Property, strategic metals, factor funds and inflation-linked bonds, mapped against the conditions where each one actually does work — and where each quietly fails.
Where Else to Put Money: Property, Metals and Factor Funds
If GPT-class models answered every Google query, would we need a new grid? A first-principles calculation arrives at a smaller, weirder number than most press coverage.
We pulled the citation graph of every machine-learning paper indexed on Semantic Scholar between 2018 and 2025. The shape of influence is more uneven, more delayed, and less correlated with hype than…
The Investors Who Shaped Modern Markets
Investment Taxes: Keeping More of What You Earn
Building an Investment Portfolio as a Tech Professional
Open-Source AI vs Proprietary Models: Business Models and Developer Trade-offs
ARC-AGI-2 and the BIG-Bench-Hard 2026 refresh quietly shipped two months ago. The headline numbers are uninspiring; the per-category breakdown is genuinely informative about where the field is — and isn't — making…
Three years after RAG became the default architecture for grounding LLMs in private data, the systems shipping in production are quietly being unwound. A look at what's replacing them, and why most…
OpenAI, Anthropic and xAI together announced $51B of FY2026 cap-ex. Their combined annual revenue is approximately $12B. The ratio is unusual but not unprecedented — and the history of comparable spending booms…
A short letter from the editor on what kind of publication this is — what we cover, what we don't, what an article here is allowed to claim, and how to read…
Value Investing in an AI-Dominated Market
If GPT-class models answered every Google query, would we need a new grid? A first-principles calculation arrives at a smaller, weirder number than most press coverage.
We pulled the citation graph of every machine-learning paper indexed on Semantic Scholar between 2018 and 2025. The shape of influence is more uneven, more delayed, and less correlated with hype than the press cycle suggests.
ARC-AGI-2 and the BIG-Bench-Hard 2026 refresh quietly shipped two months ago. The headline numbers are uninspiring; the per-category breakdown is genuinely informative about where the field is — and isn't — making progress.
Three years after RAG became the default architecture for grounding LLMs in private data, the systems shipping in production are quietly being unwound. A look at what's replacing them, and why most of the original promises were architectural illusions.
OpenAI, Anthropic and xAI together announced $51B of FY2026 cap-ex. Their combined annual revenue is approximately $12B. The ratio is unusual but not unprecedented — and the history of comparable spending booms is worth understanding before pricing the equities.