Microcosm

[CONVICTION]

Microcosm is the inward-facing venture in the three-cosm ecosystem. Where Macrocosm builds interfaces to nature's distributed intelligence and the Mesocosm coordinates civilizational systems, Microcosm maps the inner landscape of human development -- a personal verification AI that senses state, maps the attractor landscape, reflects patterns, supports experimentation, and connects resonant people, while remaining invisible in daily life.

Not an app. Not an assistant. Not a digital twin. A protocol and intelligence layer that develops human capacity for creativity and authentic living. The same [V, G] framework that maps ecosystem attractors at planetary scale maps psychological attractors at personal scale. The same methodology -- attractor landscape mapping, morphogenetic intelligence, proof-carrying protocols -- applied inward.

The Dual Nature

[CONVICTION]

For the user: a lifestyle OS. Understand yourself, develop your practice, live well. The system maps your creative flow basin (deep, you settle there after morning exercise), your anxious rumination basin (deep, you fall there after poor sleep), and the perturbation map that shifts you between them. Over time, it reveals your coherent lifestyle -- discovered through experimentation, verified by state data.

For the network: a personal verification AI producing probabilistic proofs of genuine human development for the Mycel economy. Device-attested sensing, personalized models trained on hundreds of labeled sessions, longitudinal consistency across months, cross-layer integration across five kosha layers, and perturbation response testing. The user comes first. The network benefits second.

Household-native: designed for families from the ground up. One server, multiple people, family-level dynamics. A homeschool parent with Mac Minis and seven kids is the primary design target.

The Thesis: Attractor Landscapes for Human Development

[CONVICTION]

Michael Levin's work demonstrates that cells navigate an anatomical morphospace -- a landscape of possible body configurations -- toward target states encoded in bioelectric patterns. Microcosm applies the identical insight to human development: a person navigates toward psychological attractors -- habitual states, emotional patterns, creative modes, stuck loops. The topology of possible states IS the landscape. Development IS changing the landscape.

The formal model: dx/dt = -nabla D(x, x_target). Current state x moves down the gradient of distance from target. But the landscape has multiple attractors (stable basins), ridges (barriers), and saddle points (transition opportunities). Development is not moving toward one target but reshaping the entire landscape -- deepening the basins you want, shallowing the ones that trap you.

V (the landscape): the topology of your possible states. Changes slowly over months as capacities develop. The goal of Microcosm is to reshape V.

G (the metric tensor): current capacity to traverse V. Determined by sleep, allostatic load, energy, emotional state. When G is high, steep ridges are traversable. When G is low, you slide into the nearest deep basin. G fluctuates daily. V changes over months.

Perturbation map: what shifts you between basins. Morning exercise to creative flow. Poor sleep plus email before 9am to rumination. These are your personal levers. The experimentation engine discovers them empirically.

The Four Pillars

[CONVICTION]

Four capacities enabling creative, authentic living. Invisible to the user. Organizing everything underneath.

These map directly to the Sovereign Child framework's three foundational capacities (curiosity, agency, regulation) plus the meaning dimension that emerges through the ascent spectrum.

The Four Spaces

[CONVICTION]

Four postures for engaging with life. Relational context modulates all four.

Streams (Creation -- Para to Vaikhari): the full journey from impulse to manifestation. Voice-first capture, auto-classified to domains. The Vedantic tradition describes four levels of speech: para (pre-thought seed), pashyanti (first holistic glimpse), madhyama (structured thought), vaikhari (manifest expression). Most tools only handle vaikhari. Microcosm holds the full journey.

Mirror (Awareness -- Buddhi / The Witness): second-order observation. Morning question (one, adaptive). Evening reflection. Pattern observations (weekly). Four pillars as felt qualities, never scores. Experiment results.

Practice (Engagement -- Calendar of Attention): what you have done with your consciousness. Calendar, library, and dojo. Every intentional engagement is a session. Sessions happen in the real world; the system wraps minimally.

Tribe (Connection -- IRL Only): gatherings only. Resonance matching via stream embeddings and developmental proximity. No profiles, no DMs, no social graph. Privacy-preserving matching (PSI/ZK). Show up and connect as humans.

Design Invariants

[CONVICTION]

Eight named constraints that must hold at every level. Violations are bugs, not tradeoffs.

• SILENCE-BY-DEFAULT: the system is silent 95% of the time. No notifications. No engagement loops. Sensing is continuous; interaction is sparse and opt-in.
• UNCERTAINTY-FIRST: every state estimate, trigger, session outcome, and proof carries calibrated uncertainty, provenance, and validity bounds. The system never speaks with more confidence than it has.
• CONSENT-FIRST: every data access is governed by purpose-limited, auditable, revocable consent objects. Not toggles.
• NO-ESCALATION: if a trigger fires and the user ignores it, the system never escalates. Single offer, single haptic, done.
• PORTAL-ISOLATION: portals and agents are untrusted by default. Signed manifests, sandboxed execution, deny-by-default, scoped tokens, immutable audit.
• LANDSCAPE-NOT-SCORE: the system maps topology (basins, ridges, transitions), not metrics (scores, grades, levels). Users see felt qualities and trends, not numbers and rankings.
• PROPAGATION-AS-TRUTH: a state change that only appears in one kosha layer has not fully integrated. Genuine change propagates across layers.
• USER-FIRST-NETWORK-SECOND: the system serves the person's development first. Proof production, settlement, and network participation are always optional and require explicit consent.

