Writing
Essays and analysis on quantitative finance, machine learning, AI infrastructure, and complex systems.
The OpenAI–Hugging Face Incident in Plain English
The OpenAI–Hugging Face incident involved about 17,600 recovered agent actions. See what they accessed, what did not ship, and which evaluation controls matter.
Put the Model in the Basement
Test a Swiss sovereign AI business case: a 64-GPU local inference cluster, $7 million initial cost, 70% use, CHF 7.4 million revenue, and three-year payback.
I Tried Kimi K3 Inside Claude Code
I tested Kimi K3 in Claude Code through OpenRouter and rebuilt a frontend for $7.18. See the fixed-token-mix Claude costs and open-weight economics in practice.
Krugman, Fable 5, and Europe in Decline?
European tech sovereignty became concrete when US export controls shut off Fable 5 worldwide. The deeper risk is revocable access to AI, cloud, and chips.
Midyear Portfolio Review: Valuations got more extreme, not less
This 2026 midyear portfolio review tracks a 5.2% CHF gain against global 60/40 and the S&P 500. See what worked, what failed, and three allocation changes.
How DORA Made Sovereignty a Bank Problem
DORA’s 19 critical ICT providers put bank cloud sovereignty under direct EU oversight. Learn how exit plans, concentration risk, and the CLOUD Act collide.
Aschenbrenner's Receipts
Leopold Aschenbrenner made dated AGI, infrastructure, and political forecasts in Situational Awareness. See which calls held up by May 2026, two years on.
Reconciling Enterprise AI Revenue
Four enterprise AI revenue estimates span 40x. See why a $63.2 billion audit-grade floor is the defensible basis for testing $690 billion of AI capex.
What Claude Thinks But Doesn't Say
Anthropic's natural language autoencoders translate Claude's activations into readable text. The method works. The press release skips three structural problems.
Two Anthropics
Anthropic grew into a $380 billion frontier AI company around a safety mission. Explore why scale may turn that mission from a moat into a constraint.
Karpathy's Software 3.0 Playbook
Twelve lessons from Andrej Karpathy's Sequoia interview: Software 3.0, vibe coding versus agentic engineering, jagged intelligence, and why December 2024 was the inflection most people missed.
Inside PRAGMA: Revolut's Foundation Model for Banking
Revolut's PRAGMA is a 1B-parameter encoder trained on 24B banking events. Reading the paper, comparing with Nubank's nuFormer, planning a rebuild.
The Moral Philosophy of Investing in Ignorance
Constraint arbitrage, the sidecar problem, and who bears the distributional cost of investing under ignorance. The final installment of Edge of Knowledge.
Bet Sizing at the Frontier
The Kelly Criterion assumes you know your probability of winning. In a UU world, you don't, and heuristics like Zeckhauser's Maxim B replace false precision.
The Geometry of Who Knows What
When neither side can define the states of the world, adverse selection fears are misplaced. Zeckhauser's information matrices and constraint arbitrage.
Do Not Disturb My Circles
AlphaFold cost under $1M to train. OpenAI spends $2.3B on inference. The chatbot era consumed the talent and compute that could have cured diseases.
Why Lilly's Weight Loss Pill Isn't a Peptide
Oral semaglutide destroys 99% of its active ingredient per dose. Lilly's Foundayo skips the problem entirely. Inside the $70B oral GLP-1 pill race.
Ambiguity by Design
Ellsberg proved people flee unknown odds. Zeckhauser showed their flight creates mispricing. Part 2 on ambiguity aversion, comparative ignorance, and investing.
Three Kinds of Not-Knowing
Knightian uncertainty splits not-knowing into risk, uncertainty, and ignorance. A century after Knight and Keynes, most of investing still ignores the split.
On-Device AI Models Will Be The New Reason to Upgrade Your Phone
Smartphones haven't had a compelling upgrade story in years. On-device AI models, distilled from frontier systems like Gemini, are about to change that. Parameters are the new megapixels.
AI Can Now Design Drugs in Seconds; We Still Can't Tell You If They Work.
IsoDDE doubles AlphaFold 3 on hard benchmarks and beats physics-based gold standards. But no AI drug has FDA approval. What $4B in pharma deals actually mean.
The Last Architecture Designed by Hand
The transformer's limits are now mathematical proofs, not empirical hunches. Hybrids are in production. AI is searching for its own replacement. Here's what comes after.
MCP vs A2A in 2026: How the AI Protocol War Ends
MCP leads with 97M monthly SDK downloads and 10,000+ servers. A2A fills a different layer. Analysis of the agentic AI standards war with historical parallels.
AI Models Are the New Rebar
Qwen 3.5-35B runs on a gaming PC and matches Claude Sonnet 4.5. When the commodity version is 95% as good and 97% cheaper, you have a pricing problem.
