Hacker News Reader

Comment by ayaros | Hacker News Reader

This is great, but I'd rather they make Windows 11 open-source instead.

Funnily enough you can easily find the Windows XP source code on GitHub. Not endorsed by Microsoft of course, but they've ignored it sitting on their own service for years, along with ignoring all the modern Windows and Office piracy tools which are also on GitHub. Microsoft works in mysterious ways.

Most of the money to be made is by licensing software to organisations that can afford the risk of pirating (practically anything bigger than SMBs: enterprises, governments, armies, etc). The moat of everyone used to your platform worths a lot more. So they just regulate enough so it won’t seem like they don’t give a shit at all.

If AGI ever comes close to fruition I can't wait to just dump this code into some AI, tell it to fix all security bugs and make it work on M Series processors. Would finally achieve a computing environment that would be perfect for me. Until then, I will continue to dream.

If we ever get to the point of having a tool that could do something that complex, we're well past the point of using human-written operating systems or using M-series processors.

Which is to say, very, very, very far away.

Hmm I seem to recall Karpathy talk about the next era of software being like this: Essentially everything is made on demand exactly the way the user expects or something like that. it would definitely be a paradigm shift that requires rethinking how we approach software.

What do we really need? If you look at what Ai can solve today, smaller problems seem to be solved really well by AI. So just an AI with a large enough context, adequate speed to made the changes in near real time and a reasoning ability that can better handle problems vs whats available today might be adequate to realize this relality in some way.

What we call "AI" right now could never do something of the complexity we're discussing. It has no memory, no experiences, no consciousness, no capability to continually learn, no ability to abstractly understand ideas and apply them.

It's text probabilities in a box. Until something comes around that is NOT an LLM, it doesn't matter how much memory or training data you throw at it, the fundamental idea can't work.

Why not use AI to make ReactOS better? Is there something in original Windows XP that ReactOS doesn’t want to implement?

I suspect this question would be asked different ways depending on the historical context of the user. I grew up using Windows XP, dabbled in using ReactOS maybe once or twice, saw that didn't meet my needs when I tried it, and then proceeded to not really spend anymore time thinking about it.

Why not just try and compile it yourself, see what happens?

One does not simply replicate a Windows build lab at home. (insert Boromir meme)

Not with that attitude, old son.

> along with ignoring all the modern Windows and Office piracy tools which are also on GitHub

You weren't going to buy it anyway. No-one cares about you. Pirate it if you like. Take your warezed copy of Office Home Edition and be blessed, no-one is going to miss your 120 bucks.

An organisation with maybe 100,000 users each paying a per-seat licence? Yeah, that's the sale they want. Not your one-off copy.

Yup, they will care when a huge % of people start doing it.

Not really, no.

No-one buys Windows. No-one buys Office. It's a thing that comes bundled with a computer, or that you "acquire" if you need it.

It's only interesting if Barclays are pirating Windows on a massive scale.

Oh shit did I say the name out loud?

Do companies really not buy licenses?

Reputable ones do.

If so they would have kicked this off github - https://github.com/massgravel/Microsoft-Activation-Scripts

157k stars

They are not this cynical over it, it's part of a plan. What they figure is that they can keep MS the de-facto standard this way. Photoshop worked the same way for very long.

I think it would be safe for Microsoft to release (intentionally or unintentionally) the source for just about any product. I bet it's incredibly difficult to run a successful build. From talking to someone who used to work on Excel, it took them around 1 day to build Excel from source. And that's if everything goes perfectly and you know exactly what you're doing and are using the build system and setup and configuration that the Excel team has in place.

People have built working operating systems from both the XP/Server 2003 and NT 4 leaks. Here's someone building Server 2003 on Windows 11: https://www.youtube.com/watch?v=AWZe00v2Rs0

I guess they are much better than I am. I remember working on a team at Microsoft where the product took 8 hours to build, and on my own dev machine I made several attempts but never got the build to finish successfully. I worked around the problem by taking a known good build from the build master and integrating the part I was working on into that build. But the experience left me thinking "good luck with that" for building more complex products from scratch.

Reply on news.ycombinator.com