ken

yo

[Joshua Reed Eakle 🗽 @JoshEakle]

Today, I went to the grocery store and purchased $20 worth of vegetables.

I now have a $20 trade deficit with the grocery store.

They clearly ripped me off.

I will now cut off all trade with the grocery store and let my family starve in order to Make My Household Great Again.

[∞ @TheFalseNein]

What Viktor Gyökeres taught me about B2B sales

[Tamale @256Rootyherman]

Arteta giving Arsenal fans bragging rights again

[C. @ArsecaI]

How it feels seeing an Arsenal title

[Karen Resorcé @hr_unhinged]

I noticed one employee wearing noise-canceling headphones.

Not earbuds.

The big, cushioned over-ear kind that create a tiny personal universe.

I asked if everything was alright.

He said yes, he’s just trying to focus.

I told him we value focus, but isolation can misread as resistance to collaboration.

He said he’s literally sitting at his desk doing his job.

I told him we track human presence, not just output.

He asked how presence is measured.

I said imperfectly, which is why it's so important.

Then I logged “avoiding spontaneous culture building opportunities” in his engagement profile.

[HLTCO @HLTCO]

I can’t possibly imagine who this award might have been invented for. 🙃

[amrit @amritwt]

what did i just read bro

[laurence @functi0nZer0]

These will always make me laugh

[Jacob @Jsize18]

Apparently Kerkez vented in electrical. Not sure how true it is

[MAX @MaxGrant__]

last night my girl told me to grab protection, but when i looked in the drawer, it was empty.

as she laid there waiting, primed for action while i fumbled frantically around the nightstand, i realized something about email marketing for e-commerce brands.

your customers have a hole that they NEED you to fill. they are READY for your product. there’s a problem they need solved right now. they WANT your automated, seductive email flows deep inside their customer lifecycle, turning cart abandoners into repeat addicts.

they all WANT you to succeed. they don’t care about your brand palette or texture of your packaging or how much you raised or your reach or how many ABGs you have making UGC content.

when an customer needs your product, they think okay, how much can we spend on it, and what are the options available?

it doesn’t matter how much you charge, just be the cheapest or the most expensive.

the important thing is that you BE THERE when a customer is fumbling around in the dark.

you can’t solve their problem if you’re lost under the bed with the rest of the ran-through brands.

[Rachin Kalakheti @rachinkalakheti]

How did we end up polling on Discord for Nepal’s next leader? Here’s a rundown:

tldr; Nepali government was overthrown in 48 hours and I think i just witnessed the first internet native revolution.

Over two decades ago, Nepali citizens overthrew the Monarchy to establish a multiparty democracy hoping for a better future. However, two decades later, the leaders of three major democratic parties, who overthrew the monarchy, were still fighting for power and greed.

They went from no slippers on their feet to living in lavish mansions, but the common men were still poor, with no sign of improvements.

However, something changed from around late 2010s. People who relied on national TV and radio for their information started getting wide access to internet. People could see every mishaps these politicians made in real time, every frustrated candid stories of every Nepali citizens, and the rapid development of every other country that used to be on a similar level to Nepal. Frustration started accumulating rapidly among people, all enabled by the internet.

Fast forward to around August 2025, Nepali youths, frustrated by the corruption and incompetencies of the government, started a trend on TikTok called "Nepo Babies" where they created reels comparing the lavish lifestyle of politicians' kids with common citizens of Nepal who couldn't even afford a proper restaurant meal.

This exposé got widespread traction on TikTok and Instagram, which infuriated all the citizens of Nepal who had already lost hope from the current politicians. Everyone started cyberbullying their kids for their lifestyle.

Coincidentally, Nepali government imposed a ban on almost all major social media sites on September 4th 2025 saying they need to register their entities in Nepal and follow a strict content moderation rule imposed by the government, which obviously, these sites declined to.

The problem is over 20% population of Nepal resides outside the country and suddenly they had no proper way to communicate with their relatives back home. This lack of empathy, combined by the suspicious timing around the rise of nepo baby trend and a long standing disappointment in the current establishment was too much for everyone.

Suddenly, on September 5th, a call for a peaceful protest on September 8th started circulating on TikTok, Reddit, Discord, Instagram, etc.

What’s interesting is, this was a decentralized protest, all organized via internet without any central figure. Since it was organized collectively by youths on internet, this protest was given the title of “Gen-Z protest.”

A huge 12,000+ mass appeared on September 8th protest. It started peaceful, but things suddenly got violent. The government gave an order to shoot which led to the death of 20+ people and 300+ injuries by end of the day. Gut wrenching pictures and videos from the protest started floating on internet which led to a widespread outrage against the police and the government.

Later that night, Government gave a very lackluster and apathetic response to such a devastating situation. With this, peace wasn’t an option anymore.

Reddit, Discord, and TikTok was filled with rage and call for revenge on September 9th. People started sharing tutorials on how to create a Molotov cocktail, how to disarm tear gas, etc. This felt like the heat of the moment, but no one expected what was about to come the next day.

