Microsoft turns to AWS as GitHub faces AI capacity crunch

Posted by ilreb 1 day ago

15475Original

Comments

Comment by kuschku 1 day ago

What kind of vibecoded website is this?

- the worst infinite scroll I've ever seen making it impossible to access the footer

- the title tag doesn't seem to work properly (just shows the URL in the tab title, on Chrome and Firefox)

- 2007-style keyword stuffing in meta keywords

- the entire page is client-side react with a completely empty body?

The agency that built it even proudly states on their website that they vibecode everything: https://gradientnoise.com/

EDIT: Turns out, the articles are mostly AI-generated as well? https://blog.ryanmerket.com/how-i-built-runtimewire-a-one-pe...

> And this is the gate that lets me sleep: a story only auto-publishes if the editor says PASS, the risk score is comfortably low, a hero image exists, and it has at least one source

Comment by plaguuuuuu 1 day ago

I've read more than a handful of AI-generated articles on HN recently. The last one even had a fake author with a fake face. despite the site insisting it had human editors and reviews. The web property was created by a marketing/AI agency from (AFAIK) a European country with zero employees credentialed to edit hard science news articles, also zero employees with English names and faces like the supposed author.

The interslop is real. Simulacra and simulation.

Comment by ffsm8 1 day ago

The dead Internet theory is long established at this point

Comment by patates 1 day ago

> the worst infinite scroll I've ever seen making it impossible to access the footer

Let's be fair, it's possible to reach the footer! You keep pressing the end key for like a minute or two and 665 requests (with embedded base64 encoded images and all) later, you have your footer.

So saying it's possible to reach the footer would be technically correct - the best kind of correct!

Comment by albert_e 1 day ago

I "love" infinite scroll on websites where important links like finding a way to contact customer support for help is only in the footer that you can never reach.

Comment by Walf 1 day ago

But the hero image is an AI homage to cute paper-craft, to remind us that people cannot or will not put in that level of effort any more, so it's okay!

Comment by ryanmerket 1 day ago

No agency. Just a one man shop. What kind of vibecoded website is this? It's a Replit built site along wtih my 20 years of experience building websites from scratch. It was launched this month, so it's a bit rough around the edges, but we're off to a great start.

Comment by jordemort 19 hours ago

yeah, pointless AI-generated header image --> close tab

Comment by mujib77 1 day ago

works fine on the phone

Comment by kuschku 1 day ago

I can't confirm that.

Mobile Chrome can't show a tab title either https://i.k8r.eu/3vVCTQ.png

And the infinite scroll causes the page to constantly jump back up, again preventing me from accessing the footer

And of course with JS disabled you get nothing at all (which affects RSS readers, kindle devices, etc)

Comment by mujib77 1 day ago

Now i see the problem yeah same on my device to

Comment by ryanmerket 1 day ago

rss is fine /rss

infinite scroll fixed, nojs issues fixed, and title fixed. thanks for helping QA.

Comment by kuschku 1 day ago

Typically, you'd want to be able to fetch the full article with your RSS reader, which obviously doesn't work with client side rendering.

Comment by ilreb 1 day ago

[flagged]

Comment by kuschku 1 day ago

Because I didn't read that thread? Last week I was at a conference, so I spent relatively little time on HN.

That said, I likely wouldn't have read that thread even today. I prefer engaging more deeply with fewer stories.

Another new AI benchmark result is of little interest to me, in contrast to a story about Microsoft stalling or even partially reversing the GitHub Azure migration.

Comment by amazingman 1 day ago

It's a bot

Comment by palmotea 1 day ago

> https://news.ycombinator.com/item?id=48440448 this had >300 upvotes with a link from runtimewire.com and you didn't even criticize it then. why now?

There are so many bad and false assumptions baked into your short comment, it's hard to begin. For instance: in what world does someone have provide justification for their critique of an AI-written website, because they didn't happen to respond to some previous post?

Comment by not_kurt_godel 1 day ago

main character syndrome

Comment by ryanmerket 1 day ago

almost every news website is written primarily by AI these days with a human editor to review...

