I’m building kn0wthyself to give machines usable knowledge of their own capabilities, condition, dependencies, purpose, and limits.

That knowledge matters when a system chooses an action, adapts to changed conditions, or works with another system. An agent (wither it's agentic software, or autonomous hardware like a drone) needs to establish what a potential collaborator can actually do. A system considering a change to itself needs to account for what depends on the part it is changing, and how the result will be checked.

At kn0wthyself, we’re developing s0ma, a language and runtime that make self-knowledge part of how systems are built and operated. People and AI can consult a shared model of the system they’re working with. From here, we will move on to more efficient languages, models, and hardware.

My longer-term focus is how this foundation can support better performance, self-adaptation, and agents finding suitable collaborators for a task. That includes representing uncertainty, respecting the boundaries people set, and checking whether actions achieve their intended purpose.

People should be able to understand, redirect, correct, and stop systems acting on their behalf. I want those abilities built into increasingly capable machines.

I welcome conversations with builders and researchers working on these problems.