Samsung and SK hynix have been building 12-high HBM4 (36 GB) as the default stack for NVIDIA’s Vera Rubin. NVIDIA has asked both suppliers to raise the share of 8-high HBM4 in second-half 2026 shipments, and that HBM4E may also lean 8-high. The stated reasons are thermal headroom, packaging yield, and supply flexibility: HBM4 doubles the I/O width to 2,048 pins versus HBM3E, which raises bandwidth but also heat and power in already dense GPU packages. An 8-high part uses fewer DRAM dies per cube, so more cubes can be produced from the same constrained wafer starts.

Major trend I'm seeing in hiring: people are asking to be part-time, come in three days a week, and have a 4-day weekend to work on their personal projects. Let that sink in: people are asking for 60% of the salary of a full-time job! Positive that people are upfront that they're not interested in doing a 40- or 50-hour workweek I could see the United States moving to a standard 4-day workweek in the age of AI, with people getting fewer named vacation, sick, personal and holidays, etc. Abstract the work agreement to 4 days a week × 50 weeks a year = 200 days of work for X salary But you have to Come to an office and actually work there's going to be a grand negotiation that occurs over the next decade or two

The US murder rate didn’t quite set a record low in 2025, but it may get there this year. The reason is elusive, though fewer young men may be part of the puzzle, @foxjust says (via @opinion)

