Jensen story only made investors unhappy because of the high volatility and uncertainty. Noone who's well informed dumped Nvidia because of technological fears of competitors.
OTOH developers keep confusing R&D vs production. Nvidia is going to keep its dominance in R&D. But the vast majority of Nvidia chips are used for training and inference. Using Nvidia for that saves you 3 days of work porting from R&D to production.
It’s over if they don’t put a technocrat in charge a bean counter isn’t gonna keep them on top. See Apple when Jobs was fired or the clown lineup of bean counters at Intel Intel…
Maybe they wouldn't be the laughing-stock of AI if someone better than Tim Cook was leading them. Siri is pretty bad, "Apple Intelligence" is a misnomer.
Apple is currently the most valuable company in the world and unlike many of the companies that aren't "laughing-stock of AI" it actually has a defensible moat for its products for the foreseeable future.
I'm generally not an Apple customer at all (I own an old Mac Mini to do occasional iOS development, but rarely use it), but think they are a better bet long term than any of the AI labs.
Tim Cook did not make Apple the most valuable company, they are still riding the coattails of the reality distortion field eminating from Steve Jobs grave.
To be fair, ram really is cyclical, and a commodity.
The Nvidia ecosystem… not so much. Most all of these pieces fail to recognise their already dominant position in the edge inference space - and most all of these pieces fall into the trap of AI=LLMs=ChatBots, and really, who wants to chat with a computer any more.
I don’t mind. This has been a great week to stock up on the equity, and there was enough blood in the water before the salacious spy sex story that I bought a bunch of very, very cheap $150 puts, as hysterical overreaction is the plat du jour, which I sold on Thursday.
Nvidia biggest problem going forward they aren’t as versatile now essentially a one trick pony and now that they don’t have their CEO, they will be rudderless. That latest team up with Microsoft was a dud and a bean counter CEO isn’t gonna be able to fix that.
Having done all three over my career (local machines as student, University's HPC cluster as grad/postgrad and eventually cloud in the private sector), I'd 100% rent out compute. Buying a custom rig for local ML training makes zero sense in 2026. With renting you'll be able to train bigger models/datasets faster on much better hardware while paying far less in total. Especially if you're doing project-based work.
If you use a custom disk image too, you can have it boot, apply security updates, reboot, then ssh in within minutes.
Work for however long, shut down. I do this with linode and their GPU VMs. If you shutdown nightly and on weekends, the costs drop dramatically on a monthly spend.
I only need it a few hours per day, so very cheap for custom compute.
The usage pattern i've seen on flexible compute is more aligned with 'I need this process to execute, run it when it's cheap' not 'I'm going to kick off this process, stop it if it get's expensive.' for example AWS Spot Instances
Friends of mine wanted to buy a GPU server last year for their new project. Management told them no, that there was a deal with cloud providers and that investing the budget in cloud compute was better.
The project ran out of budget roughly three months in. From what I heard, status meetings consist of everyone getting together once a month, asking "have we gotten any more budget yet? no?", and then wait until the project runs out.
Cloud providers need to make a profit too, and they've gotten very good at ensuring that their clients forget this point. The whole advantage of cloud compute is that you don't pay for the time you're not using it, but once you reach a certain point you're the one covering the overhead for everybody else.
reply