Weekly 3x3: CapEx Boom. Agentic Checkouts. Quantum Optimization in Portfolio Allocation
Big Tech’s $725B infrastructure bet is crushing free cash flow as "Agentic Commerce" protocols move from pilot to production.
Latest
Weekly 3x3: Hyperscalers bet $600B on AI. MCP becomes infrastructure. Context engineering replaces prompt engineering.
Trillion-dollar infrastructure bets are under pressure as firms learn that scaling spend doesn’t scale performance without better ways to manage what models actually see.
Weekly 3x3: Labour market jitters. Altman on spending commits. AI benchmarks under scrutiny.
A volatile week across markets and AI. Tech valuations wobble, central banks stay cautious, Big Tech defends trillion-dollar AI bets, and new research questions how we measure progress. Signals are getting noisier — not clearer.
The economics of intelligence: AI and the efficient frontier
The contemporary narrative around AI is dominated by exponential capabilities: the emergence of reasoning, the human-level performance on standardized tests, the spectacular hallucinations. This focus on performance has obscured a far more fundamental and immediate constraint: the economics of production. We have industrialized a new class of cognitive labor, transforming
Weekly 3x3: Global stocks surge. AI spending gets practical. Teens lean on chatbots.
Markets rally on US–China trade optimism as stocks hit new highs, but risks linger beneath the surface. AI spending shifts from hype to ROI, while new research flags both accelerating scientific automation and growing concerns over concentration, stability, and control.
ASICS on the track and ASICs in the data center: Specialization and the pursuit of record performance
I was in Tokyo this week for the STAC Summit. By coincidence, the World Athletics Championships were happening at the same time. One evening I was at the Japan National Stadium watching sprinters and distance runners push human limits. The next morning I was at our Summit, where firms shared
Fast and curious: Tsinghua researchers beat Dijkstra’s legendary shortest path algorithm, opening the door to faster routing everywhere
This week, a team led by Dan Qiao from Tsinghua University achieved something that seemed impossible: they broke through a computational barrier that had stood for four decades. Their breakthrough was a new shortest path algorithm that runs faster than the legendary Dijkstra's algorithm—the gold standard that&
The global AI race: Chips, capital, and kilowatts
The AI arms race is shaping up like a three-way tug-of-war — and it’s not yet clear who’s going to fall in the mud first. The US, the Middle East, and China each bring something essential to the table. But none of them hold all the cards. That’s
The great unbundling of data: Open formats and composable platforms
The data warehouse is no longer the centre of gravity. It’s just one node in a growing constellation. Modern data strategies aren’t about centralising everything in one place. They’re about composability—the ability to mix and match components, adapt to change, and build around the edges. At
Fairness in the fast lane: As trading speeds hit new highs, is it possible to keep the playing field level?
In financial markets, fairness is easy to believe in but hard to define — especially when you're dealing with electronic trading systems that operate faster than the blink of an eye. For most of history, fairness in trading meant access: If you could make it to the floor of
Decoding the data lakehouse: Why apache parquet and iceberg matter
The data lakehouse is no longer an experiment. It’s fast becoming the blueprint for enterprise data architecture. At the center of this shift are two open technologies: Apache Parquet and Apache Iceberg. Understanding their role today — and where they’re headed — is critical for sizing the opportunity. Parquet: The
Opening the black box: Explaining explainability in AI
I've often found explainability in AI to be an elusive concept—essential, yet difficult to pin down. It underpins trust, adoption, and regulatory compliance, but how does it actually work? A recent paper, Explainable Artificial Intelligence (XAI): From Inherent Explainability to Large Language Models by Fuseini Mumuni and
Quantum
Quantum quests: Marching towards quantum, qubit by qubit
The world of quantum computing is experiencing a surge of momentum, with recent announcements from tech giants Amazon and Microsoft adding fuel to the already rapid pace of innovation. These developments, alongside ongoing research breakthroughs, signal a potential transformation in how we approach computation. One of the most significant trends
Crypto
Reserves and stockpiles: US cultivating crypto confidence
President Trump's recent announcement of a "Crypto Strategic Reserve" has set the cryptocurrency world alight. This initiative outlines plans for the U.S. government to actively engage in the purchase and sale of cryptocurrencies, aiming to establish the nation as a global leader in the digital
AI
LLMs get their dance on: Bytedance open-sources Kubernetes-based LLM inferencing platform
ByteDance's release of AIBrix appears to be a significant advancement in the practical application of Large Language Models (LLMs), particularly within enterprise environments. This Kubernetes-based serving stack, specifically designed to augment vLLM, leverages ByteDance's extensive experience deploying LLMs at scale. The project addresses the critical transition
AI
The great AI knowledge transfer: Apple researchers quantify optimal conditions for teacher-student model distillation
Recent discourse within the AI community has centered on model distillation, notably fueled by speculation surrounding DeepSeek's R1 model. Distillation, in essence, is a technique whereby the outputs of a large, high-performing model are utilized to train a smaller, more efficient model. This process, exemplified by the rumored
Weekly 3x3
Weekly 3x3: CME to launch Solana futures. Texas needs reactors. ChatGPT beats DeepSeek on stock predictions.
From Solana futures and fragile payments infrastructure to negative power prices and soaring data-center energy demand, markets are testing new edges. AI research pushes into finance and long-context reliability as real-world constraints bite.
AI
AI memory reboot: Titans are the new transformers
Think about how we learn. We absorb facts, connect them, remember key details, and build a bigger picture over time. Current AI struggles with this. LLMs are good at processing information in front of them, but they have a hard time remembering things from earlier in a long text or
AI
Back to the future in time series forecasting: Sundial points to a new dawn of foundation models
A recent paper from China introduces Sundial, a family of time series models that appears to represent a significant leap forward in time series forecasting. To understand Sundial's potential impact, it's helpful to trace the evolution of this field, which has deep roots stretching back centuries.
Weekly 3x3
Weekly 3x3: Chips act is dead. Vance on AI opportunity. Hardware-aligned models.
Stagflation risk is back on the agenda as chip policy wobbles and pressure builds to reshore manufacturing. In AI, leaders warn against overregulation, hyperscalers move into chip design, and research pushes long-context efficiency and reasoning under constraints.
AI
Train smarter, not harder: Clever combinations of existing techniques are driving the next wave of AI optimization
The buzz around DeepSeek is undeniable, and for good reason. But the really extraordinary thing isn't some magical new algorithm. It's their clever application and combination of existing techniques. DeepSeek has shown us that there’s still plenty of optimization out there to discover. Their approach
AI
Total recall, AI edition: How caching enhances LLM interactions
The field of Large Language Model (LLM) application development is a hotbed of innovation, with researchers and engineers constantly developing new software engineering practices and architectural patterns to optimize performance. Retrieval-Augmented Generation (RAG) was a significant early advancement, enabling LLMs to access and integrate external knowledge. However, for long-context LLMs,
Blog
Thinking through paradoxes: Less is more
If your LinkedIn feed is anything like mine, you’ll have read a lot about Jevons Paradox over the last few days. This paradox suggests that increasing the efficiency of resource use can sometimes lead to an increase in overall consumption. In the context of AI, this means that even
AI
Frugal computing FTW: Resourceful engineering in a hardware-constrained world
AI is evolving rapidly, with new models and engineering techniques emerging thick and fast. Recently, a Chinese company called DeepSeek has been making waves, not just for the impressive results they've achieved, but for how they've achieved them. Their approach offers a valuable lesson in innovation,