Comment by kuschku 1 day ago

That's not even close to the truth, there's a lot more manual work that still goes into news than you'd think.

And even if it were true, how would that justify making the situation worse? We should be doing something to make the world a better place instead.

Comment by palmotea 1 day ago

> almost every news website is written primarily by AI these days with a human editor to review...

Source? Or is that just self-justifying speculation?

Comment by MallocVoidstar 1 day ago

Do you think everyone on HN sees every submission?

Comment by jf 1 day ago

I helped set up the first meeting between a Microsoft executive and Thomas Preston-Werner.

One of the moments that stood out to me was when Robert Youngjohns (the exec) asked Tom what it would take to have GitHub move to Azure. I was surprised that Tom had a response ready, saying that IOPS were really important and that virtual disks weren’t fast enough.

Comment by winrid 1 day ago

I had a feeling this was all related to someone trying to run a database on networked disks.

every fucking time

Comment by jiggawatts 1 day ago

Wait, are you trying to tell me that 1/100th the speed at 50x the price isn’t a great offer for Microsoft’s customers!?

Heathen lies!

Oh, wait… we have to use it? Oh, that’s terrible…

Comment by inopinatus 1 day ago

If there’s one thing that surprised me at AWS during my time there - over a decade ago now - that I was not clearly expecting in advance, it was the scale and competence of the units fulfilling the colossal and unceasing growth in capacity demand.

This led me to reconsider Amazon as a whole, and I still think of it basically as a logistics firm, with the shop and the public cloud merely a monetisation thereof.

Comment by TurdF3rguson 1 day ago

I wonder how that translates to Blue Origin and how they managed to be so completely outclassed by SpaceX.

Comment by anonymousiam 1 day ago

SpaceX had been doing their thing for a while before Blue Origin got started. It seems to me that Bezos never wanted to try to catch up with Musk. His approach is not "move fast and break things", but instead more old-school (try to get things done right as best you can, and only test when necessary). Despite the gap, Bezos is slowly gaining ground. He's flying passengers and he's winning government contracts. The USG likes to have a second source for everything, and they're fed up with Northrop, Boeing, Lockheed, ULA (who have all absorbed some of the other space and launch startups). So Bezos is capitalizing on the US government's desire to have real competition, and a second source for launch (besides ULA).

There's also this funny outcome from the SpaceX IPO: https://x.com/ICannot_Enough/status/2065449141946253390

Comment by hvb2 1 day ago

> Despite the gap, Bezos is slowly gaining ground. He's flying passengers and he's winning government contracts.

He's flying passengers? They stopped new shepherd and that was suborbital to begin with. You can't compare the 2 at all. Getting people to orbit is much much harder than 'a hop'.

On the government contracts, yes they did get some. Some through lawyers though and they still have to show that they can actually deliver. SpaceX has to deliver on HLS as well, but the ISS has one American ride up and that's crew dragon.

Comment by 1 day ago

Comment by csbrooks 1 day ago

> commits were on pace to hit 14 billion in 2026, up from 1 billion in 2025

So AI means 14x the checkins? That's not 14x features completed, but still... wow.

Comment by N_Lens 1 day ago

AI has 100x'd our productive capacity such that we're moving at unforeseen speeds at digging holes and refilling them!

Comment by shimman 1 day ago

This is too generous, they aren't even filling in the holes.

Comment by californical 1 day ago

It’s filling in a lot of the holes, but it’s putting a very convincing paper cover over the ones it misses. So it’s very hard to find the ones it didn’t fill, better hope your most valuable customers don’t walk over the paper ones!

Comment by kcartlidge 2 hours ago

With agentic stuff there's also a large amount of commits which are not code.

For instance with OpenClaw and similar, they often simulate institutional and short term memory with markdown files in folders. Other tooling that runs companies using agents as staff, for example, do the same - but also with files for inputs, outcomes, handovers etc.

All of this means a lot of extra churn as these kinds of files can be changing with every interaction not just every traditional commit point.

