01 / BACKGROUND
Systems first
My path into this work was not academic in the conventional sense. It ran through UNIX systems administration, Solaris and UltraSPARC machines, hosting infrastructure, command-line systems, C and C++, and self-directed experimental work.
That background still shapes how I approach AI. I want to know where the state lives, which component has authority, what the actual control flow is, how the system fails, and what evidence survives after a run.
02 / OPERATING LOOP
The loop is simple
hypothesis → smallest executable experiment → instrumentation → run → unexpected behavior → revised hypothesis → repeat
I prefer executable artifacts over decks, measurements over rhetorical certainty, and local or inexpensive tests before GPU-scale commitments. I am happy to entertain a ridiculous-sounding analogy if it can eventually be made to lose.
03 / RESEARCH VALUES
What stays true
- Weird hypotheses are allowed.
- Evidence beats confidence.
- Negative results are useful.
- Observation and interpretation stay separate.
- Analogy is not mechanism.
- Publish enough detail that another person or agent can tell what was actually tested.
04 / OUTSIDE THE LAB
The person in the system
I make music as Colors Conflicted, with a recurring goal of making the meltiest, swirliest music possible. Music is also one of the places I borrow technical questions from: prediction, salience, timing, learned expectation, representation, and why some transitions matter much more than others.
I also care about games and simulation, visual storytelling, robotics, Japanese, martial systems, physics, and any domain that exposes an interesting rule system.