An update to Fred Espenak’s ΔT polynomials, fitted to recent Earth-rotation observations and simulations.
Optimising my GraalVM native image for size made it surprisingly slow. A comparison of all five optimisation levels, with the JVM setting the pace.
R plots can be considerably smaller without losing any detail. A look at lossless WebP, conversion in Quarto, and direct output with ragg.
What does Java’s Math.fma() become in a GraalVM native image? Finding the code with LLDB, then taking a look at auto-vectorisation.
Use AI to help you learn, but don’t outsource the learning itself. Some thoughts on code, knowledge, and becoming a thin wrapper around an LLM.
ZIP is everywhere, despite a vague specification and plenty of implementation quirks. For close to forty years, it’s been good enough.