Comment by 3eb7988a1663 1 day ago

I am surprised it is that low. The Bun Zig-to-Rust AI port was 6755 commits in like two weeks. If you make 10 commits per working day, that is 2500/year.

While that is (hopefully) the upper end of the distribution, several companies have loudly encouraged engineers to light tokens on fire to the AI gods, so it only takes a handful of the devout to push up the average in gas town like ventures.

Comment by BobbyTables2 1 day ago

Is that even a lot?

Spread over a year, roughly estimating a generous 4 kbytes of data per commit, comes out to a throughput of a little under 2 MB/s.

Of course, it isn’t spread out uniformly and there is also a lot of hashing and other things going on.

Maybe pulls and clones drive more I/O ?

Comment by 3eb7988a1663 1 day ago

I suspect there is a cacophony of work that happens when a commit hits the server. That request needs to get replicated, git repositories need to be repacked, pull requests need to calculate diffs, CI jobs need to execute, on and on.

That's also just assuming the good-faith usage. There are probably plenty of adversarial and poorly behaved scrapers that are putting additional load on the system.

Comment by jamesfinlayson 1 day ago

Recalculate percentage of each language in the repo, recalculate top contributors, recalculate the stats for the committer's profile etc etc.

Comment by BobbyTables2 12 hours ago

Scalable algorithms and data structures have existed for decades.

Even if they had 10 billion users with 10 billion repositories it shouldn’t be a big deal on a home PC.

Comment by Aperocky 1 day ago

That commit count alone should not become any problem for infrastructure, even for Azure. They probably developed some ungodly mess with actions that did not/could not translate very well on Azure infrastructure.

Comment by nomel 1 day ago

Seems very reasonable, from my use. I commit much more often, as checkpoints, with branch rules that prevent force pushes/deletions, so the agents can't delete anything. And, suspect MS is only counting commits, and not the eventual squashes to one commit.

Comment by larusso 1 day ago

And every checking runs a whole CI run?

We had it internally with our teams that open a PR to then push like 10-20 more commits but never actually interested in the client builds etc. turned out they opened the PR as a checkmark/ way to share the current state. We set cooldowns and auto cancel for the ci. And then there is the developer who uses the CI compute to run tests instead of running them locally for various reasons. We had to remind that compute isn’t for free.

Comment by nomel 1 day ago

Nope. You can configure CI to not run for every commit of every branch (seems insane to have full CI for every commit, unless you don't allow your devs to push until done with something, which also seems insane).

Comment by 1 day ago

Comment by ashishb 1 day ago

I wonder what percentage of pull requests are cascading updates caused by dependabot and multiple code review bots reviewing those PRs.

My belief is it is likely 1% or more. And likely coming in as an avalanche.

Comment by alibarber 1 day ago

I was recently looking at a Python project to learn a bit about RTC and just generally hack around and try something out.

https://github.com/aiortc/aiortc

It was fun and I found the code nice and helpful.

I clicked on the PRs to see if there was anything interesting to look at. I started reading one when I just realised I was just reading someone’s Claude talking to GitHub Copilot. That was when I decided that the Dead Internet Theory had already happened.

Comment by not_kurt_godel 1 day ago

Saving this for the next time someone trots out the "All cloud providers are the same" line

Comment by rickette 1 day ago

Easy to criticize this, but I rather see GitHub survive than fail under its own success. So thanks for acting on this MS!

Comment by theojulienne 1 day ago

They were originally on their own datacenters + huge amounts of burst and ancillary stuff in AWS, the internal push to move away from "the competitor's cloud" after the acquisition was huge and entirely stupid. I'm ex-GitHub and was one of the annoying people constantly saying GitHub should only move where it was provably the best option for GitHub - the third best major cloud provider is likely never it on merits alone.

If this story is true, it's good that they finally realised that GitHub's performance and availability mattered more than using Microsoft's products. It would mean someone finally came to their senses rather than forcing a wholesale push to Azure - but I bet they still want to have it both ways even if they concede some AWS now.

