TS

Twishmay Shankar

Building formal frameworks at the intersection of statistical mechanics, active inference, and autonomous agent design. Exploring how possibility spaces unify physics, signals, and intelligence.

Research & Open Source

Possibility Space Theory

A 36-paper research program unifying statistical mechanics, active inference, and agent design through one formal language. 6 clusters spanning foundations, thermodynamics, signals, inference, AGI benchmarks, and applied systems.

36 papers 6 clusters 1,878 files LIVE

KHAOS / KPL

A model of computation for meaning-native programming. Meaning addresses, stored programs, and the AGI infrastructure thesis — the computational substrate underlying possibility spaces.

6 working papers formal language IN PROGRESS

Autonomous Research Pipeline

Training data infrastructure for teaching models to do autonomous research. Structured task traces, source-target pairs, and evaluation schemas for fine-tuning and RLHF.

JSON schemas training pipeline IN PROGRESS

Research Themes

possibility spaces ergodic theory free energy principle active inference Bayesian inference dynamical systems state space models information geometry phase transitions emergence entropy & complexity cellular automata formal languages digital twins world models variational methods Markov blankets thermodynamic computing ARC-AGI autonomous agents