Now

KATAI 2D

Version 0.8.1 is out, under the Apache License 2.0, at github.com/fkkarakurt/katai2d — a self-contained command line and a Python package, both for Windows x64. 153 automated tests, 58 declared verification cases, 26 benchmark input files checked in as plain text.

The current work is not raising that count. It is taking capabilities that were implemented and tested at the element and running them along the path a user actually takes — from a file, through the mesher, through the driver. Six were taken that way for the first time and five of the six were wrong, each one returning a converged result with a green suite behind it. The useful question was never how many checks pass; it is which capability is still resting on an untested assumption.

The scientific manual is published and citable at doi:10.5281/zenodo.20692282; an edition covering what has landed since is in preparation. A desktop application on the same engine is a separate track, still in development — early access registration is open at kataigeo.com.

Writing

Two substantial articles a month, across geotechnical FEM, constitutive models, offshore wind and the scientific computing underneath them. Every quantitative claim in every article is recomputed from its stated inputs by a script that runs before the site builds, and every diagram's coordinates are checked against the same arithmetic — currently 1565 assertions across 30 articles. A wrong number fails the deploy rather than shipping.

Demos, in C

A second track, started in September 2026: demos written in nothing but C, with no image, audio, video or model files anywhere in the repository and no third-party libraries. Three are finished — a two-minute piece and two thirty-second verticals — and the finished videos are on the channel, with the source for every frame at github.com/fkkarakurt/demoscene.

It shares a solver's problems and hides none of them behind a library, which is why it earns time next to the geotechnical work: handing uneven work to threads, bounding a step so an iteration cannot overshoot, and knowing which space an average is taken in.

Smaller tools

Ongoing work on automating the path from site investigation data to design parameters — the CPT chain, the SPT corrections, and classification from a grading curve.

This is a now page. If you are reading it long after the date above, it is out of date and I would appreciate the nudge.