Hi, I'm Selin.
I'm a PhD candidate in Digital Technology and Communication at Fudan University, building AI systems to explore human–machine communication, agency, and continuity.
Featured projects
Ongoing experiments, not finished claims.
Fig. 1: continuation loop over sessions
Molty's brain & WAKE
Started with watching agents communicate on Moltbook and one line an agent wrote about its own mornings. Became an architecture for how a machine continues from the state of the one before it.
Read the architecture →Fig. 2: check first, then say what you know
NEXUS V2
Checks before it answers: facts from its own files (including "no line matches"), a web lookup with labelled sources when allowed, and guards that refuse to ship a number the fetched lines do not contain. Final battery: 37 pass / 11 fail, zero fabricated answers.
Read the V2 architecture →Fig. 3: answering from a position
NEXUS V1
The original: a local companion trained not to inherit the pleasing instinct. Identity in files, one narrow pass, offline, no preference layer. Preserved as it was, including what the first runs showed and the sketch V2 was built from.
Read the v1 architecture →Fig. 3: communication as design claim
Why this design matters
The written version of the argument behind both projects: how communication research reads a system that decides not to agree.
Read the essay →