For context, this isn’t 100% true.
This is from a third-party tracker, who took the downtime of every GitHub hosted service and added them together. So a bunch of stuff that was down 0.5% of time and/or some beta+new release services down for longer make it look way worse than the core service uptime actually is.
For even more context: That means that 89% of the time all parts that make up github work without issue. 11% of the time at least one component has issues/downtime.
https://mrshu.github.io/github-statuses/ shows the breakdown, git push/pull operations for example have 98.98% uptime.
Good info. Either way, switch to Codeberg. 😉
Switched to self-hosted Forgejo already so now I’m just waiting for my dependencies to switch.
10 minutes ago my forgejo test failed because github returned a 502 for the home-manager repo •-•
Yea, I run Gitea for my personal work and we run our for our business, mainly because I started before the Forgejo fork. At some point I’ll migrate over.
I see two nines there.
good job vibe coding your ass to losing customers… hopefully.
Low to average reliability is fine if the service is cheap, and if that avoid the need for backup diesel generators in datacenters.
I doubt this applied to Github:
Microsoft to use diesel-fired generators as backup power for data centers
This is what happens to nearly every business Microsoft buys or invest in. They’re the enshittiers.
Sony is a close second, BTW 😁
Sony is a close second, BTW 😁
Crunchyroll had a major security breach earlier this week
Cloudflare has been pretty bad too lately. Its like the big companies are not even trying anymore. And if a real war broke out…their facilities are all centralized and VERY easy to target. Keeps me up at night sometimes since a vast majority of services at work are all on central servers.
In theory the internet self corrects. But in practice, if AWS/Cloudflare/MS/etc…goes down, a LOT of other services you dont even know about are effected. Last AWS issue took down Azure as well as people were scrambling to get servers back up and running (among other things).
it’s because all these major service providers are desperately attempting to inject AI into their operations without a goal in mind. they’re literally just throwing money at the wall and wishing something will stick.
at my current employer they are mandating AI be used by every department. didn’t say how or what the goal was, just “use it or get fired”. last year it became a KPI.
so far this year, we’re $13m behind projected goals. labor is down 15% because we stopped hiring people. productivity is down 30% because the ones that remain are getting burnt out by the increased workload.
I have hourly techs tell me they are being encouraged to work off the clock.
2026 is going to be worse than 2025, even if the bubble pops.
I highly doubt it’s just that, I wouldn’t be surprised if most of the devs who actually put in all the work upfront have left or been removed. See the same shit at Google, remember when it was actually a good search engine? Priorities shifted to making money above all else instead of making good products.
Who needs external malicious actors when you jam AI into your workflow?
with friends like ai, who needs enemies?
Ah the classic EEE strategy: Embrace, expand, enshittify.
That’s what AI first development gets you.
They were doing that before “AI” development was a thing
https://mrshu.github.io/github-statuses/
Look through the history. In 2024 and before that they consistently had 97+%, only rarely dipping down to 95 and once 92%.
Since February 2026 they have been constantly in the 80s.
Ok…? That doesn’t change the fact that Microsoft was enshitifying the software they bought before “AI” was a thing. They didn’t suddenly start doing it when LLMs happened.
They enshittified, yes, but uptime isn’t a topic of enshittification. You don’t make more money by reducing uptime.
Uptime, especially for a critical service like Github, that tons of businesses actually depend on, is a sign of not being able to keep your infrastructure and development under control. And getting below 90% is really, really shameful.
Sony is a close second, BTW
Do you have any examples? I’m not familiar with any major acquisitions Sony has made. Afaik the “acquire company, fire everyone, and run the business into the ground” strategy is mostly an American phenomenon.
Sony pictures just closed Pixelmondo after running it further into the ground
Don’t forget closing Bluepoint after multiple very successful remakes!
If they just let them remake Bloodborn, they would’ve been allowed to just print money
There’s two nines there: “95 issues in last 90 days”
Nein Nein Nein!
There are two more surrounding the decimal point, too! Four nines!
And if you sum four 9s you get 36. Sum those digits and you get 9 again. I think that’s how numerology works?
and a partridge in a pear tree
What? I see two nines!
picard.jpeg
Technically still 9’s just in the wrong places.
Technically it has exceeded 9% as well.
What an absolute achievement, nice work Microslop!
Really makes it want to migrate to Azure for your hosting, eh?
I work with an azure database and 15% (!) of the time the connection handshake times out for no reason known to man.
This is well documented on their end. The solution? ‘Implement a retry logic’Azure’s documentation is the worst fucking bullshit that I’ve ever read in all my days, and just about every single page or tool (including the CLI) has an integrated slopbot that routinely recommends commands and REST endpoints that don’t exist; it’s slow as fuck, and to do even the simplest things is agonising. But to give them their dues, their recent uptime has been pretty good.
Truth be told, I’ve even come round to thinking that I prefer using Azure to Google Cloud Platform. Using any of Azure’s features is a pleasure akin to cutting yourself with a rusty nail and then falling in a sewer, but at least it has some features. GCP is like they implemented a quarter of the very basic functionality and then got fed up, decided to call it a day.
Deleting or creating resources will randomly take 20-45 minutes sometimes lol
I worked with some Azure devs a few years ago. they looked tired. I mean real fuckin worn down.
compared to the Amazon guys, it was like night and day. not to say AWS docs are much better, but they are better.
I’ve been using it for a while now and am surprised how shitty it is. I remember not having a routing table and it allowed a single packet to be routed to the endpoint and then stop, it doesnt even make sense how that can happen. Intune is even worse, its missing extremely basic functionality, you cant even clone a device config its that bad.
