I needed some sparse linear algebra code for a project (more on that later). Using AI (Claude Opus 5.5), it suggested I just convert qdldl to Fortran. I had never seen this library before, but it looked like a pretty straightforward C code. So, I had the AI develop a plan for converting it to Fortran and making it an FPM package.
The result is here: GitHub - jacobwilliams/qdldl-fortran: Modern Fortran edition of QDLDL · GitHub It’s pretty amazing. I haven’t touched the code at all. The AI ported it over to Fortran, and wrote unit tests to validate the results. I suggested it build the original C and compare the timings, and it did all that too. Note that the Fortran is a little bit faster in general. In less than an hour I was able to pull it in as a dependency (via FPM) into my other project and continue what I was doing, and it seems to be working great.