ody4242

7 ms·

1.
by ody4242 1mo ago
then you switch provider, it's not a monopoly
2.
by ody4242 2mo ago
The upfront cost of facility is ~30%, network infra 10-15%, land/utilities is small percentage, power could be significant for an AI DC. The servers are ~50-60% only.
3.
by ody4242 2mo ago
They are building new datacenters for the AI demand, so around half of this CAPEX is not for the GPU-s, and those will not be replaced every 3-5 years. Also, TPUv2 was introduced in 2018, and still not completely retired in all regions, fro
4.
by ody4242 2mo ago
This is a good chart that shows historical CAPEX spending. Hyperscalers have been through a couple CAPEX cycles like this, they all know what they are doing. https://eco3min.fr/en/big-tech-capex-revenue-ratio-quarterly.
5.
by ody4242 2mo ago
Why would they go into price war when they are compute constrained?
6.
by ody4242 5mo ago
try to add it here: https://gemini.google.com/saved-info
7.
by ody4242 6mo ago
Idk how an AWS region would respond to a power outage, but i have tested this in AWS Outpost, and there, if you power down a rack, then power it back again, the baremetal instances will not be recreated. (I was surprised as I was expecting
8.
by ody4242 6mo ago
I would have benchmarked with an instance that has local nvme, like c8gd.4xlarge.
9.
by ody4242 7mo ago
I'm 100% sure that all providers are playing with the quantization, kv cache and other parameters of the models to be able to serve the demand. One of the biggest advantage of running a local model is that you get predictable behavior.
10.
by ody4242 7mo ago
Yesterday GPT 5.2 wrote a python function for me that had the import in the middle of the code, for no reason. (It was a simple import of requests module in a REST client...) Claude I agree is a lot better for backend,Gemini is very good f
11.
by ody4242 8mo ago
You don't need hundreds of services. Give me virtual machines, reliable block storage, file storage and object storage, networking, dns, managed kubernetes, and it will cover the majority of workloads in Europe that run on Openshift or
12.
by ody4242 10mo ago
Sure, EBS or any network-attached storage is expected to be a lot slower than a local SSD for synchronous writes or random reads, as there is a network stack in between. But my understanding is that for those usecases, you can use metal ins
13.
by ody4242 2y ago
MS does sell ads, they have an advertising platform.
14.
by ody4242 3y ago
Terraform Crossplane (although it's quite new, and missing a lot of things that Terraform has)
15.
by ody4242 3y ago
The Model3 has a dedicated button for the windscreen wiper on the steering wheel, and then you can adjust with the scroll buttons. It certainly has a learning curve, and it would be better if you could rely on muscle memory for these things
16.
by ody4242 3y ago
What makes you stand out from this list of other managed openstack providers? https://www.openstack.org/marketplace/public-clouds/
17.
by ody4242 3y ago
I have seen similar numbers from multiple sensors across our city during the fireworks at New Years eve.
18.
by ody4242 4y ago
Imho, if Tesla could train a huge neural network to demonstrate that self driving is possible with their sw stack, they would have done that already to let Elon Musk have it's road show with it's capabilities. (Running ChatGPT doe
19.
by ody4242 4y ago
Where is wallstreetbets? :)
20.
by ody4242 4y ago
Telco vendors have been using Openstack for 6-8 years already, and virtualization (and even containerisation, mainly with LXC, or Solaris zones earlier) was a thing back then. Kubernetes is also used for the last couple of years, mainly ins
21.
by ody4242 4y ago
There are some exotic scenarios, where copy-paste is the good(fast) solution. For example, one of my clients can be accessed via a VPN. Once I'm on the VPN, I have to start their VDI. Inside the VDI, I need to start another VPN, that
22.
by ody4242 5y ago
Interestingly, the site is blocked by Symantec Endpoint Protection.
23.
by ody4242 5y ago
Here you can choose between MDI and pump.
24.
by ody4242 5y ago
I have no idea, that's why I am not using an insulin pump either:)
25.
by ody4242 5y ago
I use Dexcom G6 CGMS with Nightscout and xDrip+. xDrip+ runs on Android, gets the data from the Dexcom transmitter, and sends it to a Pebble watch and also to a mongodb database. xDrip itself supports a lot of CGMS systems, I've been u
26.
by ody4242 5y ago
Their repo has like 300 open issues today... I have type 1 diabetes, and use open source tech for getting the glucose readings from my sensor and pushing it to a Pebble, but insulin dosing is a way more complex story.
27.
by ody4242 5y ago
As someone who has an Intel NUC w. Ubuntu, a Thinkpad with Windows 10, and a Macbook Pro 2015, I have a totally different experience. I spend like 5% of my total effort on OSX to make things working, the remaining 95% is spent to make thing
28.
by ody4242 5y ago
I would expect that if you read data - the GIF image in this case - from the block device, it will stay in pagecache, until there is memory pressure.
29.
by ody4242 5y ago
You don't have to be a victim. If you earn x, and she earns y, it's perfectly OK in a relationship to get a job that pays only y. I think your wife doesn't want you to be a victim either, maybe she is just not good with finan
30.
by ody4242 6y ago
I did not claim, that it's causing a lot of deaths, I said that it's been linked to the radiation. Thyroid cancer is not fun, even if you survive it. The treatment and followup is a financial burden (on the person or the society,

More ›