Intelligence Learns How to Find Out
Perception, action, and the making of evidence.
Plato Is Not a Space
A research note on predictive interfaces, event structure, and intervention-facing probes, rather than a single universal embedding space.
Where Do Good Vision Targets Come From?
Notes in the margins of James Chen's pure-vision scaling post, written after the CVPR Bitter Lessons workshop. The search for a vision scaling law is not a search through losses; it is a search through targets, and where their invariance comes from.
A Platonic — and Umwelt — reading of color bias in VLMs
An interpretive companion to “Seeing Red, Thinking Bad”: why an inert surface feature—text color—moves a vision language model's semantic judgment, read through the debate on representational convergence.