Now

Almost everything on this page exists because I wanted to understand something and building the thing turned out to be the only honest way to find it out. None of it is my job, which is exactly why it works. There is no deadline and no roadmap to answer to, so I get to follow whatever question I am chasing all the way down and ship the result while it is still rough, without anyone deciding whether a half-finished thing counts as a release. Most of what I actually know came out of those messy versions, the ones I would never have been allowed to publish anywhere that was keeping score, and the through line is that I read the spec properly and try to leave behind something small and correct that will still run in twenty years.

Right now that looks like:

• TTAS, an open assembler for Tenstorrent's Tensix engine, because assembling a single instruction should not require their entire compiler tree.
• Ernest, compiling Fortran down onto real quantum hardware to find out exactly where the quantum stops and the classical begins.
• BarraCUDA, a CUDA compiler with no vendor lock-in, tested by simulating nuclear reactors.
• z390, the portable mainframe emulator, where I get to build alongside the people who first wrote it.

And because it amused me, here is that same set of things laid out as an agile board, ceremony and all.