Is MS365...
...the amount of outages it has?
Microsoft has reported two Azure service wobbles in as many days, including a disruption affecting Virtual Machine management ops yesterday and a Managed Identity for Azure resources outage in East US and West US regions today. According to Microsoft, today's Managed Identity for Azure resources issue impacted the East US and …
There was an interesting article last year about Atlassian running a DR exercise and discovering their architecture had _tons_ of circular dependencies: https://www.theregister.com/2025/11/25/atlassian_dependency_migration/
Atlassian remedied some of those issues by essentially turning the dependency graph into a DAG where possible. I've never seen a similar analysis for Azure, but I'd be surprised if they're much better.
I don't think anyone knows how long a restart of these distributed systems would take after a total failure (e.g high-altitude nuclear detonations, Carrington event). Hopefully we'll eventually have plans similar to what utilities have for "black starts" of the power grid.
Maybe, just maybe, we are all (dare I say it) becoming just a tat too dependent on third party SPOFs hosted by overseas tech giants who care about nothing but their stranglehold on the market and their bottom lines? Cloudflare, AWS, Azure and what not have lately proven to be more of a liability than a boon to the IT using world. Or maybe that's just me.
I have heard (off the record from someone who worked at MS ) the internal MS staff have to use AI tools for all coding, and not allowed to write code by hand
so I wonder how much of this is relying on AI to generate the code for these updates and patches, then using AI tools to test it before deploying
if it is the case, my answer would be that if they can't make it work reliably why should we use it
I have already seen the mess one of our devs is making using various "AI coding tools" and how much longer it is taking to clean up when it all goes wrong, as it does more often than not
I have already seen the mess one of our devs is making using various "AI coding tools" and how much longer it is taking to clean up when it all goes wrong, as it does more often than not.
This seems like a rather big risk, ie developers turn into auditors, until that function is AI'd. AIs verify their own code (for a fee, naturally) and there's fewer people who can figure out what the code is or does on production systems. Which might then mean developers who've cut their teeth on debugging & fixing AI slop will be able to pick & choose their own contracts.
I would compare it to the chip industry except... Synthesis became a thing quite awhile ago. The logic blobs created were huge from some RTL code. No human was going to sort thru the blob to verify it. The difference was it was not an AI, but a highly reliable algo that did the transformation. And even with that, a logic sim is run on the output to validate, along with timing of course. Pity AI is not more deterministic.
> Those same developers would likely have some advice for Microsoft on testing changes before deploying to production.
How does one test a configuration change on a cloud platform. Seeing as it can't ever be turned off or rolled back to a previous instance.