GERALD & THE FEATHERED BANDITS — PART IV: THE CROW AND THE CANDLE Three knocks. That's how last night ended. Three slow, deliberate knocks from beneath my home. THUMP. THUMP. THUMP. Ill fortune always bears three fruits. And apparently... one of them lives in my fucking basement now. Fantastic. Sleep did not come to me. For Kitten it did. How? I could not tell you. I spent majority of the night sitting on my couch staring at the ground while Gerald remained perched outside my window. Watching. Waiting. No movement in the slightest. Every so often I'd glance over at him. He'd glance at me. Neither of us hummed. Our relationship had reached the silent-treatment phase within the marriage. Around 3:17 AM, Kitten wandered into the room nuzzled in a blanket. "Has Satan's pigeon moved?" "No." "Anymore knocking?" "No." "Fantastic." She turned back around. "Where are you going?" "Back to bed." "THERE IS AN ANCIENT EVIL UNDER MY HOUSE!" She took a pregnant pause. Looked at me. "And what exactly would you like me to do about that at three in the fucking morning?" ... Valid. She disappeared down the hall. I looked back toward Gerald. "Your recruit is incredibly useful." From outside: Hmmm. "Fuck off." But something had been bothering me. Gerald knew that thing was down there. He had known from the very beginning. The map. The crown. The scouting. The army. He didn't wake it by accident. He wanted it to be awoken. And suddenly... I wasn't sure whose side I was actually on. At sunrise, Gerald finally moved. He tapped the window. Once. Twice. Thrice. "Absolutely fucking not." Tap. Tap. Tap. "Gerald, we've already established three is incredibly ominous." Hmmm. Then he flew away. I sighed. Grabbed my stick crown. And followed him. Because apparently survival instincts are most optional equipment and mine didn't come with factory-installment. Kitten followed behind me. "Where are we going?" "Your guess is as good as mine." "Excellent. Loving that for us." Gerald led us back into the woods. But not toward the dungeon. Not toward the stone. Somewhere else. Somewhere deeper. Eventually we reached an behemoth dead oak. Its branches were twisted together like fingers. Hundreds of strips of black cloth hung from them. At the base of the tree sat something strange. A candle. Just one. Gerald landed beside it. Kitten stopped. "No." "He hasn't even done anything yet." "There is a single candle beneath a decayed tree covered in black ribbons. I've seen enough horror movies to understand basic fucking foreshadowing." Gerald ignored her. He tapped the ground. Scratched beneath the leaves. And uncovered something. A small stone tablet. There was writing carved into it. No symbols this time. Just words. Old and weathered. But legible. At the top: THE CROW AND THE CANDLE Gerald perched beside it. And somehow... I knew he wanted me to read. So I did. Once there was a crow who ruled over a starving flock. Winter had swallowed the land. The trees bore no fruit. The rivers became stone. And each morning the crow awoke to fewer voices than the morning before. One night, the crow saw a candle burning inside the home of Man. The crow knew that fire had belonged to the Man. The crow knew taking it was forbidden. But the crow also knew his children were freezing. So he stole the flame. He carried it home in his beak. The fire warmed the first nest. Then the second. Then the third. And the flock rejoiced. The crow had saved them. But wherever the candle burned... a shadow appeared. At first, the shadow was small. So the flock ignored it. As the flame grew... so did the shadow. Until eventually... the shadow became hungry. The flock begged their King to extinguish the flame. But winter still surrounded them. Without the fire, they would freeze. With the fire... the shadow would feed. And so the King was given a choice. Extinguish the flame and allow many to perish. Or... choose one bird each night for the shadow. I stopped reading. Kitten stopped making jokes. The latter concerned me more than anything. I looked at Gerald. "What did the King choose?" Gerald didn't move. I looked back at the stone. There was one final line. I brushed the dirt away. And read it. The King fed the shadow. Silence. "Jesus Christ." Kitten whispered. Gerald looked at me. I suddenly understood why he'd brought me here. "No." Gerald tilted his head. "No, Gerald." Hmmm. "I know what you're asking, and I am saying NO." Kitten looked between us. "Well, fantastic, because I fucking don't." I pointed toward the story. "He's asking what I would choose." Gerald nodded. Kitten stared at him. "You're giving her a fucking ethics exam?" Hmmm. "YOU KIDNAPPED ME FOR THIS ORGANIZATION AND NOW THERE'S HOMEWORK?" Normally I would have laughed. I couldn't. Because Gerald was still staring at me. Waiting. I looked back at the tablet. One life. For many. Simple mathematics. Horrifying mathematics. And suddenly the crown on my head felt weighted. "I wouldn't choose." Gerald's feathers shifted. Hmmm. "I said I wouldn't choose." Gerald tapped the tablet. Harder. "There has to be another way." He tapped again. "I'm not feeding anyone to a monster just because a fucking rock says those are my options." Gerald screamed. The forest erupted. Hundreds of crows descended from the trees. Kitten grabbed my arm. Gerald flew directly in front of me. Angry. Not annoyed. Not impatient. Furious. And I finally snapped. "YOU MADE ME KING!" Silence. Even Gerald stopped. "Remember?" I pulled the crown from my head. "You brought me into this." I stepped closer. "You put this stupid fucking collection of shrubbery on my head." Kitten whispered: "Very regal." "Thank you." I pointed at Gerald. "So if I'm King, then I decide." Gerald stared at me. "And I will not sacrifice someone just because it's convenient." Hmmm. "I don't care if that's what the last King did." Gerald froze. There. Something changed. The crows began whispering amongst themselves. Tiny clicks. Soft caws. Restless wings. And suddenly I realized... The story wasn't about some ancient crow. Gerald lowered his head. "Gerald..." He wouldn't look at me. Oh. Oh fuck. I turned toward the tablet. The King fed the shadow. One bird every night. Again. And again. And again. Until winter ended. I looked at Gerald. "That was your King." Silence. "Wasn't it?" Gerald closed his eyes. Hmmm. Kitten whispered: "Oh fuck." For once...accurate and appropriately timed. Gerald hadn't brought me here to learn how to defeat the thing beneath my house. He brought me here because he'd seen this once before. His flock had survived it before. And Gerald knew the price. "How many?" Gerald didn't answer. "How many birds did he sacrifice?" Nothing. Then the forest moved. One crow descended. He was old. He was missing feathers. One of his eyes clouded white. Then another. Then another. They were scarred. They were broken. They grew silent. Survivors. There had been another war. Another King. Another shadow. And suddenly Gerald's obsession with loyalty made terrifying sense. Why betrayal had mattered. Why he tested Kitten. Why he watched me. Why he needed to know what kind of King I would become. I looked down at the crown that rested in my hands. Maybe that was the real fable. Not the crow. Not the candle. Not even the shadow. The King. Because anyone can promise to protect everyone when protection costs absolutely nothing. Leadership becomes something entirely different when the saving of many requires you to look at one innocent soul...and decide they are expendable. I placed the crown back upon my head. "I'm not your old King." Gerald stared at me. "Nobody is getting fed to anything." A long pause. Then— THUMP. The ground shook. Kitten sighed. "Of fucking course." THUMP. Birds exploded from the branches. Gerald remained. THUMP. Closer. The candle beneath the tree suddenly ignited. Nobody touched it. Its tiny flame flickered. And behind Gerald... his shadow stretched across the ground. Longer. Longer. Longer. Until it reached my feet. Except... Gerald hadn't moved. The flame hadn't moved. But the shadow had. Kitten whispered: "Bexxy." "I see it." The shadow crawled over my boots...up my legs...across my chest. Then stopped directly over the crown. Gerald screamed. The candle went out. Darkness. And from somewhere beneath the earth... something whispered. Not a hum. Not a knock. Spoken words.. "A King who will sacrifice nothing..." The voice came from every direction. The soil. The trees. My bones. Then it laughed. Low. Ancient. Starving. "…will eventually sacrifice everything." Silence. Kitten grabbed my hand. Gerald stared into the darkness. I swallowed. "Well." Kitten looked at me. "Don't." "I was just going to say—" "Whatever sarcastic bullshit you're about to say, DON'T." ... Fair. Gerald slowly turned towards me. And for once... I didn't need him to hum. I understood. The thing beneath my home had just challenged the King. And I had answered. Nobody was being sacrificed. Nobody was being offered. Nobody was becoming food for some ancient shadow because tradition said that's how this story ends. Maybe that made me brave. Maybe it made me foolish. More than likely both. But here's the thing about old fables: They're supposed to teach you how the story ends. Nobody ever said you had to obey them. I looked toward Gerald. Gerald looked toward me. I nodded. He nodded. Kitten sighed. "You two are about to do something entirely fucking stupid, aren't you?" Gerald: Hmmm. Me: "Abso-fucking-lutely." Kitten stared at us. "I want my walnut back." And somewhere deep beneath the earth... the shadow laughed again. Because apparently... Phase Two had begun.

