13 June 2026

Cold Comfort of Beautiful Game

The glass is a miracle of industrial engineering, currently sweating with the kind of condensation that suggests it has just emerged from a polar expedition. It is a heavy, dimpled pint mug—the sort that feels less like drinkware and more like a tactical commitment. Inside, a golden lager glows with the effervescence of a thousand tiny, trapped suns, topped by a head of foam as crisp and white as a fresh alpine snowfall.

You lift it. The cold bites at your palm, a stinging, pleasant shock that serves as the perfect sensory baseline for the next ninety minutes of absolute, unadulterated chaos.

You are in a pub that smells vaguely of sawdust, history, and the collective anxiety of three hundred people who have collectively decided that their mood for the next four years depends entirely on whether a man in neon-colored shorts can kick a sphere into a net. Outside, the 2026 World Cup is in full, throbbing swing. Inside, the collective atmospheric pressure is dropping faster than a lead balloon in a vacuum.

You take the first sip.

It is transformative. The liquid is crisp, biting, and aggressively carbonated. It hits the back of the throat with a sharp, hoppy slap that seems to wash away the memory of your own name, your job, and your pending tax returns. It is the liquid equivalent of a deep, satisfying sigh.

Then, the match happens.

A striker—a man whose hair has been styled with more precision than a Swiss watch—receives a long ball. The pub goes silent. The collective breathing of the room stops, suspended in a vacuum of high-stakes, sweating anticipation. You are mid-sip, the cold mug pressed against your lip, when the striker dances past a defender with the fluidity of an eel in a thunderstorm.

He winds up. Your eyes bulge. You forget the beer is hovering at a forty-five-degree angle.

He strikes the ball. It screams toward the goal, a leather comet. The keeper, a man who presumably was forged from granite and spite, launches himself horizontally through the air, his limbs flailing like a startled spider. The ball thuds against the post—a sound that resonates in your very molars—and ricochets violently back into play.

The pub erupts. It is a symphony of groans, shouts, and the frantic sloshing of beverages. You, meanwhile, have become a casualty of the excitement. Because you were frozen in the 'sip' position, the kinetic energy of the crowd's collective jump has resulted in a tiny, refreshing waterfall of lager cascading down your chin and onto your shirt.

You don’t care. You wipe it away with the back of your hand, heart hammering against your ribs. You look back at the mug. The foam has settled, the condensation is still doing its duty, and the beer is still, miraculously, ice-cold. You take another pull, emboldened, ready to witness whatever glorious, heart-stopping, alcohol-fueled disaster happens next. It’s glorious. It’s magnificent. It’s absolutely ridiculous.

12 June 2026

Gunpowder and Regret

The Dean of Admissions at his Ivy League university had a face like a dried apple and a handshake that felt like wet cardboard. He looked at Shawn’s file—an ROTC scholarship kid who spent his weekends doing pushups in the mud—and muttered something about "scholarly pursuits." If only he knew that Shawn’s primary extracurricular activity involved learning how to disassemble a suppressed rifle in the dark while reciting Shakespearean sonnets to keep from screaming.

Shawn wasn’t even American. He was a foreign exchange project, a tactical oddity tossed into the shark tank of elite special operations. He had a Medal of Honor and a Victoria Cross pinned to the inside of his footlocker—not because he wanted the shiny metal, but because they made excellent makeshift shims for a wobbly cot in a damp Afghan cave. Technically, he was a cross-pollination experiment—a SEAL on long-term attachment to the SAS, essentially a high-speed, low-drag piece of military lend-lease designed to ensure that the US and UK could export their particular brand of 'security’ with perfectly synchronized, multinational efficiency. He’d landed the ROTC scholarship through a bureaucratic loophole in the 'Joint-Operations Talent Exchange'—an administrative nightmare of a program that basically allowed the Pentagon and Whitehall to trade human assets like baseball cards.

Being an international hybrid SEAL-SAS operator meant Shawn was never quite sure which accent to adopt when interrogating someone. "Listen here, mate," he’d say, "you’re going to give us the coordinates, innit, or I’m going to have to do something remarkably unpleasant to your structural integrity, buddy." It confused the hell out of the locals. They didn't know if they were being liberated by the SAS or forcibly recruited into a frat house by a SEAL.

Shawn spent six years dodging things that go bang in the night. He’d been blown up in ways that defied physics, dangled from helicopters that sounded like dying lawnmowers, and eaten enough MREs to ensure his digestive system was now roughly 40% military-grade plastic. He had seen the absolute absurdity of geopolitical posturing—two empires arguing over a patch of dirt that mostly consisted of goats and bad vibes.

Then came The Day.

They were perched on a ridge, the air thin and smelling of stale gunpowder and regret. The CO—a man whose personality was a carefully calibrated mix of Pentagon bravado and Whitehall coldness, entirely composed of high-fives and sociopathy—pointed down at a village. Through the thermal scope, Shawn saw him: a four-year-old boy, wandering near a supply crate. He was carrying a wooden stick, pretending it was a sword, likely imagining he was fighting off a dragon.

