Ab Initio Versions [ 2026 Release ]

Here’s a draft for an interesting post about ab initio versions — tailored for a computational chemistry, materials science, or ML/physics audience. Why “Ab Initio” Versions Still Matter in an AI-Driven World

The first implementation of a theory — no experimental fitting, no empirical parameters, just fundamental constants and equations. ab initio versions

So next time you run a fast, production-level calculation, thank the awkward, unoptimized, 1990s-era ab initio code that proved the physics first. Here’s a draft for an interesting post about

ML potentials are getting shockingly good. But they depend on training data — and that data comes from the expensive, “ab initio version” codes. When the ab initio version changes (e.g., higher accuracy functional, core-valence correlation), the ML model’s ceiling moves too. no empirical parameters