📖 Before you replace books with screens, read this. A massive meta-analysis of 460,000+ readers reveals a striking gap between print and digital reading—and for young kids, the difference is night and day. Synthesizing data from across two decades, researchers at the University of Valencia published their findings in the *Review of Educational Research* to examine how leisure reading habits connect with comprehension. What they found complicates the rush to move childhood reading entirely onto screens. When these results are placed alongside previous research, traditional print-reading habits show a robust connection with comprehension, whereas digital leisure reading demonstrates an association that is numerically six to eight times weaker. This difference is statistical rather than literal. Previous research has placed the correlation between print-reading frequency and comprehension at roughly 0.30 to 0.40, while in the digital meta-analysis, the corresponding association was a modest 0.05. That does not mean a child understands eight times as much after opening a physical book; it means print-reading habits show a much stronger relationship with reading comprehension. That distinction matters, in large part because of what was being measured. The study defined digital leisure reading broadly, encompassing social feeds, chats, forums, websites, and online news alongside e-books. Reading an edited novel on a dedicated e-reader is fundamentally different from moving through messages, social posts, hyperlinks, and fragments of online information. Because the underlying research grouped these varied habits together, the findings speak most clearly to children’s overall digital reading environment rather than to every individual device or text. The pattern is most concerning among younger readers. Among primary- and middle-school students, frequent digital leisure reading was negatively associated with reading comprehension. It is only later, in high school and college, that the relationship edges into positive territory—possibly because older students are better equipped to navigate digital texts and regulate their attention. One compelling explanation for the divergence lies in the cognitive habits commonly associated with each medium. Reading comprehension grows through repeated exposure to extended arguments, unfamiliar vocabulary, complex syntax, and ideas that must be held in mind across paragraphs or chapters. Much everyday digital reading exercises a different skill entirely: rapidly deciding what deserves attention, extracting a small amount of information, and moving on. While useful in its own right, rapid filtering may not provide the same sustained practice that deep comprehension requires. The design of connected devices reinforces this scanning behavior. Much of what we read casually on screens is short, fast-paced, and written in informal fragments rather than sustained, syntactically rich prose. On a connected screen, text constantly competes with notifications, links, videos, incoming messages, and the knowledge that another source is always a tap away. This environment can encourage a fragmented reading mindset. Print, by contrast, places far fewer decisions between the reader and the text. Because most of the underlying studies were correlational and relied on self-reported habits, the results cannot prove that screens directly caused lower comprehension. Nevertheless, the scale of the dataset and the consistent age-based pattern make these findings difficult to dismiss. The practical lesson is not that children must avoid screens altogether. Phones and general-purpose tablets certainly have their conveniences, while dedicated e-readers can offer a quieter, less distracting experience. But digital reading as a whole should not be treated as a one-to-one substitute for print during childhood. Digital exposure cannot simply take the place of sustained, immersive reading practice. For parents and educators, the most constructive approach is intentionality. Print should remain the default medium for longer stories, complex nonfiction, and close reading during the foundational years, while digital reading can be introduced deliberately alongside lessons in managing screen distractions. For children navigating the critical leap from learning to read to reading to learn, physical books continue to provide the quiet, dependable foundation that deep comprehension demands.