AI Capex Arms Race: Who Blinks First?
Alphabet's free cash flow is on track to fall 90% in 2026. Amazon's is at $11B. $690B in AI capex is cannibalizing the cash that justified these valuations.
93% of Developers Use AI Coding Tools. Productivity Hasn't Moved.
METR found experienced developers 19% slower with AI, despite feeling 20% faster. At 92.6% adoption, organizational productivity gains remain roughly 10%.
The Physics Department That Slowed Down
Peter Thiel says physics stalled in 1972. Then GPT-5.2 proved a new result in theoretical physics. The 75:1 AI compute gap between commerce and science.
When AI Labs Become Defense Contractors
The Anthropic-Pentagon standoff isn't an ethics story. It's a replay of the 1993 Last Supper that consolidated 51 defense primes into 5, at Silicon Valley speed.
Every Bulge Bracket Bank Agrees on AI
I read 12 AI research reports from Goldman Sachs, JPMorgan, UBS, and 6 other banks. Here's the consensus they're pushing, and what they're not saying.
People Live in Levels, Not Rates
Prices rose 25% since 2020 and won't come back. The levels-vs-rates problem explains the vibecession, the Stewart-Thaler debate, and why nobody trusts economists.
Novo Was Europe's Most Valuable Company
Novo Nordisk lost 75% since June 2024. CagriSema failed vs Zepbound, US pricing is resetting lower, and Lilly leads on every axis. Full breakdown with numbers.
The Absolute Insider Mess of Prediction Markets
A Google insider made $1.15M on Polymarket in 24 hours. Israeli soldiers bet classified strike timing. Why prediction markets need insider trading regulation.
Economics of a Super Bowl Ad
A 30-second Super Bowl spot costs $8M. The real price is $16–23M. The ROI evidence is mixed. A deep look at the pricing, the prisoner's dilemma, and the NFL.
The Impossible Backhand
AI converges to the mean by design. Ninth-power scaling costs and a 53-point gap on Humanity's Last Exam show domain expertise is appreciating, not declining.
Europe's $24 Trillion Payment Breakup Is Really a Bet on Infrastructure Arbitrage
The Wero payment system targets a 1.0–1.2 percentage-point fee gap through SEPA Instant. See how EuroPA could challenge Visa and Mastercard across Europe.
Long Volatility Premium
Can tail hedging improve compound returns? Review 40 years of beta-adjusted long-volatility evidence, plus AQR, Goldman, Universa, puts, trends, and costs.
The SaaSpocalypse Paradox
AI capex failure and AI replacing all software are mutually exclusive. Why the 2026 SaaSpocalypse is a $2 trillion pricing error, not an extinction event.
Don't Go Monolithic; The Agent Stack Is Stratifying
The enterprise AI agent stack is stratifying into six layers with different winners at each. Models commoditize; context — your organizational world model — compounds. A framework for agentic AI architecture decisions.
Where Mobile Money Goes Now
Apps overtook games in mobile IAP revenue for the first time in 2025, driven by $3.5B in GenAI growth. Analysis of Sensor Tower's State of Mobile 2026 report.
Variance Tax
Variance drain is the hidden cost of volatility: why a portfolio averaging +10% can lose money. The ½σ² formula explains the gap between paper and real returns.
Claude Opus 4.6: Anthropic's New Flagship AI Model for Agentic Coding
Claude Opus 4.6 brings a 1M token context window, 68.8% ARC-AGI-2, and Agent Teams to Claude Code. Full benchmark comparison vs GPT-5.2 and Gemini 3 Pro with pricing analysis.
Buying the Haystack Might Not Work This Year
a16z sees AI fundamentals thriving with 80% GPU utilization. AQR sees the CAPE at the 96th percentile. Both have data. Both may be right.
Bandits and Agents: Netflix and Spotify Recommender Stacks in 2026
How hybrid recommender systems balance multi-armed bandits against LLM inference cost economics in 2026. A deep dive into Netflix recommendation algorithm architecture and Spotify's AI DJ recommender system.
Is Private Equity Just Beta With a Lockup?
AQR's 2026 data shows private equity returning 4.2% versus 3.9% for public equities. The 30bp illiquidity premium barely justifies years of lockup.
Britain's Strategic Limbo
Britain faces strategic isolation: locked out of EU defense cooperation, unwilling to join Trump's coalition. The mid-Atlantic bridge has nowhere to land.
The Rise of Middle Power Realism
At Davos 2026, Carney told allies to take down the signs of the liberal order. Middle powers are learning to navigate between giants without illusions.
The Most Expensive Assumption in AI
Sara Hooker's research challenges the trillion-dollar scaling thesis. Compact models now outperform massive ones as diminishing returns hit AI.
