Box2D Updated 2025-07-16
Wikipedia dumps Updated 2025-07-16
Per-table dumps created with mysqldump and listed at: dumps.wikimedia.org/. Most notably, for the English Wikipedia: dumps.wikimedia.org/enwiki/latest/
The tables are "documented" under: www.mediawiki.org/wiki/Manual:Database_layout, e.g. the central "page" table: www.mediawiki.org/wiki/Manual:Page_table. But in many cases it is impossible to deduce what fields are from those docs.
William Shockley Updated 2025-07-16
This dude is the charicature of the evil scientist! It is so funny. Brilliant, ambitious and petty!
Quantum compiler benchmark Updated 2025-07-16
These appear to be benchmarks that don't involve running anything concretely, just compiling and likely then counting gates:
Video 1.
The Spider's Web: Britain's Second Empire
. Source.
2017. Directed by Michael Oswald. Adam Curtis vibes.
Some notable points:
XPath Updated 2025-07-16
XPath kind of died with the rise of CSS selectors around the beginnning of the 2010's. But that is a shame. XPath is a good standard, and was generally more powerful than CSS selectors for many many years.
House mouse Updated 2025-07-16
SELECT FOR UPDATE Updated 2025-07-16
An example where SELECT FOR UPDATE is a good solution to an use case can be seen at: nodejs/sequelize/raw/parallel_select_and_update.js.
The horrors of Sequelize Updated 2025-07-16
Microsoft Updated 2025-07-16
And also their monopolistic practices: United States v. Microsoft Corp.
However, like all big tech companies with infinite money, they do end up doing some cool things in their research department, Microsoft Research, notably for Ciro Santilli being:
Tin compound Updated 2025-07-16
torchvision ResNet Updated 2025-07-16
That example uses a ResNet pre-trained on the COCO dataset to do some inference, tested on Ubuntu 22.10:
cd python/pytorch
wget -O resnet_demo_in.jpg https://upload.wikimedia.org/wikipedia/commons/thumb/6/60/Rooster_portrait2.jpg/400px-Rooster_portrait2.jpg
./resnet_demo.py resnet_demo_in.jpg resnet_demo_out.jpg
This first downloads the model, which is currently 167 MB.
We know it is COCO because of the docs: pytorch.org/vision/0.13/models/generated/torchvision.models.detection.fasterrcnn_resnet50_fpn_v2.html which explains that
FasterRCNN_ResNet50_FPN_V2_Weights.DEFAULT
is an alias for:
FasterRCNN_ResNet50_FPN_V2_Weights.COCO_V1
The runtime is relatively slow on P51, about 4.7s.
After it finishes, the program prints the recognized classes:
['bird', 'banana']
so we get the expected bird, but also the more intriguing banana.
By looking at the output image with bounding boxes, we understand where the banana came from!
Figure 1.
python/pytorch/resnet_demo_in.jpg
. Source.
Figure 2.
python/pytorch/resnet_demo_out.jpg
. The beak was of course a banana, not a beak!
Wikipedia mentions "Since animal mtDNA evolves faster than nuclear genetic markers" with a few sources.
Some sources:
The problem with virtually all sleeping masks on the market is that they leave a lot of room near your nose for light to come in.
Ciro Santilli discovered a useful workaround for that: make the mask tighter with a swimming goggles!
Just make the goggles as loose as possible to not put pressure on your eyes, and then strap them over the sleeping mask.
If you are a back sleeper, put the googles forward as normal. If you are a stomach sleeper, put the googles on the back of your head, and the straps over the mask. This way you wont get your head squished by the goggles and the bed.
Once Ciro understood the idea, Googling "swimming googles sleeping mask" led to: mantasleep.uk/ might be a good option.
Another good Google is "sleeping mask blackout".

There are unlisted articles, also show them or only show them.