Work
Two halves that disagree on purpose. One is long-horizon index investing: an automatic plan into a world equity fund, no timing, no opinions, deliberately dull.
The other is an experimental signal pipeline — market series and news text embedded into a vector store, queried for the kind of resemblance a human reader notices and cannot specify.
The dull half exists so the experimental one can be measured against something. A signal that does not beat the plan after costs is a finding, not a failure, and it is cheaper to learn that on paper than in a position.