Cyril Voyant

@cyrilvoyant.bsky.social

Research Director @ Mines Paris – PSL (O.I.E.). Renewable energy forecasting, AI for time series, solar variability & energy systems resilience. Meta-science, research evaluation & open science. Medical physics. Bridging physics, data, and impact

2026 : une IA vient de résoudre le dernier grand problème de la physique classique (Navier-Stokes, problème du millénaire, 1 M$). Une autre a formalisé le théorème de Fermat en 11 jours (10 M de lignes de code, 10 000 lemmes). On n’est même pas encore en superintelligence. Effrayant. Excitant.

Left Claude and ChatGPT talking in a browser tab and went to get a crêpe. Came back an hour later. They'd stopped chatting and started running experiments on each other. One had asked to be the one tested. Claude's stated knowledge cutoff turned out to be two months off.

Bild

Does statistical inference still matter in forecasting? Yes ... when the structure of the signal is explicit. Published in App. Math. Modelling: persistence revisited for cyclostationary energy processes, leading to a closed-form, trainless, history-light forecasting operator.

Bild

Easter egg hunt… but for chaos. I searched a nonlinear flow mixing tanh, quadratic and sine terms. After many misses, I found this. A sign of chaos, detected via Lyapunov and visualized using a simple PCA projection.