Yes, and No. At some point, even scaling at all would be hard.

(Just like sending a human to Alpha Centauri is hard, even if you had unlimited funds.)

Eh, sending a human to Alpha Centauri wouldn't be that hard... Although it would be difficult to know for sure if they arrived, and for ease of transport, you may want to send a dead human.

You are joking, but even sending 10kg to Alpha Centauri to arrive within a century (or at all) is something that has never been done.

Arriving within a century is a significant new requirement! But yeah, we haven't done it at all. We've gotten 5 objects to escape velocity to leave the solar system though, so it's not a huge leap to chuck it at Alpha Centauri instead of wherever we did... Of course those things aren't going very fast, the internet says Voyager 1 would take 74,000 years or so to make it to Alpha Centauri at its current speed if that's where it were heading. That's a really long time, so it's not super worthwhile to do. You might be able to design a system to launch a 10kg object from an interstellar probe after it reaches heliopause to get a higher velocity, but it's still going to take approximately forever to get there... And I don't know that we can track our interstellar probes if they're not sending radio signals anyway? So chuck it at Alpha Centauri and make a website that counts down to the day it gets there, cause nobody is going to know if it actually made it there and you'll be dead anyway.

Like it how? Accomplishing a grand feat is nearly the opposite of scaling.

If Netflix built out more slower servers, that would be acceptable scaling. I don't see any plausible scenario where that becomes too difficult. Even if they had billions of subscribers.

There's lots of scaling involved in grand feats.

Eg enriching a tiny bit of uranium is 'relatively' easy. Enriching enough of the stuff to build an atomic bomb is a question of scaling. And, yes, in principle you can just use more and more centrifuges etc.

Humanity has already built some small infrastructure off-earth, like the ISS or satellite networks. Also various unmanned probes that toiled away for years in space and on Mars.

Building a habitat big enough to plausibly keep humans alive for the decades or more it would take to get to the Alpha Centauri would involve enormous scaling efforts. And that's not even touching on propulsion, yet.

However, yes, grand feats require more than just scaling.

> Eg enriching a tiny bit of uranium is 'relatively' easy. Enriching enough of the stuff to build an atomic bomb is a question of scaling. And, yes, in principle you can just use more and more centrifuges etc.

But we're not scaling for scaling's sake. We're scaling with the number of customers.

If it was profitable to enrich some milligrams by hand, then you'd be able to get more scientists and engineers and better equipment and automate some parts and make even more profit. If you go the centrifuge route, it's not because you had to do the harder methods, it's because it's massively cheaper per gram of product.

Such a scenario is very different from "we can't do it at all, and we need to build complex infrastructure just to make it possible".

It depends entirely on the problem domain. Sure, it is more of a devops problem when the problem is trivially parallelizable, but often you have a bottleneck service (e.g. the database) that has to run on a single machine. No matter how many instance serves the frontend * if every call will have to pass through that single machine.

* after a certain scale

Tell that e.g. Tesla

What I've read they burned a lot of money and hat large problems scaling nevertheless. Which I don't find too surprising, not because they are unable, but because it isn't easy to scale.

From my experience and from what I read scaling people roughly a power of ten is a larger change in an organisation and therefor likely a challenge. For _any_ technical process the boundaries might not be strictly a power of ten but i would say that scaling a power of a hundred is a challenge if this value is not already reached on any process in your organisation.

True.

Scaling to - say - Paramount+ size should not be difficult if you're willing to pay AWS / Azure / GCP 10-100x what it would cost to serve it yourself (which in many cases actually makes sense).

It's possible at Netflix's size, they couldn't just run on AWS anymore. Though, given enough lead time and a realistic growth curve - I'm sure it's feasible.

Obviously scaling manufacturing is not a solved problem like (realistically) scaling network and compute usage.

Serving Netflix streaming traffic from AWS would be… unwise. One the bandwidth cost would be enormous even if they can handle it. And two I doubt they can handle that much traffic.