"Take the shot," the CO hissed over the comms. "Threat neutralized."

Shawn stared at the kid. He was wearing a shirt with a cartoon cat on it. He wasn't a threat; he was a toddler who had clearly lost a war against his own shoelaces. 

"Sir," Shawn whispered, "he’s four. He’s currently losing a duel with a blade of grass."

"He’s a future insurgent," the CO barked. "Eliminate him."

That was the moment the hero fantasy—and the entire transatlantic architecture of 'global stability' folded like a cheap lawn chair. Shawn realized that if he pulled that trigger, he wouldn't just be an extension of the apparatus; he wouldn't be a soldier anymore, just a professional bully with a better tax bracket. He didn't want to explain to his future children that his greatest contribution to global security was murdering a kid who hadn't even mastered long division yet.

Shawn put the rifle down, stood up, and walked off the ridge. He retired right there, that night. His integrity, it turned out, was more important to him, and the only thing he hadn't lost in the field. And honestly? He preferred it that way. It was much quieter than a battlefield, and the only people he had to shoot now were the ones who put pineapple on pizza.

Years on, Shawn made sure that boy had a bright future. He tracked him down through a private network of NGOs and old-school contacts, funneling anonymous support to ensure he had access to education, clean water, and a life far away from the shadows of the operations. He’s currently studying engineering in a city where the only explosives he’ll ever encounter are in a chemistry lab. It’s the most important mission Shawn ever completed, and it didn't require a single bullet.

11 June 2026

Grand, Overstuffed, Border-Hopping Spectacle

The 2026 FIFA World Cup, where the football is world-class, the logistics are a logistical nightmare, and the geographic sprawl is so vast you might need a passport, a travel visa, and a sheer sense of wonder just to see two group-stage matches in the same week.

For the first time in history, the beautiful game has decided that three countries are better than one—mostly because, let’s be honest, hosting a 48-team, 104-match behemoth in a single nation is essentially an invitation for national infrastructure to spontaneously combust. So, we have the United States, Canada, and Mexico joining forces under the banner of "United As One." It is a tournament of superlatives: the largest, the most expensive, and undeniably the most complex circus act FIFA has ever performed.

The criticism, naturally, has been loud and well-deserved. FIFA’s decision to expand the tournament to 48 teams has turned the World Cup into a sprawling marathon. We are looking at 104 matches packed into 39 days. For the players, it’s a grueling physical test; for the fans, it’s an expensive geopolitical obstacle course.

The three-host model, while convenient for the budget, has birthed a travel headache that could test the patience of a saint. Teams are crisscrossing an entire continent, and the fan experience is increasingly dictated by dynamic ticket pricing that seems designed to leave the average supporter watching from a pub in their home country rather than the stands. Add in the complex web of visa requirements and regional travel, and you have a tournament that feels less like a global celebration and more like a high-stakes corporate summit for football fans.

Despite the bureaucratic chaos, the football remains the heart of the matter. With so many teams in the mix, we are seeing the rise of first-time debutants like Curaçao and the return of forgotten giants, all fighting for their moment in the sun.

If you are looking for the favorites, keep your eyes fixed on:

  • France: The perennial juggernaut. With a tactical mastermind in the dugout and Kylian Mbappé leading a squad of terrifyingly deep talent, Les Bleus are once again the team everyone else is trying to dethrone.

  • Brazil: Always the heartbeat of the tournament. The Seleção are looking to reclaim their aura of dominance. With their flair and constant ability to unearth new superstars, they remain the ultimate test for any defense.

  • Argentina: Defending champions and masters of the modern pressure cooker. They know how to grind out results, and in a tournament of this length, their tournament experience will be their greatest asset.

  • Germany: After years of a transition period, the German machine appears to be clicking into gear again, blending tactical discipline with a fresh, hungry generation of talent.

As the matches kick off, we’ll see if the spectacle justifies the stress. It’s the World Cup, after all—a tournament that somehow survives its own excess to deliver, time and time again, the magic that keeps the world watching.



World Cup 2026

CS-AI Research Papers

Papers with Code

Why Neo4j Sucks

In the race to implement AI-driven knowledge management, many enterprises are falling into a dangerous architectural trap: choosing Neo4j as the backbone for large-scale GraphRAG and agentic workflows. While Neo4j remains the safe procurement choice due to its market dominance, it is fundamentally ill-equipped for the demands of modern, high-concurrency AI systems. Designing a greenfield enterprise knowledge graph on Neo4j is a decision that essentially mandates future failure. The core issue lies in the database’s architectural DNA. Neo4j was designed for deep-path analytics on static datasets, not for the high-frequency, read-heavy and write-heavy cycles of an agentic RAG pipeline.

Agentic workflows depend on high-concurrency, iterative feedback loops. When you subject Neo4j to these demands, it hits a performance ceiling almost immediately. Its write-path is notoriously heavy; ensuring consistency across replicas for every agent-initiated update induces severe locking contention. As you scale to multiple agents, the database morphs into a system-wide bottleneck, strangling the parallelism necessary for effective reasoning. Furthermore, Neo4j’s reliance on memory-locality means that as data volume grows, the system demands excessive RAM. When the working set exceeds physical memory, performance collapses into disk-swap latency. In an agentic loop, where every millisecond of LLM thinking time is costly, a 500ms delay per graph hop due to cache misses is catastrophic. Agents become brittle, timeouts proliferate, and the system fails under even moderate load.

