Legibility Kills What It Measures
What happens when AI makes everything legible? The things that worked because they couldn't be seen clearly may stop working once they can.
I grew up in Argentina, where you learn early that money breaks, institutions fail, and the systems people depend on are more fragile than anyone admits. That shaped everything I build now.
Through LambdaClass I build Ethrex, one of the fastest Ethereum clients, a new programming language called Concrete designed around a formally verified kernel, and Lambdaworks, a cryptographic proof library used in production. I built the underwriting engine for Levenue, Europe's largest revenue-based financing platform. I'm helping develop a bank and a payments platform replacing broken financial rails in Latin America. Through Ergodic Group I hold companies across distributed systems, AI, gaming, wine, and culture. This site is where I write about what compounds over time rather than what trends today.
Every previous coordination technology that dissolved a form of selfhood also cultivated the next one. Current algorithmic systems only dissolve. The problem is the speed mismatch.
What happens when AI makes everything legible? The things that worked because they couldn't be seen clearly may stop working once they can.
A 400-year journey through logarithms, Kelly, ergodicity, and tail risk to show why geometry sits at the core of finance and why Jensen's inequality ties the wh…
We tested Spitznagel's tail hedging strategy and AQR's critique with 17 years of real SPY options data. AQR's published tests use near-ATM puts in a no-leverage…
We built fatcrash, a Rust+Python toolkit with 15 crash detection methods: LPPLS, DFA, EVT, Hill, Kappa, Hurst, GSADF, momentum/reversal, price velocity, and mor…
We assembled forex-centuries, an open dataset of exchange rates, gold, silver, interest rates, commodity prices, GDP, sovereign debt, and more — 27 sources span…
The scarcity that matters most in a post AGI world won't be compute or energy. It will be desire itself.
What happens to the West's services advantage when strong AI models are free, portable, and running on every laptop?
Our stack and practices for building SaaS applications: Elixir on the backend, React on the frontend, Nix for everything else. No Docker. No Kubernetes.
Humanity is completely unprepared for what's coming. The pace of AI advancement might give people months to adapt, not decades.
A practical guide through the landscape of type systems, from everyday generics to dependent types that prove correctness, with examples in Rust, Scala, and Idr…
Concrete is a systems language proposal built for machine reasoning, aiming to combine low-level performance with formal verification, safety, and expressive de…
Individuality is a coordination technology. It emerged under specific historical conditions, and those conditions are weakening as computation, capital, and aut…
A framework for building institutions that compound under pressure: treat time as information, preserve loop integrity, and focus on domains where repetition im…
Ethereum is emerging as a neutral financial backend, lowering the cost of global financial services by encoding ownership and obligations in shared infrastructu…
The internet scaled information but not ownership institutions; Ethereum addresses this gap by embedding rights, records, and enforcement into programmable econ…
Discussion about Ethereum's native rollup roadmap
Crypto found product-market fit where trust is weakest: inflationary or censored economies, and internet-native communities that need programmable coordination …
The proving-verification asymmetry in zero-knowledge proofs is what makes them economically meaningful: proving is expensive, verification is cheap, and that ga…