AIEN Research
Preprints, experiments, and public evidence from the AIEN project. The central thread: the Turing (T), a protocol-bound measure of net held-out explanatory compression — one measurable component of understanding, with Turing yield (T/J) normalizing it by energy. All preprints; none peer reviewed yet.
Crack the code, run the Rule Lab, duel the memorizer — and see the first real measurement, the understanding-versus-search research direction, and why nature invented the loop.
The reference document: Turing gain defined (1 T = 1 bit of net held-out description-length improvement against a declared baseline), assumptions, measurement protocol with sealed data and recomputable evidence, related work, and falsification criteria. First canonical measurement recorded 2026-09-29: TY-1+TY-2 PASS, +2,559,679.825 T under profile V0. Preregistered experiment ladder EXP-001 through EXP-003 still pending.
The engineering history the measure emerged from: AIEN, Omega, PHYSICS, FORGE, AEGIS, ARGUS, Cortex, and the resident reaction architecture — with the public evidence receipts (R13, R14, R15, TURING W0) and the first experimental proposals.
The philosophy behind the architecture, stated as a position paper. It names and answers the claim “English is a better programming language than Ruby” (David Heinemeier Hansson, Rails World 2026): English is the conversation, not the language, and the universal artifact is the verified program. Weights suggest, programs explain, Omega verifies, physics authorizes, evidence teaches. The stack is stripped layer by layer to a foundation owned outright, because whoever owns the interpreter decides what your words mean.
A companion concept paper. It proposes that integration peaks inside a bounded band, the candidate integration zone (CIZ), defines the quantities Gamma, R, D, P, and H, and lays out experiments CIZ-000 through CIZ-011 to test the idea. A hypothesis with named experiments, not a result.
EXP-002 of the experiment ladder: a hidden evaluator generates quantized trajectories from an undisclosed stochastic law — pure diffusion, diffusion with drift, mean reversion. The candidate must earn complexity on sealed held-out trajectories without inventing structure inside noise. No profile is frozen and no measurement may be reported until one is.