🪈Rooster 🪈Cogburn🪈 retweeted
Sorry guys the way the banks are right now I will not be able to pay anyone at all.
So don’t ask! Ask the banks not me.
Sorry and also when they are back up and running I still won’t be able to sorry this inconvenience is insane.
BREAKING: DN notice issued in 48 of 50 states.
Authorities are advising residents DEEZ NUTS will be in effect indefinitely.
2
44
🪈Rooster 🪈Cogburn🪈 retweeted
As a big 'monitor the situation' guy, there are currently way too many situations to monitor.
🪈Rooster 🪈Cogburn🪈 retweeted
I think people seriously underestimate how much information was moving around North America long before Europeans arrived.
Ideas, materials, technologies, people. Across enormous distances.
🪈Rooster 🪈Cogburn🪈 retweeted
I am highly skeptical of this claim from a former OpenAI employee who claims he can look at the open weights of an AI model and determine which chipset the model was trained on. Sounds like another OpenAI psyop to me, to push the narrative that Nvidia chip exports need to be restricted even further. I very much doubt the claimed pattern here even exists at all. The weights have no knowledge of the GPU that altered them. Those elements are several abstract layers apart during the model training process.
I have a major result to share.
I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China.
This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model.
Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves.
I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems.
I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used.
It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20.
These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs.
I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at unemploymentlabs.com/checkpo….
There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.
every once in a while @Zai_org GLM just randomly starts outputing answers in Chinese for me
nbd roight 
21
entire customer experience programs could be built based on the principles in this thread
We're investigating reports of users experiencing issues accessing Exchange Online mailboxes through multiple connection methods in the Americas. More details can be found in the Microsoft 365 admin center under EX1550543.
26
the context window on codex can’t be more than like 256k right now
shit compacts every other question while
GLM chugs along
1
26
kind of sounds like Iran scaled up those cyberattacks?
Non-nerd explanation of the Outlook, etc, outage today for those who care
An ominous *something* caused a whole metric shit ton of Microsoft cloud servers to shit the bed. This caused all the data they would normally have handled to be sent to the servers that didn’t shit the bed. THOSE servers screamed holy terror “TOO MUCH!! TOO MUCH!!”, like a virgin taking Mandingo and everything shit the bed.
They still haven’t fixed it and appear to not know what caused it yet.
52
boy howdy
Replying to @MSFT365Status
We’re applying infrastructure capacity improvement where possible to help support traffic in regions coming online. We’re continuing to monitor the results of changes applied so we can identify strategic actions that can help to improve service health across all affected regions. More info under EX1550543 and MO1550821 in the Microsoft 365 admin center, as well as msft.it/6016xBHGu if the admin center is not available to you.
74
🪈Rooster 🪈Cogburn🪈 retweeted
A portion of their infrastructure is experiencing elevated load because a number of servers had unexpectedly become unavailable
Those servers were failing automated health checks, causing them to be removed from service
Health checks are automated tests that infrastructure uses to determine whether a server is functioning properly and should receive requests
When a server fails these checks, traffic can be redirected elsewhere to protect users from an unhealthy system
But when large numbers of servers become unavailable, the remaining infrastructure inherits more work
Microsoft confirmed that this was happening across multiple regions
The company also discovered that many affected servers had become unresponsive, making it difficult to verify whether its attempted fix was working
It began restarting affected servers, redistributing traffic across its infrastructure and testing an alternative health-check method
Microsoft also paused deployments and configuration changes while investigating whether recent configuration updates had contributed to the incident
The alternative health-check method worked on a subset of servers, and Microsoft began applying it more broadly
There was another strange part of the incident
Downdetector showed simultaneous spikes in reports involving Steam, AWS, Google and Cloudflare
But simultaneous reports don't establish that these companies suffered a common outage
Cloudflare's status reporting showed no corresponding widespread incident, Google's cloud status page also showed no issue, and AWS reported a localized Tokyo problem
The Microsoft disruption was independently confirmed
The wider internet outage narrative was not
This incident is a useful reminder of how cloud infrastructure can fail in unexpected ways
A health-check mechanism is supposed to protect availability by identifying unhealthy servers
But if the mechanism itself starts producing incorrect results or affected servers cannot respond to it, the recovery process can become part of the problem
And at Microsoft's scale, losing enough infrastructure across regions can affect services millions of people depend on every day
the beatings will continue until morale improves
🚨 BREAKING: MICROSOFT'S OUTAGE IS PROVING HARDER TO FIX THAN EXPECTED.
Microsoft admits its recovery efforts have failed to deliver the expected relief.
A core server process is being overwhelmed, causing servers to stop responding.
Engineers are isolating infrastructure and attempting controlled restarts.
Still no confirmed restoration.
Microsoft's cloud is struggling to recover.
42




