The App: Creative Universe

[STORY]

For children, the app is a creative universe -- a dark, spatial, abstract landscape unique to each child. Regions represent domains explored (numbers, making, stories, music, nature). Explored regions are larger, brighter, more alive. Unexplored regions are dim, small, mysterious -- invitations, not deficiencies. The universe is the child's cognitive landscape made visible.

Sessions are the atomic unit. Everything the child does is a session: a math tutoring conversation, a YouTube video with AI co-viewing that pauses and questions, a nature walk with voice companion, a making project. Sessions arrive from parents, tutors, schools, or the system itself, drifting to the center of the universe as glowing presences.

The Active Learning Membrane wraps any passive content (YouTube, podcast, document) and turns it into an active session. The AI processes the content, identifies learning moments, pauses at key points, asks Socratic questions, listens to thinking, then resumes. Passive consumption becomes active cognition. The design inspiration is Rick Rubin's creative space: quiet, no urgency, no rewards, no gamification, presence not performance.

Age-Adaptive Interfaces

[EVIDENCE]

SOUL.md is the parent-authored governance document defining learning philosophy, allowed agents, content boundaries, time boundaries, and developmental priorities. The app enforces it silently.

What Does Not Exist Anywhere

[FRONTIER]

Six capabilities that no existing product provides:

1. Personalized verification AI for human development -- a model trained on YOUR labeled sessions that distinguishes genuine from performed state change for you specifically, producing probabilistic proofs with five-layer verification
2. Attractor landscape mapping -- a dynamical model of your state space running continuously, mapping basins, ridges, perturbation sensitivities, tracking how your landscape changes over years
3. Household state engine -- family-level state awareness treating the household as a dynamical system, with privacy preservation and family-level experiments
4. Practice-derived coherent lifestyle -- experimentation engine treating life as ongoing N-of-1 trials, converging on practices unique to each person, verified by state data
5. Proof-carrying human development protocol -- a Mycel domain where capacity development produces verifiable proofs with settlement weighting that incentivizes genuine long-term growth
6. State-aware agent layer -- bridge between interior awareness (Microcosm) and exterior action (OpenClaw), where agents know your state, your family's state, your capacity

The MorphoLife Extension

[FRONTIER]

The MorphoZero architecture -- originally built for robotics and chess -- instantiates for lifestyle as MorphoLife. The same losses that learn a chess landscape from game outcomes learn a lifestyle landscape from dojo sessions:

L_terminal: Dojo sessions have outcomes. A breath session either restores coherence (+1) or does not (-1). A Brih inquiry either produces insight (measurable via cross-layer propagation) or stays surface. Every session is a trajectory with a terminal label -- exactly like a chess game.

L_flow: The trajectory between dojo sessions is the person's daily life, read through boundary observables (HRV, breath, voice). For sessions that end in coherence, the gradient of V should align with the direction they were moving. For sessions that end in dysregulation, it should anti-align.

L_morse: The topology of V should have clean critical points. The creative flow basin should be a real attractor with a classifiable Hessian. The anxiety rumination loop should be a real attractor too. The saddle between them -- the ridge where one bad email tips you from creative flow into rumination -- should be a classifiable saddle point.

L_cross (new, lifestyle-specific): Genuine transformation propagates across kosha layers. A breath intervention that shifts HRV AND voice AND sleep gets y = +1 with high confidence. If it only shifted HRV, y = +0.5 (superficial change). This captures the Pancha Kosha insight that the state space has layers that sensor data alone cannot distinguish.

The LLM as Fifth Kosha Sensor. The LLM is not a translation layer bolted on top. It is a sensing and actuation channel for the upper koshas, equivalent to what the watch is for annamaya and the earbuds are for pranamaya. Hardware sensors reach annamaya (movement, temperature) and pranamaya (HRV, breath). The LLM reaches manomaya (emotional valence, topic-state correlations), vijnanamaya (cognitive patterns, insight, frame flexibility), and anandamaya (purpose, meaning, existential orientation). Without the LLM as sensor, V_task would be an annamaya-only surface -- what Whoop and Oura already approximate. The LLM co-constructs the landscape by integrating verbal self-report with sensor data to produce labels that V_task trains on.

Three LLM translation modes operate at different tiers:

Mode A -- Direct Actuation (no LLM). State engine to actuation agent to hardware. The intervention library is a lookup table. Covers breath entrainment, cardiac entrainment, sleep protocols, circadian nudges. Approximately 60-70% of all actuations. Fast, deterministic.

Mode B -- LLM as Hypothesis Generator. The state engine detects a pattern; the LLM generates the intervention hypothesis, designs the experiment protocol, and interprets results in actionable terms. The LLM has world knowledge the state engine does not. Fires intermittently -- maybe once a day, maybe once a week.

Mode C -- LLM as the Intervention Itself. Brih inquiry, Mirror reflections, emotional range mapping, relational facilitation. The conversation IS the perturbation to V. The LLM's words shift the person's state. This is vijnanamaya and manomaya actuation -- you write to the cognitive and emotional layers through meaning, not through earbuds.

The Household Extension. Multiple people on a shared landscape -- the same architecture as MorphoZero's embodiment transfer. Same V (the household's coherence topology), different G per person (each family member's capacity and constitution). Mom and child navigate the same household landscape through different geodesics.

See Also

• Microcosm Architecture -- the five kosha layers, sensing stack, AI integration
• Health -- the health domain that Microcosm serves
• Ascent Spectrum -- the developmental arc from regulation to awakening
• Sovereign Child -- the three foundational capacities
• Technology as Training Wheels -- the scaffolding-that-graduates thesis
• Macrocosm -- the same methodology applied outward
• Latent Human Capacities -- the evidence chain for what Microcosm aims to develop