I work in seasons.
There are seasons of writing, when the ideas arrive faster than I can organize them. Seasons of thinking, when the work mostly looks like walking, reading, and staring at a whiteboard. Seasons of creating, when the output has less to do with software and more to do with music, design, or making something exist because it should.
And then there are seasons of coding.
This is one of those.
The easiest way to see it is the contribution graph. For years it rises at a familiar angle. In 2026 it bends upward and starts to look nearly vertical. When I checked on July 17, Commit History placed me at #1,769 globally for private contributions, with 12,640 recorded. The rank will move. The shape is the more interesting part.
The writing has been quieter because the building has become louder.
Most of that energy is concentrated in three systems. Punk is the runtime: it learns from repeated agent work, proves cheaper execution paths, and keeps the live model available when the proof is not there. Pantasso is the context layer: it turns approved company sources into cited answers, canonical records, honest gaps, and scoped access. Plasmate is the web layer: it gives agents structure and meaning without making them reverse-engineer a page designed for eyes.
They are not three unrelated products. They are three answers to the same operational question.
What has to exist beneath an agent before you can trust it to do real work?
It needs a runtime that can learn without gambling on production. It needs company context with provenance and control. It needs a web representation it can understand without spending most of its tokens on presentation noise.
Runtime. Context. Web. The thesis is becoming an operating stack.Punk and Pantasso alone have absorbed more than 1,200 repository changes in the past month. That number is not a customer outcome, a benchmark result, or proof that the market cares. Commits are a poor substitute for impact. But they are a useful signal of where attention is going, and right now the signal is unambiguous.
The work is in implementation, integration, testing, hardening, and the thousand small decisions that separate an argument from a system someone can depend on.
This is also why a personal site should not be confused with a publication schedule. The essays are part of the work, but they are not the whole work. A quiet month on the blog can be the loudest month in the repositories. The site should make the current season legible without pretending every season is permanent.
Writing produced the map. Thinking clarified the layers. Coding is forcing every vague part of the thesis to become specific.
There will be another writing season. It will be better because this one happened first.