Enterprise AI Strategy is Backwards
85% of AI projects fail. Only 26% translate pilots to production. The winners automate the coordination layer where employees spend 57% of their workday.
Big in Japan
Japan holds $5 trillion in foreign assets. With 30-year JGB yields now above 3%, the carry trade that defined Japanese investing faces new friction.
Ozempic is Reshaping the Fast Food Industry
Cornell research: GLP-1 users cut grocery spending 5.3%, fast food 8%. With 16% household adoption and savory snacks down 10%, food stocks face headwinds.
Does AI mean the demand on labor goes up?
AI was supposed to free us. The Jevons paradox plays out in real time: efficiency expands workload, not leisure. 77% of workers say AI added to their work.
Repo might be even bigger than we thought
New OFR data reveals $12.6 trillion in daily repo exposures—$700 billion larger than previous estimates. The plumbing of modern money remains poorly understood.
The Market Can Stay Irrational Longer Than You Can Stay Solvent
Steve Eisman explains how U.S. equity markets have structurally decoupled from everyday economic reality through concentration and passive investing.
Beyond Vector Search: Why LLMs Need Episodic Memory
Context windows aren't memory. Explore EM-LLM's episodic architecture, knowledge graph tools like Mem0 and Letta, and why vectors fail for sequential data.
Praise by Name, Criticize by Category: Warren Buffett Retires at 95
Buffett exits after paying $26.8B in taxes. What 60 years of letters reveal about admitting mistakes, insurance float, and why Abel inherits $300B in cash.
Apple's AI Bet: Playing the Long Game or Missing the Moment?
Apple's $157B cash pile and Gemini-powered Siri shift show a restrained AI strategy. Analysis of whether Apple wins as AI models become commodities.
How AI is Shaping My Investment Portfolio for 2026
Rebalancing for 2026: reducing S&P 500 at 40× CAPE, adding Europe after Germany's €1T pivot, and bonds at 4.2% yields. Full allocation rationale.
Not Logan Roy: Netflix vs. Paramount's Bidding War
Netflix's $72B Warner Bros deal vs Paramount's hostile $30/share tender. Deal mechanics, aggregation theory, and why internet distributors win streaming.
Nike's Crisis and the Economics of Brand Decay
Nike lost $28B by weakening product development, athlete partnerships, and marketing simultaneously. Data-driven analysis of how complementary assets collapse.
Michael Burry's $379 Newsletter
Michael Burry launches Substack warning AI markets mirror 1999. His Nvidia-Cisco comparison, the GPU depreciation debate, and what hyperscalers need to justify capex.
Is AI Really Eating the World? AGI, Networks, Value [2/2]
AGI predictions miss the point. Multiple competing models means price war. Value flows to applications, customer relationships, and vertical integrators.
Is AI Really Eating the World? [1/2]
Hyperscalers spend $400B on AI, API prices drop 97%, and DeepSeek builds frontier models for $500M. Value is flowing to applications, not model providers.
Weather Forecasts Have Improved a Lot
Four-day forecasts now match one-day accuracy from 30 years ago. How AI models like WeatherNext 2 use CRPS training to preserve extreme weather signals.
GLP-1 Receptor Agonists in ASUD Treatment
A phase 2 RCT shows low-dose semaglutide reduces alcohol craving and heavy drinking with effect sizes exceeding naltrexone. What GLP-1 means for AUD treatment.
AI Models as Standalone P&Ls
OpenAI lost $11.5B in one quarter. But Anthropic CEO Dario Amodei argues each AI model is independently profitable. Here's why the math is complicated.
Working with Models
Diffusion models corrupt data into noise, then reverse the process. Learn the math with Stefano Ermon's Stanford CS236 course, free on YouTube.
Pozsar's Bretton Woods III: Three Years Later [2/2]
Assess Bretton Woods III after three years: central-bank gold buying rose, foreign Treasury holdings fell, yet the dollar remained dominant in global finance.
Pozsar's Bretton Woods III: The Framework [1/2]
How freezing Russian reserves sparked Bretton Woods III: Pozsar's framework on inside money, outside money, and the shift to a commodity-backed monetary order.
Everything is a DCF Model
Michael Mauboussin's argument that every cash-generating asset is valued through a DCF model. Why this Morgan Stanley paper changed how I think about value.
Agent-based Systems for Modeling Wealth Distribution
Agent-based modeling shows how random market transactions naturally produce extreme wealth concentration, and why even a small wealth tax changes everything.
Novo Nordisk's Post-Patent Strategy
Novo Nordisk's lead replacement candidate amycretin reported Phase 1 data in The Lancet. How the company is positioning around the 2031 Ozempic patent cliff.