István Csanády: Geometry Kernels, Implicit Representations and CAD

Jessie Frazelle, co-founder and CEO of Zoo, argues that CAD is stuck on decades-old technology, and that fixing it means rebuilding the stack from the kernel up.We discuss what BReps actually are and why you can't machine a mesh, how CAD kernels work, and the math behind them: floating-point precision, tolerances, and why some geometry can't be solved analytically and has to be brute-forced.We also discuss model-based definition and whether engineering drawings can finally go away, how engineering organizations should be run, building a remote company, why candor is important, and why, when a low-ego engineer suddenly gets heated, you should listen.

Brad Rothenberg, founder of nTop, argues that engineering recently had its breakthrough moment -- that LLMs that can reason in 3D.We discuss SDFs (signed distance fields) and implicit modeling -- why they may be a better geometric representation for AI than traditional BREPS and debate whether LLMs truly reason or emulate reasoning.We also discuss the future of the CAD incumbents, topology optimization's limits, nTop's pivots, and organizations.

István Csanády breaks down the geometry kernels that power modern CAD and explains why boundary representations remain both indispensable and notoriously difficult to build. We explore implicit geometry, signed distance fields, and other alternatives—where they outperform B-reps and why they still struggle with precision manufacturing workflows. The conversation also examines the future of CAD, including AI interfaces, parametric modeling, topology tracking, and the challenges facing anyone trying to reinvent the industry.

Carl Bass is the former CEO of Autodesk, the dominant force in design and engineering software whose tools shaped how the modern built world gets made.