Next day, everyone from every part of country started coming out of their houses and protesting against the government. Around early afternoon, news started circulating on internet that a big group was headed towards the house of the Prime Minister. Then another group headed towards another big politician, then another, then another. All of their houses were burnt.

Internet facilitated a realtime coordination for mimetic actions across the nation. This is why “burn their mansions” became the trend for the day within a couple hours across the country. Police force, who were already demoralized from last day’s actions and reactions, showed little resistance on this day.

Almost every major politician’s house was burnt, and many politicians were even beaten brutally by the crowd, on streets, on rivers, and in their mansions.

I was on Reddit and Discord all day. It truly felt like dropping at a warzone where everyone is coordinating attacks, sharing real time updates, and even telling which locations to avoid, etc. By late afternoon, the prime minister resigned and took refuge under the army alongside many other politicians. Army took control for security later that night and wanted to host negotiations with the protestors.

The problem was, there was no representative for the protest. It was all a collective decentralized effort. So people started discussing and polling on Discord and Reddit on who should represent the protestors and who they should nominate as their future leader.

It’s been a chaotic effort (who knew decentralization would be chaotic), but they’ve finally agreed upon who they want to nominate. Discussion between Army, the President, and the protestors is still ongoing, but it’s almost guaranteed now that the leader nominated by the Discord polls will take the position within a couple days.

It was an unfortunate event where many things went wrong and the destruction pushed Nepal back by years, but the sequence of events were very unique, where almost everything originated and accumulated from the internet. The dissatisfaction accumulated for years on internet, and suddenly in two days, the government was overthrown, all coordinated via Internet with no central figure. In many ways, the protest happened on the internet for years, but the government didn’t listen.

I suspect this pattern will emerge across many other countries now that the world has seen what’s possible. The Network State is becoming a real thing.

[anshuman @athletic_coder]

She dumped me last night.

Not because I don't listen.
Not because I'm always on my phone.
Not even because I forgot our anniversary (twice).

But because,

in her exact words:

"You only pay attention to the parts of what I say that you think are important."

I stared at her for a moment and realized...

She just perfectly described the attention mechanism in transformers.

Turns out I wasn't being a bad boyfriend. I was being mathematically optimal.

See, in conversations (and transformers), you don't give equal weight to every word. Some words matter more for understanding context. Attention figures out exactly HOW important each word should be.

Here's the beautiful math:

Attention(Q, K, V) = softmax(QK^T / √d_k)V

Breaking it down:

Q (Query): "What am I looking for?"
K (Key): "What info is available?"
V (Value): "What is that info?"
d_k: Key dimension (for scaling)

Think library analogy:

You have a question (Query). Books have titles (Keys) and content (Values). Attention finds which books are most relevant.

Step-by-step with "The cat sat on the mat":

Step 1: Create Q, K, VEach word → three vectors via learned matrices W_Q, W_K, W_V
For "cat":

Query: "What should I attend to when processing 'cat'?"
Key: "I am 'cat'"
Value: "Here's cat info"

Step 2: Calculate scoresQK^T = how much each word should attend to others

Processing "sat"? High similarity with "cat" (cats sit) and "mat" (where sitting happens).

Step 3: Scale by √d_kPrevents dot products from getting too large, keeps softmax balanced.

Step 4: SoftmaxConverts scores to probabilities:

"cat": 0.4 (subject)
"sat": 0.3 (action)
"mat": 0.2 (location)
"on": 0.1 (preposition)
"the": 0.1 (article)

Step 5: Weight valuesMultiply each word's value by attention weight, sum up. Now "sat" knows it's most related to "cat" and "mat".

Multi-Head Magic:Transformers do this multiple times in parallel:

Head 1: Subject-verb relationships
Head 2: Spatial ("on", "in", "under")
Head 3: Temporal ("before", "after")
Head 4: Semantic similarity

Each head learns different relationship types.

Why This Changed Everything:

Before: RNNs = reading with flashlight (one word at a time, forget the beginning)

After: Attention = floodlights on entire sentence with dimmer switches
This is why ChatGPT can:

Remember 50 messages ago

Know "it" refers to something specific
Understand "bank" = money vs river based on context
The Kicker:Models learn these patterns from data alone. Nobody programmed grammar rules. It figured out language structure just by predicting next words.
Attention is how AI learned to read between the lines.

Just like my therapist helped me understand my focus patterns, maybe understanding transformers helps us see how we decide what matters.

Now if only I could implement multi-head attention in dating... 🤖

Still waiting for "scaled dot-product listening" to be invented.

[Arsenal @Arsenal]

Welcome to Arsenal, Viktor Gyökeres 💪

[ALBY LAD 🎄🎅🏻 @AlbyLad_]

Huge news for Owl City this

[͏͏͏𝐉𝐚𝐲 ⌖ @cagiago_]

Away games on the last day of the season never made sense to me, everyone should play at home so the fans can enjoy and give a proper send-off before the long break.