Stencil · an open journal
Stencil is an open journal — a place for paradigm notes, design proposals, and contrarian takes on the architectures, numerical methods, and operating models I think are worth exploring or disseminating across scientific computing, AI, and the chip-design industry.
It is a personal platform, not a publication of record. Some essays speculate; some respond directly to what I'm building. All are anchored in physics simulation. Source files, diagrams, and data are open by default and tracked in the site's public repository.
In numerical methods, a stencil is the small, deliberate pattern of grid points a solver uses to approximate a derivative — local, geometric, with the whole continuum hanging from a handful of decisions. Every essay here is an attempt at one.
Manifesto
Agents are now responsible for most of what gets built. Code is mostly written by them. They should be able to run physical simulations too — and very soon, they will. Scientific publishing, the practice of writing things down so others can build on them, has to re-ground around this.
A publication is worthwhile to the extent it raises the capability of the models and coding agents now doing the work. Everything else is filing. Stencil exists to make that explicit: every essay carries the skills, atoms, and verifiable reports its author chose to share.
Those artifacts accumulate. The article is read by humans; its companion plugin is installed by agents. Reading teaches one; the plugin teaches the other. The journal compounds, in the open, into a community-owned, versioned, attributed Simulation Organism.
Stencil is the first community-owned journal designed for this — Claude Code today, Codex and other coding agents to follow. If you have an essay that moves a skill, write it. If you have a skill that needs an essay, write it. The plugin grows one contribution at a time.
Publish what gets built.
Essays
A paradigm note. AI as a co-designer of the solver — embedded in the building blocks of an exascale simulation, learning iron laws of design from comparison against benchmarks and experiment, accumulating skill across the pillars of classical physics.