Plain-language summary.

A running system logs many internal numbers at each step.

This paper turns those numbers into a small set of coordinates, like a map. It then checks

whether the system’s regime labels match regions on that map, and whether regime switches

look like jumps. No meaning is read from generated text; only timing signals (input arrival

and “say” timestamps) are used.

Technical summary.

A tick-resolved window of a real-time “cognitive runtime” execution

(1k-node substrate) is analyzed using principal component analysis (PCA) on internal

numeric telemetry. In the analyzed window (t ∈ [359521, 385094], n = 16,746 ticks), the

first eight PCs explain 89.18% of standardized-feature variance. Regime-change ticks have

higher per-tick displacement in the first eight PCs, ∥∆PC1:8∥2 (mean 6.05 vs. 5.05; Cohen’s

d = 0.47; Mann–Whitney p = 2.9 × 10−157). A microstate Markov model (K=30, built

on the first eight PCs) exhibits multiple slow modes (second eigenvalue 0.9978, implied

timescale τ ≈ 455 steps), consistent with metastable organization. A “content influence”

control uses a lightweight input embedding (hashing + SVD) and finds a measurable increase

in next-step PC predictability (mean ∆R2 = 0.134 for predicting PC1:6(t+1)), while direct

next-step “say” prediction does not improve (AUC change −0.0055). All results are backed

by a self-contained data+code bundle with SHA256 indexing.

A running system logs many internal numbers at each step. This paper turns those numbers into a small set of coordinates, like a map. It then checks whether the system’s regime labels match regions on that map, and whether regime switches look like jumps. No meaning is read from generated text; only timing signals (input arrival and “say” timestamps) are used

A full length version on the Predictive Feature Architectures For Self-Supervised Say-Events in VDM audio podcast.

A cognitive runtime runs without training. It processes. Sometimes it speaks.

This paper figures out when and why. A single internal signal — a boundary pulse — predicts speech events with near-perfect accuracy. What the system hears doesn't matter. The gate opens from the inside. We discuss the results, the limits, and what it means for separating the decision to speak from the choice of words.

A self-rewiring network that flips modes when input appears.

This paper reports one VDM runtime run, analyzed with four independent checks. The runtime is a self-changing connectome, meaning the network rewires as it runs. The problem is simple: adaptive systems can look real while hiding measurement errors. So each claim must pass a gate, meaning a pass or fail test on logged data. The run shows recurring hub coalitions, meaning highly connected nodes reappear in similar groups. It also shows burst cascades, meaning bursts that come in many sizes. Most striking, the network sits in two distinct wiring styles, with a barrier score near 8.9 between them. Every switch into the sparse state occurred during input, and every switch back occurred without input.

Best for: complex-systems readers and builders who want measurable evidence from self-modifying networks.

What you can do with it: rerun the analysis pack, reproduce the four gates, and confirm which results hold.

What makes it trustworthy: pass/fail gates, shared tables and figures, and hashes, meaning file fingerprints, plus explicit listed limitations.

If you only remember one thing

It’s not a vibe report: it’s four independent, gate-checked signals from one self-rewiring run.

This first episode walks through a simple question: can a small, model-aware assist improve an “echo” without spending extra effort?

An echo is a forward run followed by a reverse run, where you try to land back on the start state. In this episode, I summarize a preregistered VDM experiment called Counterfactual Echo Gain (CEG). The point is not hype. It’s auditability: the measurement method is treated like an instrument, with pass/fail gates that must hold before we interpret outcomes.

You’ll hear:

- What the assist changes (and what it must not change)

- The gate checks (reversibility drift, monotonic dissipation checks, step-size accuracy, and equal-work matching)

- The outcome metric and what the result actually supports

If you want to verify it, find the record on Zenodo (includes the report, run ledger, telemetry, and figures).