Silvia’s mission is to help independent investors make money. As these investors grow their net worth, they are creating financial security for their families. We built @cfosilvia to help a tired father manage the family’s investment portfolio. Silvia helps a busy mother make decisions about the household finances. Silvia is used daily by young couples who are planning their future, including their first home, their unborn children, or the necessary support for their aging parents. Regardless of your financial situation, Silvia has become a favorite tool for independent-minded investors. These people don’t trust the legacy financial institutions. They are misfits who feel underserved by the old firms and out-of-place in the new world of speculation. They don’t want to be told what to do by a financial advisor. And they don’t feel comfortable listening to the mainstream media. The Silvia investor is self-made. They think for themselves. They make their own financial decisions. They live and die with the results. No nonsense. No victim mindset. Just misfits with a dream of a better life for themselves and their families. It is these people who inspire our team to work tirelessly every day. The long hours and high stress are worth every second when we talk to the people who are benefiting from Silvia. That is why we are launching Silvia 2.0. We felt investors on Silvia deserved a user interface and experience that matched the impact of the platform. This new design is more modern, sleek, and effective. There is less noise and fewer distractions. Today we are launching the same great product with a new look and feel. As part of the redesign, we wanted to create a new logo that matched the Silvia approach. We wanted to create a logo that signaled to misfits around the world that Silvia was a home for them. It had to be something memorable and different. The logo had to immediately evoke an emotional response that resonated with the right person and repelled the wrong person. That is why we ultimately chose a flag. A flag is a portable identity. Before radios, printed logos, or uniforms that could be read at a distance, a piece of cloth on a pole told friend from foe, member from outsider, and “here” from “everywhere else.” Flags are littered throughout history as important signals. They tell people where to go, who to trust, and when they are safe. Nations, regiments, teams, and companies have all discovered the same technology: a piece of cloth on a pole that turns “me” into “we.” That flag represents a home for the misfits. It stands for an optimistic future that allows an individual to defy the odds and create financial security for their loved ones. The flag lets you quickly identify similar-minded people when you are out in public. The Silvia flag is a simple logo that represents a lifetime of work. It unites every misfit under the same team. It creates an opportunity for us all to learn from one another. The flag faces forward. It stands for people who show up every day, do the work, and don't wait to be rescued. Those who carry the Silvia flag have a deep self-belief that they can change the trajectory of their family. They can use the tools available to them to learn. They can take intelligent risk in the market and allow compounding to lift their loved ones into a place of security for generations to come. Normal people don’t change the world. They don’t change their family’s situation. Only misfits do that. Misfits have written history for centuries. And our hope is this flag is the rallying cry for millions more to join our winning team. Silvia misfits vs the World. I like our chances. Read blog post here: $BRR

Italy’s birth rate is collapsing to historic lows. Italy recorded just 370,000 births in 2024, the lowest level since national records began in 1861. Fertility fell to around 1.13 children per woman in early 2025 data, with the decline continuing into 2025. Births fell further to roughly 355,000 in 2025 Fertility remained near 1.14 children per woman Delayed parenthood, fewer women of childbearing age, and economic pressures are driving the decline Net immigration is currently offsetting part of Italy’s natural population decrease Italy’s demographic crisis is becoming impossible to ignore.

Interesting work on long-horizon research agents. PrimeScientists can decide where a research agent spends its budget. Achieves 10.3% more reward with 50.6% fewer research attempts, under the same budget. That is PrimeScientist against AutoResearch on 12 AI research tasks. They treat deciding where to spend a research agent's budget as part of the agent's job. PrimeScientist keeps an executable plan tree of competing research directions and their outcomes. An adaptive MCTS policy reads the experimental feedback and the remaining budget and chooses whether to explore a new direction or continue a promising one. The gains also hold on systems, code optimization and ML engineering tasks. If your research agent can propose more experiments than you can afford to run, this is a concrete method for choosing among them. Paper: