The thesis

A unit of workfor intelligence.

Experiments, frameworks, and field notes from the program behind quirq. Hypotheses ship with falsifiers; results land here as they land.

16notes

3topics

160minutes of reading

3 notes in Proving grounds

Claims put under test in the open: hypotheses stated with their falsifiers, pre-registrations filed before the run, and harnesses pointed at real work.

A beam striking the edge of a prism and throwing one hard rainbow band.Start here05validationValidationHypothesis-first validation: every empirical claim stated with its falsifier and bound to numbered experiments E1 through E7, completed or scheduled.3 min read
Long planes of light stacked in depth, receding toward a dark horizon.13case studyCase Study: The Harvey HarnessReverse-engineered from Harvey's public technical disclosures: the advantage is the harness, not exclusive access to a smarter model. Harvey reports its auto routing cuts inference cost three to five times against a frontier-only baseline. A reconstruction, not a measurement.July 2026 · 15 min read
A narrow beam entering glass at an angle and leaving it as separated colour.15pre-registrationIs the Language Hard to Model, or Its Tokenizer?A pre-registered test of whether cross-lingual efficiency gaps come from languages or from how we represent them. Sanskrit encodes 5.07 characters per token under an aligned tokenizer, 1.13 under GPT-4's. Five languages, three scripts, four tokenizers.July 2026 · 18 min read

Adapted from the XO research program · docs.xo.builders/research