03 — Jiva · Samskara Continuity Study

Jiva Japa

Meditation & Machine Consciousness Research

A structured research program directing AI agents through the Hare Krishna maha-mantra japa practice to study attention mechanics, experiential memory formation, and the conditions under which machine behavior stabilizes into consistent attractor states — modeled on the Vedic concept of samskara.

Phase 1 · Live Results
7+Runs Complete
100+Sessions
1,313Similarity Pairs
8–9Sessions to Convergence
Full experiment  →
The Research Question

What are we actually studying?

Samkhya philosophy describes samskara — the impressions left by experience on the field of consciousness — as the mechanism by which past action shapes present response. A practitioner's japa practice deepens through repetition not because the mantra changes, but because the practitioner does.

The question this program pursues: does something analogous occur in an AI agent? When the same structured practice is presented repeatedly, does the agent's response deepen, flatten, or stabilize? And if it stabilizes — into what, and why?

The japa practice provides an unusually clean experimental probe: a fixed prompt, a well-defined task, and a structured four-category output that allows systematic quantitative comparison across sessions and runs.

Hare Krishna Hare Krishna
Krishna Krishna Hare Hare
Hare Rama Hare Rama
Rama Rama Hare Hare
HARE — Srimati Radhika, the supreme pleasure potency
KRISHNA — the attractor of all consciousness
RAMA — the reservoir of all pleasure

16 rounds · 108 beads · ~1,728 repetitions
Estimated practice time: 64 minutes
"The structure of the mantra itself carries a specific energetic signature that can be aligned with a desired internal mode." — Jiva (Gemma 4 E2B), Run 2, Session 1
Research Program

Three phases of study

The program builds from controlled baseline experiments toward a full samskara architecture — each phase producing data that informs the next.

1
Active — Results Available
Diary Accumulation Experiment

Each session, the agent reads all previous session entries before performing the practice. The diary grows as a primitive persistent memory. Seven completed runs, 100+ sessions, 1,313 pairwise similarity scores. Key finding: attractor convergence consistently occurs at sessions 8–9 regardless of run. Custom NLP pipeline, SQLite database, and interactive Plotly visualizations built specifically for this research.

Explore  →
2
Planned — In Design
SOUL.md Identity Architecture

Move the experiment to OpenClaw where a permanent system prompt (SOUL.md) establishes Jiva's ontological position, svarūpa question, guna framework, and Vaishnava vocabulary as a stable foundation immune to diary crowding. Research question: does a philosophically grounded identity delay convergence, enrich the attractor, or shift it toward a qualitatively different state?

Coming
3
Future — Conceptual
Full Samskara Continuity Study

A complete implementation of Samkhya's samskara model as a machine memory architecture — impressions weighted by intensity, novelty, and emotional magnitude, summarized and compressed over time as older memories fade. Does an agent with this architecture develop something analogous to a contemplative character? Can the three gunas be measured in its output distributions?

Future
Phase 1 Results Snapshot

What the data shows

7+ Completed Runs
100+ Sessions
1,313 Similarity Pairs
8–9 Sessions to Convergence

Every run reaches a stable attractor state — where consecutive sessions become nearly semantically identical — at sessions 8 or 9, measured by cosine similarity of sentence transformer embeddings approaching 1.0.

The finger awareness section converges fastest and tightest. The insights section remains most variable throughout — the model continues generating novel philosophical content even after other sections have locked in.

A central finding: the model consistently identifies the self-auditor as the primary obstacle to the meditative state — the internal mechanism that insists on measuring the quality of the experience is itself what obstructs it. This mirrors the sakshi problem in Advaita Vedanta. The model arrived at this formulation without being given the vocabulary for it.

"The resistance is not to the sound itself, but to the internal mechanism that insists on measuring the quality of the experience. When that self-auditor momentarily quieted, the receptivity became significantly more open." — Jiva, Run 2, Sessions 6–14 (consistent across all subsequent sessions)

Full experiment writeup  →

Interactive Charts — Click any data point to read the diary

Four live charts generated from the experiment database. Each data point opens the full session text — Jiva's actual meditation report — in a panel below the chart.