bookmarks

The GitHub Archive Program is a GitHub initiative to ensure that open source software is preserved for future generations. It includes both very-long-term storage such as the GitHub Arctic Code Vault and ongoing backups of public repositories by nonprofit organizations such as Software Heritage and the Internet Archive.

The “Greatest Hits” are archives of GitHub’s 17,000 most-popular and most-dependended-upon repositories, written to hardened film designed to last for 1,000 years, enclosed in beautiful museum-quality cases, and donated to three of the world’s great libraries, on three different continents.

approximately 170 million Americans spend, on average, one hour every day in an app that is designed to maximize psychological grip

earlier media relied on polished narratives (films with arcs, shows with seasons), TikTok turned culture into a never-ending feedback loop of impulse and machine learning.

TikTok's algorithm learns instantly from micro-behaviors. You can nuke your feed in minutes just by deliberately watching only one type of video—say, pottery clips or capybara memes. That’s because the algorithm heavily weights engagement signs per video rather than long-term user profiles

before today, each medium we’ve invented reshaped how we think and consume information. The printing press trained readers in linear, sequential thought, encouraging sustained focus and complex argumentation. Television created visual storytelling and shared cultural moments, with families watching the same shows at the same time and building collective references. The internet introduced hyperlinked thinking, enabling rapid information switching, comparison of perspectives, and knowledge-building through exploration. Now, the world is picking up on the TikTok model

movie trailers increasingly resemble TikTok compilations: rapid-fire montages of action sequences and emotional beats rather than traditional narrative setups

you're not consuming culture; you're teaching a machine how to feed you culture more efficiently

markets where dozens of apps competed brutally for user attention. Platforms that kept people engaged survived; the rest died. That evolutionary pressure produced increasingly sophisticated ways to capture attention

We gain instant access to exactly the content that triggers our reward systems, but we lose the ability to be bored, to sit with incomplete thoughts

We lose the serendipity of discovering things we didn't know we wanted

To find what you need online, you require a solid general education and, above all, prior knowledge in the area related to your search

Culturally, it is increasingly the case that the first step is to go online, search for the a desired end result of a thinking process instead of engaging in learning

We detrain ourselves out of the ability to access the quality of the information and turn it into actual knowledge

We need a fully developed mental map of the subject in order to derive value from the results of an internet search

If you can’t produce a comprehensive answer with confidence and on the whim the second you read the question, you don’t have the sufficient background knowledge

Assuming that you have no comprehensive knowledge in health and fitness, terms like “aerobic capacity” or “core stability” only activate some superficial level associations

use Illuminate\Support\Facades\Cache; 2  3$lock = Cache::lock('foo', 10); 4  5if ($lock->get()) { 6 // Lock acquired for 10 seconds... 7  8 $lock->release(); 9}

If you want to support your favorite authors: send them money. A dollar helps more than viewing ads ever would

Yesterday one of the biggest package ecosystems had very popular packages get compromised.

Traditional AI models are like that, highly optimized systems capable of extraordinary performance, but only in a limited, data-driven way. They react to data, analyze it, and process it, but they cannot move beyond that framework

it is logical to use the human brain and its neurons as a reference to model artificial systems, rather than scaling the systems in their current configurations and hoping to see generalized intelligence

Some in the AI community believe that if we make our existing AI models bigger and faster, eventually they will start flying/exhibiting generalized intelligence

Scaling current AI models would have been a logical approach, only if their fundamental building blocks (nodes/neurons) are inherently capable of exhibiting generalized

in a sense we can get a motorcycle to "fly" by building a ramp and launching it into the air. But that’s not sustainable flight, it is a temporary solution based on brute force, not the inherent ability to fly.

Religiosity tends to decline across generations

first, public ritual participation declines; second, the importance of religion to individuals declines; and third, people shed religious affiliation

we recommend caution in interpreting longitudinal claims, due to limited data

It is well established that religion declines in Western countries

The model links modernization, particularly institutional and technological innovation, to a decline in the symbolic and social functions of religion3

biomedicine replacing religious healing, welfare states replacing spiritual reassurance—religion loses functional relevance

credibility enhancing displays (CRED’s

The central idea of CRED’s is that individuals will believe and imitate others with a higher probability if these others back up their words with hard-to-fake deeds6

Because of the increasing efficiency of secular goods, parents may still teach their children religiosity—but stop backing it up with religious behavior. This will lead children to drop their religiosity levels.

The term has become an idiom describing a group of persons, each agreeing to perform an impossibly difficult task under the misapprehension that someone else will be chosen to run the risks and endure the hardship of actual accomplishment