The problems are compounded by Neo4j’s lack of native vector integration. Because vector support is an add-on, engineers are forced to maintain a two-tier architecture, coordinating between a vector index and a graph store. This results in fragmented data, synchronization nightmares, and massive complexity in agent orchestration. Instead of a cohesive data fabric, teams are forced to build glue code to patch over these architectural gaps. Consequently, the entire programme team is handicapped from day zero. The Platform Team spends 90% of their time over-provisioning hardware and tuning Cypher queries just to stave off memory pressure, rather than delivering platform value. The Agentic Team is forced to artificially simplify the graph context—effectively lobotomizing the agent's intelligence—to stay within latency bounds. The Quality Team is left chasing phantom inconsistencies, struggling to maintain provenance in a system that lacks native, sharded, transactional integrity.

By binding a knowledge model to a tool incapable of true horizontal sharding, the architecture is effectively setting itself up for millions of dollars in re-platforming event. Within 18 to 24 months, as the graph grows and agentic traffic increases, the technical debt will become unsustainable. Cypher is an excellent query language, but it is not a system architecture. Choosing Neo4j today, when distributed-native MPP (Massively Parallel Processing) graph stores exist, is not just a technical oversight; it is an act of institutional negligence. True enterprise innovation requires choosing the right tool for the future, not the safest one from the past.

What is Left to Build When Software Is Free

What is Left to Build When Software Is Free

Strategic Angel Investing in AI Applications

Angel investing is far more than a financial transaction; it is a strategic commitment to the architects of the future. As we stand at the precipice of a shift toward Artificial General Intelligence, the role of the individual investor has become a critical mechanism for steering technological evolution. Becoming an angel investor in the AI space requires moving beyond traditional venture metrics and adopting a forensic approach to evaluating systemic viability. You must essentially become a judge of the startup's underlying philosophy, as the choices made during the early stages of model development will dictate the long-term ethical trajectory of the firm.

When scouting startups in the AI sector, the first sign of a company with true longevity is ethical alignment by design. You should actively avoid firms that treat AI merely as a superficial feature to automate cost-cutting or maximize short-term engagement. Instead, look for founders who are deeply grappling with the Alignment Problem. Ask yourself if their internal architecture prioritizes stability and balance. Companies that integrate foundational principles—those that seek equilibrium in their decision-making processes rather than just raw, unchecked output—are inherently more resilient. These startups prioritize the logic of their reasoning as much as the result, ensuring that as their models scale, they do not become unaligned or disproportionately destructive to their users or the broader ecosystem.

A superior AI startup should be able to explain the why behind its model's outputs with clarity. If a founder cannot articulate the specific guardrails they have implemented to ensure their AI remains proportional and balanced, they are likely building an opaque black box that carries significant liability. Furthermore, look for teams that have embedded self-reflecting mechanisms. The best AI startups treat errors not as failures to be swept under the rug, but as vital data points to refine their moral frameworks. This signifies a team that understands AI is not a static tool, but a continuously evolving intelligence that requires constant, iterative correction. The most valuable AI applications are those that augment human agency rather than liquidate it; prioritize startups that design technology to respect the boundaries of human autonomy.

To be an effective angel investor, you must adopt the mindset of an auditor. Dig deeply into their technical infrastructure. Ask hard questions about their data sources, the diversity of their training sets, and the specific protocols they use to prevent drift, which is the tendency for an AI to stray from its core ethical constraints as it processes more information. Understand the market, but prioritize the systemic health of the startup above all else. A company with a brilliant product but a dark or exploitative operational culture will eventually collapse under the weight of its own misalignment. Conversely, a startup grounded in principles of proportionality and universal balance possesses the structural DNA to survive and thrive in an increasingly automated world. By looking for these markers of balance and transparency, you are not just placing a bet on a balance sheet; you are contributing to the necessary architecture of a stable, ethical future.

Beyond your role as an auditor of systemic integrity, you should prioritize partnerships with founders who operate with a high degree of independent competence. The most promising AI ventures are not those that require excessive mentorship or hand-holding, but those led by visionaries who already possess a profound grasp of their technical domain. Your value as an investor here is not to provide remedial guidance, but to act as a bridge between their high-level vision and the reality of the market. By fostering a direct connection between their advanced ideas and the specific, often overlooked needs of the end-user, you enable a form of customer-centricity that bridges the gap between abstract innovation and practical implementation. This involves identifying active market voids—the white space where a product solution is desperately needed but currently ignored by incumbents—and facilitating the engagement necessary to bring that solution to scale. In this dynamic, you are not a manager, but a strategic conduit who empowers capable founders to turn their equilibrium-seeking models into dominant, human-centric industry standards whether that be solving hard problems in deep  tech or pioneering new paradigms in practical, high-utility application layers.