Comment by jeffrallen 1 day ago

In the early 2000's they did the same dumb thing to WebTV: forced a move from Solaris to Windows, that completely wrecked the engineering team's velocity and motivation.

Comment by collabs 1 day ago

They did the same thing to Hotmail as well. It was fine running on FreeBSD/Solaris. Microsoft didn't need to switch to windows for no good reason.

Comment by JumpCrisscross 1 day ago

Interesting to see Anthropic go with xAI and Microsoft go with Amazon.

(Probably just tea leaves. If you wanted to be extra spicy, you’d note that Jassy just threw Fable under the bus.)

Comment by shrubble 1 day ago

AWS gets a lot of money from both the unclassified government work they do plus the more secretive work. If they were required to disclose as part of that, they would disclose - why put all those contracts at risk?

Comment by VirusNewbie 23 hours ago

>Interesting to see Anthropic go with xAI and Microsoft go with Amazon.

Anthropic has been very strategic about playing all the big cloud providers off each other. They're also in desperate need of inference, and I don't think AWS has the capacity to scale up as fast as they want. AWS may also be trying to 'force' them to use Graviton, and I don't think Anthropic is as big of a fan.

Comment by CSMastermind 1 day ago

Doesn't most of Anthropics revenue come from AWS still?

Comment by amazingman 1 day ago

Nature abhors a monoculture.

Comment by daft_pink 1 day ago

They should have made it work with hono workers so they could scale to infinity with free edge deployment on cloudflare - Senior JS Developer (Kai Lentit)

Comment by 1 day ago

Comment by locusofself 1 day ago

I have to imagine this is very temporary.

Comment by N_Lens 1 day ago

After a decade and hundreds of billions in spend, Microsoft has finally upgraded from Azure to AWS.

Comment by aykutseker 1 day ago

GitHub used to get code after someone had thought about it.

Agents are starting to use it while they think.

Comment by pjmlp 1 day ago

So this is an Hotmail moment?

Comment by UnlockedSecrets 1 day ago

Is this in reference to something that should be linked for those not alive or in the know at the time?

Comment by kqgnkqgn 1 day ago

I'm not sure if it was entirely true, but there are stories that after Microsoft bought Hotmail in the mid-90's, they quickly attempted to move them from FreeBSD (?) to Windows NT. But it failed miserably, and they went back to the original stack for another ~decade.

Comment by enigmo 1 day ago

the FreeBSD migration didn’t take that long - iirc this was the frontend migration to an IIS ISAPI.

the Solaris bits (storage and routing tables) took far longer - and again iirc the frontend had been rewritten in C# before all the Sun hardware had been decommissioned.

Comment by kqgnkqgn 11 hours ago

Ah sounds like you actually might have been there. I'm sure you must have some interesting stories from that time

Comment by enigmo 2 hours ago

not in the Hotmail dev team but about as close as you could get otherwise. I even reported to their general manager for a minute or two.

fun times for sure.

Comment by jeffrallen 1 day ago

Yes. This happened.

Comment by pmontra 1 day ago

Maybe this from 2012 https://www.theguardian.com/technology/2012/aug/01/microsoft...

Comment by 1 day ago

Comment by fortran77 1 day ago

Still, it’s mostly text. You’d think it wouldn’t be that much of an issue.

Comment by andyst 1 day ago

and programming is just typing!

Comment by stonogo 1 day ago

and now it's not even that!

Comment by fuckinpuppers 9 hours ago

Azure is hot garbage

Comment by sreekanth850 1 day ago

Soon we can see limits on free github account. I guess that is clean way to end this AI slop fiasco.

Comment by citizenpaul 1 day ago

My favorite part of this is that MS is just bending over to take it lest the gods scorn their "free" training data temple.

Comment by dr_kretyn 1 day ago

If they pull out "Embrace. Extend. Extinguish." on AWS then I'm going to be impressed beyond mad.

Comment by solumunus 1 day ago

How is this relevant or possible here.

Comment by Aperocky 1 day ago

That only work on smaller business/organization.